Merge pull request #84 from yusufipk/word-timestamps-for-models-without-segments

Build subtitle cues out of word times where a model marks no segments

.github/workflows/whisper-vulkan.yml

@@ -0,0 +1,222 @@
+name: whisper.cpp Vulkan bundle
+
+# Only what the bundle is built from. Compiling the Vulkan shaders takes
+# tens of minutes, and a README typo is not worth one: what ties ggml.py to
+# this release is a handful of assertions in tests/test_packaging.py, and
+# those run on every pull request in milliseconds.
+on:
+  pull_request:
+    paths:
+      - packaging/whisper-vulkan/**
+      - .github/workflows/whisper-vulkan.yml
+  workflow_dispatch:
+    inputs:
+      whisper_version:
+        description: Upstream whisper.cpp version (without v)
+        required: true
+        default: "1.9.3"
+        type: string
+      whisper_commit:
+        description: Peeled commit SHA for that reviewed upstream tag
+        required: true
+        default: "371b5a7561823ab2bb32142d2751e35e7534727b"
+        type: string
+      expected_sha256:
+        description: >-
+          Reviewed archive SHA-256. Leave empty for a version this file has
+          not reviewed: the digest of what was built is reported instead of
+          being checked, and publishing is refused.
+        required: false
+        default: ""
+        type: string
+      publish:
+        description: Publish a Dikte dependency release
+        required: true
+        default: false
+        type: boolean
+
+permissions:
+  contents: read
+
+concurrency:
+  group: whisper-vulkan-${{ github.event.pull_request.number || github.ref }}
+  cancel-in-progress: ${{ github.event_name == 'pull_request' }}
+
+env:
+  WHISPER_VERSION: ${{ inputs.whisper_version || '1.9.3' }}
+  WHISPER_COMMIT: ${{ inputs.whisper_commit || '371b5a7561823ab2bb32142d2751e35e7534727b' }}
+  REVIEWED_WHISPER_VERSION: "1.9.3"
+  REVIEWED_WHISPER_SHA256: c25ca76504144da488eb74441390a7b9aa7ce547e5f2f391cbd831253c9b54d8
+
+jobs:
+  build:
+    runs-on: ubuntu-22.04
+    timeout-minutes: 45
+    steps:
+      - name: Check out Dikte
+        uses: actions/checkout@11d5960a326750d5838078e36cf38b85af677262 # v4
+        with:
+          persist-credentials: false
+
+      - name: Validate source coordinates
+        shell: bash
+        run: |
+          set -euo pipefail
+          [[ "$WHISPER_VERSION" =~ ^[0-9]+\.[0-9]+\.[0-9]+$ ]]
+          [[ "$WHISPER_COMMIT" =~ ^[0-9a-f]{40}$ ]]
+
+      - name: Check out pinned whisper.cpp source
+        uses: actions/checkout@11d5960a326750d5838078e36cf38b85af677262 # v4
+        with:
+          repository: ggml-org/whisper.cpp
+          ref: ${{ env.WHISPER_COMMIT }}
+          path: vendor/whisper.cpp
+          fetch-depth: 0
+          persist-credentials: false
+
+      - name: Verify source version and commit
+        shell: bash
+        run: |
+          set -euo pipefail
+          test "$(git -C vendor/whisper.cpp rev-parse HEAD)" = "$WHISPER_COMMIT"
+          git -C vendor/whisper.cpp fetch --depth=1 origin \
+            "refs/tags/v$WHISPER_VERSION:refs/tags/v$WHISPER_VERSION"
+          test "$(git -C vendor/whisper.cpp rev-list -n1 "v$WHISPER_VERSION")" = "$WHISPER_COMMIT"
+          echo "SOURCE_DATE_EPOCH=$(git -C vendor/whisper.cpp show -s --format=%ct HEAD)" >> "$GITHUB_ENV"
+
+      - name: Build pinned build environment
+        run: docker build --pull=false -f packaging/whisper-vulkan/Dockerfile.build -t dikte-whisper-builder packaging/whisper-vulkan
+
+      - name: Build deterministic archive
+        run: |
+          docker run --rm \
+            -e WHISPER_VERSION -e WHISPER_COMMIT -e SOURCE_DATE_EPOCH \
+            -v "$PWD/vendor/whisper.cpp:/src:ro" \
+            -v "$PWD/packaging/whisper-vulkan:/packaging:ro" \
+            -v "$PWD/work:/work" \
+            dikte-whisper-builder \
+            bash /packaging/build-package.sh
+          mkdir -p dist
+          cp work/out/whisper-bin-ubuntu-vulkan-x64.* dist/
+
+      - name: Verify reviewed archive digest
+        shell: bash
+        env:
+          EXPECTED_SHA256: ${{ inputs.expected_sha256 }}
+          PUBLISH: ${{ inputs.publish }}
+        run: |
+          set -euo pipefail
+          read -r actual _ < dist/whisper-bin-ubuntu-vulkan-x64.tar.gz.sha256
+          echo "built archive sha256: $actual"
+          expected="$EXPECTED_SHA256"
+          if [ -z "$expected" ] \
+              && [ "$WHISPER_VERSION" = "$REVIEWED_WHISPER_VERSION" ]; then
+            expected="$REVIEWED_WHISPER_SHA256"
+          fi
+          if [ -z "$expected" ]; then
+            # The digest of a version nobody has reviewed yet cannot be known
+            # before it is built. Reporting it is the whole point of the run;
+            # a release out of it is not.
+            if [ "${PUBLISH:-false}" = true ]; then
+              echo "refusing to publish an archive whose digest has not been reviewed" >&2
+              exit 1
+            fi
+            echo "::notice::no reviewed digest for $WHISPER_VERSION." \
+                 "Review the one above, then dispatch again with expected_sha256."
+            exit 0
+          fi
+          test "$actual" = "$expected"
+
+      - name: Validate archive and ELF contract
+        run: OUT_DIR=dist packaging/whisper-vulkan/validate-package.sh
+
+      - name: Schema-validate CycloneDX 1.6 SBOM
+        run: |
+          docker run --rm \
+            -v "$PWD/dist/whisper-bin-ubuntu-vulkan-x64.cdx.json:/sbom.json:ro" \
+            cyclonedx/cyclonedx-cli@sha256:252c2e26f468c25fea1e63ecde1bc3198ad6e9dbb57f5ed3236bddcb2281b3a7 \
+            validate --input-file /sbom.json --input-format json \
+            --input-version v1_6 --fail-on-errors
+
+      - name: CPU fallback smoke test (no Vulkan loader)
+        run: OUT_DIR=dist packaging/whisper-vulkan/smoke-runtime.sh cpu
+
+      - name: Vulkan loader present, no device smoke test
+        run: OUT_DIR=dist packaging/whisper-vulkan/smoke-runtime.sh noicd
+
+      - name: Vulkan plugin-load smoke test (Mesa llvmpipe)
+        run: OUT_DIR=dist packaging/whisper-vulkan/smoke-runtime.sh vulkan
+
+      - name: Upload reviewed outputs
+        uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02 # v4
+        with:
+          name: whisper-bin-ubuntu-vulkan-x64
+          path: dist/*
+          if-no-files-found: error
+          retention-days: 14
+
+  publish:
+    if: >-
+      github.event_name == 'workflow_dispatch' && inputs.publish &&
+      github.ref == 'refs/heads/master'
+    needs: build
+    runs-on: ubuntu-22.04
+    environment: dependency-release
+    permissions:
+      contents: write
+      id-token: write
+      attestations: write
+      artifact-metadata: write
+    steps:
+      - name: Download the exact tested outputs
+        uses: actions/download-artifact@d3f86a106a0bac45b974a628896c90dbdf5c8093 # v4
+        with:
+          name: whisper-bin-ubuntu-vulkan-x64
+          path: dist
+
+      - name: Verify digest sidecar
+        run: (cd dist && sha256sum --check whisper-bin-ubuntu-vulkan-x64.tar.gz.sha256)
+
+      - name: Attest build provenance
+        uses: actions/attest@1e69f48acb82d1966a394da916b4c1698aa569d6 # v4
+        with:
+          subject-path: dist/whisper-bin-ubuntu-vulkan-x64.tar.gz
+
+      - name: Attest SBOM to archive
+        uses: actions/attest@1e69f48acb82d1966a394da916b4c1698aa569d6 # v4
+        with:
+          subject-path: dist/whisper-bin-ubuntu-vulkan-x64.tar.gz
+          sbom-path: dist/whisper-bin-ubuntu-vulkan-x64.cdx.json
+
+      - name: Publish dependency release
+        env:
+          GH_TOKEN: ${{ github.token }}
+          RELEASE_TAG: whisper.cpp-v${{ inputs.whisper_version }}
+          RELEASE_TITLE: whisper.cpp v${{ inputs.whisper_version }} Vulkan bundle
+          RELEASE_NOTES: >-
+            Pinned source: ggml-org/whisper.cpp@${{ inputs.whisper_commit }}.
+            Verify with: gh attestation verify
+            whisper-bin-ubuntu-vulkan-x64.tar.gz
+            --repo ${{ github.repository }}
+        run: |
+          set -euo pipefail
+          if gh release view "$RELEASE_TAG" --repo "$GITHUB_REPOSITORY" >/dev/null 2>&1; then
+            echo "refusing to replace existing release $RELEASE_TAG" >&2
+            exit 1
+          fi
+          if gh api "repos/$GITHUB_REPOSITORY/git/ref/tags/$RELEASE_TAG" >/dev/null 2>&1; then
+            echo "refusing to replace existing tag $RELEASE_TAG" >&2
+            exit 1
+          fi
+          gh api --method POST "repos/$GITHUB_REPOSITORY/git/refs" \
+            -f ref="refs/tags/$RELEASE_TAG" \
+            -f sha="$GITHUB_SHA" >/dev/null
+          test "$(gh api "repos/$GITHUB_REPOSITORY/git/ref/tags/$RELEASE_TAG" \
+            --jq .object.sha)" = "$GITHUB_SHA"
+          gh release create "$RELEASE_TAG" dist/* \
+            --repo "$GITHUB_REPOSITORY" \
+            --verify-tag \
+            --prerelease \
+            --latest=false \
+            --title "$RELEASE_TITLE" \
+            --notes "$RELEASE_NOTES"

README.md

@@ -161,9 +161,14 @@ running.
 - **It all runs on this machine by default.** Speech to text on whisper.cpp and
   cleanup on llama.cpp, neither installed beforehand: the settings window fetches
   the program and the model, verifies the sha256 and refuses a download published
-  without one, then keeps a server alive while you dictate. The graphics card is
+  without one, then keeps a server alive while you dictate and hands the memory
+  back once it has sat unused for ten minutes. The model list is
+  grouped by model rather than by file size, and the row this machine's memory
+  and graphics can take is marked. The graphics card is
   reached through CUDA, ROCm or Vulkan where the build allows. No key, no
-  account, nothing leaving the machine.
+  account, nothing leaving the machine. On x86_64 Linux the same button fetches
+  a Vulkan build of whisper-server that Dikte publishes itself, because
+  upstream's Linux archive is processor-only.
 - **Silence never reaches the API.** Handed near-silence, a transcription model
   invents a sentence instead of returning nothing ("Thanks for watching", or in
   Turkish "Altyazı M.K."). A recording is dropped when nothing rose 10 dB above

README.tr.md

@@ -158,8 +158,13 @@ olmasını ister.
   whisper.cpp, temizleme llama.cpp üzerinde; ikisini de önceden kurman gerekmez:
   ayarlar penceresi programı ve modeli indirir, sha256'sını doğrular,
   checksum'suz yayınlanmış bir indirmeyi reddeder, sen dikte ettikçe sunucuyu
-  ayakta tutar. Derleme destekliyorsa ekran kartına CUDA, ROCm ya da Vulkan
+  ayakta tutar ve on dakika kullanılmayan modelin belleğini geri verir. Model
+  listesi dosya boyutuna değil modele göre gruplanır ve bu
+  makinenin belleğine ve ekran kartına uyan satır işaretlenir. Derleme
+  destekliyorsa ekran kartına CUDA, ROCm ya da Vulkan
   üzerinden ulaşılır. Anahtar yok, hesap yok, makineden çıkan bir şey yok.
+  x86_64 Linux'ta aynı düğme, whisper-server'ın Dikte'nin kendi yayınladığı
+  Vulkan derlemesini indirir; upstream'in Linux arşivi yalnızca işlemci için.
 - **Sessizlik API'ye gitmez.** Sessize yakın bir ses verildiğinde model boş dize
   döndürmez, bir cümle uydurur ("Altyazı M.K.", "Thanks for watching"). *O
   kaydın kendi* gürültü tabanının 10 dB üstüne en az 0,3 saniye çıkan bir şey

dikte/__init__.py

@@ -10,4 +10,4 @@ business loading Qt to answer one question.
 # both the .dmg's Info.plist and the AppImage's file name are built from it. A
 # build off master rather than off a tag appends the commit to it, so that a
 # bug report from someone running "latest" names a commit.
-__version__ = "1.0.2"
+__version__ = "1.2.0"

dikte/api.py

@@ -48,22 +48,33 @@ LOCAL_TIMEOUT = 3600
 
 # Where a transcription request goes; built by config.Config.transcribe_target().
 # `service` is the name the user sees in an error, `provider` the one the code
-# branches on.
-Target = collections.namedtuple("Target", "provider service api_key base_url model")
+# branches on. `file_model` is what a timestamped run asks for instead of
+# `model`, where the two differ; empty means the provider's own whisper.
+Target = collections.namedtuple(
+    "Target", "provider service api_key base_url model file_model",
+    defaults=[""])
+
+# What answers with segment times on OpenRouter when nothing else was chosen.
+OPENROUTER_FILE_MODEL = "openai/whisper-1"
 
 
-def timestamp_model(provider, selected=""):
+def timestamp_model(provider, selected="", file_model=""):
     """Which model answers with segment times.
 
-    OpenAI keeps them to whisper-1 and OpenRouter namespaces that id. Everything
-    Groq transcribes with is a whisper, so the model already chosen does it and
-    the fallback is only for a provider left on its default. So is everything the
-    local server runs, whatever the file is called, and there asking for another
-    model would name one it has never heard of.
+    OpenAI keeps them to whisper-1. Everything Groq transcribes with is a
+    whisper, so the model already chosen does it and the fallback is only for a
+    provider left on its default. So is everything the local server runs,
+    whatever the file is called, and there asking for another model would name
+    one it has never heard of. OpenRouter fronts several models that do times
+    and several that do not, and a request to the wrong one gets a transcript
+    with no segments in it, so the one to use is a setting of its own
+    (`file_model`) and whisper-1 is only where that setting is left empty.
     """
     if provider in ("groq", "local"):
         return selected or "whisper-large-v3-turbo"
-    return "openai/whisper-1" if provider == "openrouter" else "whisper-1"
+    if provider == "openrouter":
+        return file_model or OPENROUTER_FILE_MODEL
+    return "whisper-1"
 
 
 # What a gateway in front of the model answers of its own accord: the request
@@ -373,15 +384,20 @@ def _transcribe_request(target, audio_path, language, prompt, response_format,
     # takes it as the initial prompt, the way OpenAI does.
     if prompt and target.provider != "openrouter":
         fields.append(("prompt", prompt))
-    if granularity:
-        fields.append(("timestamp_granularities[]", granularity))
+    for level in granularity or ():
+        fields.append(("timestamp_granularities[]", level))
     body, ctype = _multipart(fields, "file", audio_path)
+    # An hour of meeting takes the local server a while, and the idle unload has
+    # to count that as the model being used rather than as nobody wanting it.
+    held = (ggml.whisper.busy() if target.provider == "local"
+            else contextlib.nullcontext())
     try:
-        return _request(
-            f"{target.base_url.rstrip('/')}/audio/transcriptions", body,
-            _headers(target.provider, target.api_key, ctype), timeout=timeout,
-            aborter=aborter,
-        )
+        with held:
+            return _request(
+                f"{target.base_url.rstrip('/')}/audio/transcriptions", body,
+                _headers(target.provider, target.api_key, ctype), timeout=timeout,
+                aborter=aborter,
+            )
     except ApiError as exc:
         if target.provider == "local":
             raise local_failure(target.service, ggml.whisper, exc) from None
@@ -427,6 +443,96 @@ def _merge_word_splits(segments):
     return merged
 
 
+# A cue built here is one a reader has time for: about two lines of subtitle,
+# and no longer on screen than a sentence takes to say. Neither is a hard rule
+# for a sentence that ends early, only the point past which one is broken.
+MAX_CUE_SECONDS = 7.0
+MAX_CUE_CHARS = 84
+# The other end of it: a cue nobody can read because it was gone before they
+# looked. A full stop this early in a cue is not the end of anything worth
+# breaking on, which is what "1." and "Dr." are, and a cue that ends up short
+# anyway is held on screen until the next one needs the space.
+MIN_CUE_SECONDS = 1.2
+# No whisper segment is longer than the window it was heard in, so a segment
+# that runs past this came from a model that is not marking segments at all.
+WHISPER_WINDOW = 30.0
+SENTENCE_END = ".!?…"
+
+
+def _too_coarse(segments):
+    """Whether these segments are too long to be cues, or are not there at all.
+
+    Not every model behind /audio/transcriptions marks segments the way whisper
+    does. Some fill the field with one entry per paragraph, or with a single one
+    covering the whole file, which turns a fourteen minute video into three
+    subtitles. Word times are what those models do give, and cues built from
+    them are better than what the segments would have been.
+    """
+    if not segments:
+        return True
+    return any(float(seg.get("end") or 0.0) - float(seg.get("start") or 0.0)
+               > WHISPER_WINDOW for seg in segments)
+
+
+def cues_from_words(words):
+    """[(start, end, text)] cut out of word times, where segments were no use.
+
+    A cue ends where a sentence does, and failing that wherever it has grown too
+    long to read or too long to leave up. Nothing is ever cut between two words:
+    the times that arrive are per word, and so are the ones that leave.
+    """
+    cues = []
+    start = end = 0.0
+    current = []
+
+    def flush():
+        nonlocal current
+        if current:
+            cues.append((start, max(end, start), " ".join(current)))
+            current = []
+
+    for word in words:
+        text = (word.get("word") or "").strip()
+        if not text:
+            continue
+        at = float(word.get("start") or 0.0)
+        until = float(word.get("end") or at)
+        if current:
+            grown = len(" ".join(current)) + 1 + len(text)
+            if grown > MAX_CUE_CHARS or until - start > MAX_CUE_SECONDS:
+                flush()
+        if not current:
+            start = at
+        current.append(text)
+        end = until
+        # A sentence can end inside the punctuation that closes a quote. What
+        # is too short to have been a sentence is a list marker or a shortened
+        # word, and the cue goes on rather than ending on it.
+        if (end - start >= MIN_CUE_SECONDS
+                and text.rstrip("\"')]»”’").endswith(tuple(SENTENCE_END))):
+            flush()
+    flush()
+    return _held(cues)
+
+
+def _held(cues):
+    """Keep a cue that is still too short on screen, without covering the next.
+
+    A one word sentence is a fifth of a second of audio and so a fifth of a
+    second of subtitle, which is a flicker. It stays up until the cue after it
+    starts, or for as long as it takes to read, whichever comes first.
+    """
+    out = []
+    for index, (start, end, text) in enumerate(cues):
+        if end - start < MIN_CUE_SECONDS:
+            room = start + MIN_CUE_SECONDS
+            if index + 1 < len(cues):
+                room = min(room, cues[index + 1][0])
+            end = max(end, room)
+        out.append((start, end, text))
+    return out
+
+
 def transcribe(target, audio_path, language="", prompt="", timeout=300, aborter=None):
     data = _transcribe_request(
         target, audio_path, language, prompt, "json", timeout=timeout, aborter=aborter
@@ -443,14 +549,34 @@ def transcribe(target, audio_path, language="", prompt="", timeout=300, aborter=
 def transcribe_segments(target, audio_path, language="", prompt="", timeout=300,
                         aborter=None):
     """[(start_seconds, end_seconds, text)] using whisper-1's verbose response."""
-    data = _transcribe_request(
-        target._replace(model=timestamp_model(target.provider, target.model)),
-        audio_path, language, prompt, "verbose_json",
-        granularity="segment", timeout=timeout, aborter=aborter,
-    )
+    target = target._replace(model=timestamp_model(target.provider, target.model,
+                                                   target.file_model))
+    ask = dict(language=language, prompt=prompt, response_format="verbose_json",
+               timeout=timeout, aborter=aborter)
+    # Word times are the way out of a model that does not mark segments, and
+    # whisper.cpp is not one of those, so the local server is only ever asked
+    # for what it has always been asked for. A hosted model that refuses the
+    # field says so with a 400, and the request it used to answer is still
+    # there to fall back on rather than losing the run over a field it did not
+    # need in the first place.
+    if target.provider == "local":
+        data = _transcribe_request(target, audio_path, granularity=("segment",), **ask)
+    else:
+        try:
+            data = _transcribe_request(target, audio_path,
+                                       granularity=("segment", "word"), **ask)
+        except ApiError as exc:
+            if exc.status != 400:
+                raise
+            data = _transcribe_request(target, audio_path,
+                                       granularity=("segment",), **ask)
     segments = data.get("segments") or []
     if target.provider == "local":
         segments = _merge_word_splits(segments)
+    if _too_coarse(segments):
+        cues = cues_from_words(data.get("words") or [])
+        if cues:
+            return cues
     out = []
     for seg in segments:
         text = (seg.get("text") or "").strip()
@@ -506,22 +632,65 @@ def _thinking(payload, provider, reasoning):
         payload["reasoning"] = {"effort": reasoning, "exclude": True}
 
 
-def local_ceiling(text):
+# Room for the thinking on this machine, one budget per rung of the settings
+# ladder. llama.cpp counts the thinking towards max_tokens along with the answer
+# it precedes, so a ceiling sized for the answer alone leaves a model that
+# thinks nothing to answer with. The rungs double, starting where a small model
+# lands when it barely thinks at all: cleanup is punctuation, and locally every
+# one of these tokens is also a second of somebody standing in front of the
+# screen, so the low rungs are the ones meant to be used.
+THINKING_ROOM = {
+    "minimal": 256, "low": 512, "medium": 1024,
+    "high": 2048, "xhigh": 4096, "max": 8192,
+}
+# An empty setting leaves it to the model, and the templates that can think
+# think by default. Room for a middling amount of it, since there is no way to
+# ask which kind of model this is.
+DEFAULT_THINKING_ROOM = THINKING_ROOM["medium"]
+
+
+def local_ceiling(text, reasoning="", context=0, prompt=""):
     """How much of a reply is worth waiting for from a model on this machine.
 
     Cleanup gives back what it was given, near enough, so a reply several times
     the length of the transcript is a model that has lost the thread rather than
     one doing the job. A small one will happily repeat the transcript until the
     context is full, and every one of those tokens is a second of somebody
-    waiting. A hosted model is left alone: there the same runaway is rare, and a
-    ceiling would cut the minutes short instead.
+    waiting, with only the hour-long local timeout underneath. A hosted model is
+    left alone: there the same runaway is rare, and a ceiling would cut the
+    minutes short instead.
+
+    The answer's share is the transcript's length in characters spent as a
+    budget in tokens, so what it really allows is two to four times the
+    transcript depending on how well the language tokenises. Turkish sits at the
+    tight end of that and still has room to spare for a reply that is meant to
+    come back the same length it went in.
+
+    Thinking is added on top of that share rather than taken out of it. Sharing
+    one budget is what makes turning thinking up quietly cost the answer, and on
+    a short dictation the 512 floor is the whole budget, so the answer is what
+    goes missing first.
+
+    `context` is what the server was started with, and the whole of it is the
+    real limit whatever is asked for here: a ceiling above it is not a ceiling,
+    because the runaway it exists to stop would run to the end of the context
+    instead. So the ceiling is held below what the prompt leaves. Two characters
+    to the token is under any tokeniser's rate for natural language, Turkish
+    included, which makes the reserve an over-estimate rather than a promise of
+    room that is not there.
     """
-    return max(512, len(text))
+    answer = max(512, len(text))
+    if reasoning != "none":
+        answer += THINKING_ROOM.get(reasoning, DEFAULT_THINKING_ROOM)
+    context = int(context or 0)
+    if not context:
+        return answer
+    return max(256, min(answer, context - (len(prompt) + len(text)) // 2))
 
 
 def cleanup(text, api_key, model, system_prompt, reasoning="",
             base_url=OPENROUTER_URL, timeout=180, provider="openrouter",
-            service="OpenRouter", aborter=None):
+            service="OpenRouter", aborter=None, context=0):
     if not api_key and provider != "local-llm":
         raise ApiError(t("{service} API key is empty. Add it in Settings.",
                          service=service))
@@ -534,7 +703,8 @@ def cleanup(text, api_key, model, system_prompt, reasoning="",
         ],
     }
     if provider == "local-llm":
-        payload["max_tokens"] = local_ceiling(text)
+        payload["max_tokens"] = local_ceiling(text, reasoning, context,
+                                              system_prompt)
     _thinking(payload, provider, reasoning)
     try:
         data = _request(
@@ -558,6 +728,13 @@ def cleanup(text, api_key, model, system_prompt, reasoning="",
             raise ApiError(t("The cleanup model spent its whole reply on "
                              "thinking. Set Thinking to \u201cOff\u201d."))
         raise ApiError(t("The cleanup model returned an empty reply."))
+    if choices[0].get("finish_reason") == "length":
+        # Cut off at somebody's ceiling: ours locally, the provider's otherwise.
+        # What came back is a sentence that stops mid-word, and cleanup is meant
+        # to hand back the whole dictation, so the half is refused rather than
+        # returned. The callers keep the transcript they started with, which is
+        # the better of the two.
+        raise ApiError(t("The cleanup model was cut off before it finished."))
     return content
 
 
@@ -593,6 +770,11 @@ def chat(messages, api_key, model, system_prompt, reasoning="",
     content = ((choices[0].get("message") or {}).get("content") or "").strip()
     if not content:
         raise ApiError(t("The model returned an empty reply."))
+    if choices[0].get("finish_reason") == "length":
+        # An answer that stops mid-sentence reads like a whole one once it has
+        # been pasted, so it is refused here for the same reason cleanup refuses
+        # a half transcript.
+        raise ApiError(t("The model was cut off before it finished."))
     return content
 
 

dikte/app.py

@@ -164,12 +164,14 @@ class Dikte:
         self._front_watch = None
 
         self.overlay = Overlay(self.conf["overlay_corner"],
-                               screen_name=self.conf["overlay_screen"])
+                               screen_name=self.conf["overlay_screen"],
+                               follow_pointer=self.conf["overlay_follows_pointer"])
         # The agent's indicator sits on top of the dictation one when both are
         # up, and drops into the corner when it is alone there.
         self.ask_overlay = Overlay(self.conf["overlay_corner"], below=self.overlay,
                                    dismissable=True,
-                                   screen_name=self.conf["overlay_screen"])
+                                   screen_name=self.conf["overlay_screen"],
+                                   follow_pointer=self.conf["overlay_follows_pointer"])
         self.recorder = audio.Recorder()
         self.pipeline = Pipeline(self.conf)
         self.ask_pipeline = Pipeline(self.conf)
@@ -287,6 +289,11 @@ class Dikte:
         self.update_action.triggered.connect(self.open_release_page)
         self.menu.addAction(self.update_action)
 
+        # Named in _refresh_tray, which is where the loaded models are known.
+        self.unload_action = QAction("", self.menu)
+        self.unload_action.triggered.connect(self.unload_models)
+        self.menu.addAction(self.unload_action)
+
         self.settings_action = QAction(t("Settings…"), self.menu)
         self.settings_action.triggered.connect(self.open_settings)
         self.menu.addAction(self.settings_action)
@@ -301,6 +308,10 @@ class Dikte:
         self.menu.addAction(self.quit_action)
 
         self.tray.setContextMenu(self.menu)
+        # A model unloads itself in the background, so what the unload row says
+        # goes stale between state changes. Refreshed as the menu opens, which
+        # is the only moment anybody reads it.
+        self.menu.aboutToShow.connect(self._refresh_tray)
         self.tray.setToolTip(t("Dikte: ready"))
         self.tray.activated.connect(self._tray_clicked)
         self._refresh_update()
@@ -396,6 +407,21 @@ class Dikte:
         )
         self.ask_cancel_action.setEnabled(self.ask_state == BUSY)
 
+        # A local model holds its memory whether or not anything is using it, so
+        # the menu says which of the two are loaded and offers to give it back.
+        # Hidden on a machine that runs neither: there is nothing to unload and
+        # nothing to report.
+        loaded = [server for server in (ggml.whisper, ggml.llm) if server.running]
+        self.unload_action.setVisible(
+            self.conf["transcribe_provider"] == "local" or self.conf.uses_local_llm()
+        )
+        self.unload_action.setText(
+            t("Unload the models") if len(loaded) > 1
+            else t("Unload the model") if loaded
+            else t("No model loaded")
+        )
+        self.unload_action.setEnabled(bool(loaded))
+
         # The agent speaks through the icon only when dictation has nothing to
         # say, since dictation is the one being waited on in front of a screen.
         if self.state == IDLE and self.ask_state != IDLE:
@@ -1207,6 +1233,18 @@ class Dikte:
         QDesktopServices.openUrl(
             QUrl(release.url if release is not None else update.RELEASES_PAGE))
 
+    def unload_models(self):
+        """Give the memory back now rather than when the idle window closes."""
+        held = [server for server in (ggml.whisper, ggml.llm)
+                if not server.unload()]
+        self._refresh_tray()
+        if held:
+            self.tray.showMessage(
+                "Dikte",
+                t("A model is loading or answering right now. Try again in a "
+                  "moment."),
+                QSystemTrayIcon.MessageIcon.Information, 5000)
+
     # ---- settings ---------------------------------------------------------
 
     def open_settings(self):
@@ -1297,10 +1335,10 @@ class Dikte:
             threading.Thread(target=warm, daemon=True).start()
 
     def _apply_settings(self):
-        self.overlay.corner = self.conf["overlay_corner"]
-        self.overlay.screen_name = self.conf["overlay_screen"]
-        self.ask_overlay.corner = self.conf["overlay_corner"]
-        self.ask_overlay.screen_name = self.conf["overlay_screen"]
+        for indicator in (self.overlay, self.ask_overlay):
+            indicator.corner = self.conf["overlay_corner"]
+            indicator.screen_name = self.conf["overlay_screen"]
+            indicator.follow_pointer = self.conf["overlay_follows_pointer"]
         self._apply_local()
         self._build_tray()
         self._refresh_tray()

dikte/cleanup.py

@@ -115,13 +115,20 @@ def _local(text, conf, system_prompt, timeout, aborter=None):
     """
     service = t("Local model")
     try:
-        return api.cleanup(
-            text, "", conf["local_llm_model"], system_prompt,
-            reasoning=conf["local_llm_reasoning"],
-            base_url=api.serving(ggml.llm),
-            timeout=max(timeout, api.LOCAL_TIMEOUT),
-            provider="local-llm", service=service, aborter=aborter,
-        )
+        # Held for the length of the request so that the idle unload does not
+        # take the model away from a block still being cleaned up.
+        with ggml.llm.busy():
+            return api.cleanup(
+                text, "", conf["local_llm_model"], system_prompt,
+                reasoning=conf["local_llm_reasoning"],
+                base_url=api.serving(ggml.llm),
+                timeout=max(timeout, api.LOCAL_TIMEOUT),
+                provider="local-llm", service=service, aborter=aborter,
+                # The ceiling is only a ceiling while it sits under what the
+                # server was started with; above that the context is what stops
+                # the reply.
+                context=ggml.llm.settings()["context"],
+            )
     except api.ApiError as exc:
         # A server that died mid-request would otherwise report only that the
         # connection dropped, when the reason is in its own output.

dikte/config.py

@@ -397,6 +397,9 @@ DEFAULTS = {
     "transcribe_model": "gpt-4o-transcribe",           # used when provider is openai
     "groq_transcribe_model": "whisper-large-v3-turbo",
     "openrouter_transcribe_model": "openai/gpt-4o-transcribe",
+    # What a timestamped run (subtitles) asks OpenRouter for: not every model
+    # there returns segment times. Empty -> openai/whisper-1.
+    "openrouter_file_model": "",
     "language": "tr",
     "transcribe_prompt": "",
 
@@ -439,6 +442,15 @@ DEFAULTS = {
     # Off rather than empty: a model trained to think will, and 300 tokens of
     # reasoning about a comma is 300 tokens of waiting.
     "local_llm_reasoning": "none",
+
+    # --- what happens to both of them when nothing is using them -------------
+    # One pair for the two servers rather than a pair each: what is being
+    # decided is whether a machine keeps gigabytes tied up between dictations,
+    # and nobody wants that answered one model at a time. On by default because
+    # a reload costs seconds and the memory costs the rest of the desktop.
+    "local_idle_unload": True,
+    "local_idle_minutes": 10,
+
     "cleanup_prompt": "",           # empty -> language-specific default
     "auto_paste": True,
     "paste_shortcut": paste.desktop().shortcuts[0],   # cmd+v on a Mac
@@ -467,6 +479,9 @@ DEFAULTS = {
     "evdev_hotkey": False,
     "overlay_corner": "bottom-left",
     "overlay_screen": "",
+    # Off, so that an indicator stays where it appeared unless it is asked to
+    # keep up with the pointer. Nothing to say when a screen is named above.
+    "overlay_follows_pointer": False,
     "keep_audio": False,
     "history_limit": 200,
     # A look at the releases page once a day, and nothing more than a look:
@@ -669,8 +684,9 @@ class Config:
             # to land on rather than reading it from there.
             name = "openai"
         who = TRANSCRIBERS[name]
+        file_model = self["openrouter_file_model"] if name == "openrouter" else ""
         return api.Target(name, who.service, self.api_key(who.key),
-                          self[who.url], self[who.model])
+                          self[who.url], self[who.model], file_model.strip())
 
     def transcribe_ready(self):
         """Whether speech to text could run right now, without opening Settings."""
@@ -703,6 +719,14 @@ class Config:
             binary=self["local_llm_binary"],
             context=int(self["local_llm_context"]),
         )
+        ggml.whisper.set_idle(self.idle_seconds())
+        ggml.llm.set_idle(self.idle_seconds())
+
+    def idle_seconds(self):
+        """How long a loaded model may sit unused. 0 means it is kept."""
+        if not self["local_idle_unload"]:
+            return 0
+        return max(1, int(self["local_idle_minutes"])) * 60
 
     def uses_local_llm(self):
         """Whether anything is set to run the local cleanup model."""

dikte/filetranscribe.py

@@ -283,11 +283,20 @@ def to_srt(text, segments):
         hours, minutes, secs = (int(g or 0) for g in match.groups())
         cues.append([hours * 3600 + minutes * 60 + secs, None, body])
 
+    # Several cues can share a whole second, so a second holds every segment
+    # that began in it and they are handed out in the order they were spoken.
     timing = {}
     for start, end, _ in segments:
-        timing.setdefault(int(start), (start, end))
+        timing.setdefault(int(start), []).append((start, end))
     for cue in cues:
-        cue[0], cue[1] = timing.get(cue[0], (float(cue[0]), 0.0))
+        found = timing.get(cue[0])
+        if found:
+            # The last one stays, so a second with more lines than it has
+            # timings hands the last of them out again rather than falling back
+            # to the bare second, which would run backwards from the line above.
+            cue[0], cue[1] = found.pop(0) if len(found) > 1 else found[0]
+        else:
+            cue[0], cue[1] = float(cue[0]), 0.0
     for index, cue in enumerate(cues):
         following = cues[index + 1][0] if index + 1 < len(cues) else 0.0
         if following > cue[0]:

dikte/ggml.py

@@ -26,6 +26,8 @@ interface already knows how to show.
 
 import atexit
 import collections
+import contextlib
+import ctypes
 import ctypes.util
 import hashlib
 import http.client
@@ -67,6 +69,10 @@ STARTUP_TIMEOUT = 180.0
 # to load takes longer than this to be read in first. The line between "worth
 # another port" and "would fail the same way again" is drawn on time.
 EARLY_EXIT_WINDOW = 5.0
+# How often the watcher looks at a model it has been asked to unload when idle.
+# Short next to any window worth setting, so the memory goes back within seconds
+# of the window closing rather than a minute after it.
+IDLE_CHECK_SECONDS = 5.0
 DOWNLOAD_CHUNK = 1 << 20
 
 # `health` is the path that answers only once the model is in memory. whisper
@@ -76,6 +82,13 @@ Program = collections.namedtuple("Program", "name repo binary health")
 
 WHISPER = Program("whisper", "ggml-org/whisper.cpp", "whisper-server", "")
 LLAMA = Program("llama", "ggml-org/llama.cpp", "llama-server", "/health")
+DIKTE_REPO = "yusufipk/dikte"
+MANAGED_WHISPER_RELEASE = "whisper.cpp-v1.9.3"
+MANAGED_WHISPER_VERSION = "v1.9.3"
+MANAGED_WHISPER_VULKAN = "whisper-bin-ubuntu-vulkan-x64.tar.gz"
+MANAGED_WHISPER_SHA256 = (
+    "c25ca76504144da488eb74441390a7b9aa7ce547e5f2f391cbd831253c9b54d8"
+)
 
 # Where the models are listed. Neither list is written into Dikte: a catalogue
 # in the source means a release of Dikte for every model somebody else
@@ -83,32 +96,139 @@ LLAMA = Program("llama", "ggml-org/llama.cpp", "llama-server", "/health")
 WHISPER_MODELS_REPO = "ggerganov/whisper.cpp"
 LLM_AUTHOR = "ggml-org"
 
+# The file llama.cpp attaches to its version releases in place of the binaries:
+# a line naming the nightly tag those are published under.
+NIGHTLY_TAG = "nightly-tag.txt"
+
 # What the whisper repository holds besides models: Core ML encoders for Apple
 # hardware and the odd loose file.
 WHISPER_PREFIX = "ggml-"
 WHISPER_SUFFIX = ".bin"
+# The mark on the whisper models trained on English alone. They are half of the
+# list, and they belong under the model they are a variant of rather than
+# scattered through it by size.
+ENGLISH_ONLY = ".en"
+
+# Full-precision weights, however they are spelled. Several times the memory of
+# a quantisation of the same model, for a difference dictation and cleanup
+# cannot see, so nothing here ever points at one.
+SIXTEEN_BIT = ("bf16", "f16", "fp16")
+
+# How many bits a weight is stored in, read off the file name. Every one of
+# these lists spells it differently, `q5_1` and `Q4_K_M` and `MXFP4` and
+# `BF16`, and the only part of that anybody choosing between two rows needs is
+# the number. Longest mark first, so `bf16` is not read as `f16`.
+BIT_DEPTHS = (("mxfp4", 4), ("bf16", 16), ("fp16", 16), ("f16", 16),
+              ("q2", 2), ("q3", 3), ("q4", 4), ("q5", 5), ("q6", 6), ("q8", 8))
 
 # What a GGUF repository holds besides the model: mmproj is the vision half of a
-# multimodal model, mtp a draft head for speculative decoding. Neither is a model
-# a server can be started on, and offering them is offering a failure.
-GGUF_SKIP = ("mmproj", "mtp-")
+# multimodal model, and mtp, dflash, dspark and eagle3 are draft heads for
+# speculative decoding. None of them is a model a server can be started on, and
+# they are the small files in the repository, so a list sorted by size puts them
+# at the top where they are likeliest to be clicked.
+GGUF_SKIP = ("mmproj", "mtp-", "dflash-", "dspark-", "eagle3-", "draft-")
 # Big enough for a 12B at Q4 and far past anything cleanup wants; the point is
 # to keep a 400 GB frontier model out of a list somebody might click.
 GGUF_MAX_BYTES = 16 << 30
 
+# Repositories that carry GGUF files but nothing a cleanup server can be started
+# on: a vision or audio tower with no text half worth running, a speech model,
+# and the base models, which continue text rather than following an instruction
+# and answer a cleanup prompt by carrying on writing the transcript.
+# Matched as plain substrings, so every one of these carries its own
+# delimiters: an unanchored "test-" is also inside "Latest-" and would drop a
+# publisher that is perfectly usable.
+LLM_REPO_SKIP = ("-Base-GGUF", "-VL-", "-Vision-", "-Omni-", "-Video-",
+                 "-TTS-", "parakeet", "/test-")
+
+GB = 1 << 30
+
 # Suggestions, not a catalogue: the list itself is fetched, and these are only
-# the rows that float to the top of it. Small instruction-following models,
-# because cleanup is punctuation and filler words rather than anything that
-# wants thinking about.
+# the rows that float to the top of it. Cleanup is punctuation, capitals and
+# filler words rather than anything that wants thinking about, so what it is
+# picked on is instruction following at a size a desktop can spare. Gemma 4
+# scores 94.6 on IFEval at E2B and 96.7 at E4B, and E2B leads here rather than
+# E4B because two points of instruction following is not worth twice the
+# weights on a job that runs while somebody waits for their sentence to appear.
+# SmolLM3 and Gemma 3 are the older pair below them. Qwen3.5 0.8B is for the
+# machines nothing else fits on; it thinks before it answers, which is what the
+# Thinking box in the settings window turns off.
 SUGGESTED_LLM = (
-    "ggml-org/gemma-3-4b-it-GGUF",
     "ggml-org/gemma-4-E2B-it-GGUF",
     "ggml-org/gemma-4-E4B-it-GGUF",
+    "ggml-org/gemma-3-4b-it-GGUF",
     "ggml-org/SmolLM3-3B-GGUF",
+    "ggml-org/Qwen3.5-0.8B-GGUF",
 )
+# Roughly what each of those weighs at the quantisation cleanup would run, to
+# the nearest half gigabyte. Not a catalogue of files: the sizes on the rows
+# come from the publisher, and this only decides which suggestion is offered
+# first on a machine that has room for some of them and not others.
+SUGGESTED_LLM_SIZE = {
+    "ggml-org/gemma-4-E2B-it-GGUF": 3 * GB,
+    "ggml-org/gemma-4-E4B-it-GGUF": 5 * GB,
+    "ggml-org/gemma-3-4b-it-GGUF": 5 * GB // 2,
+    "ggml-org/SmolLM3-3B-GGUF": 2 * GB,
+    "ggml-org/Qwen3.5-0.8B-GGUF": GB // 2,
+}
+
+# What each of them is, in the words somebody choosing between them would
+# use. A repository id says the publisher, the parameter count, the shape of
+# the weights and nothing at all about whether it is the one to click, and
+# `ggml-org/gemma-4-E2B-it-GGUF` reads as four pieces of jargon to everybody
+# who has not been reading model cards all year.
+SUGGESTED_LLM_NOTE = {
+    "ggml-org/gemma-4-E2B-it-GGUF":
+        "Google Gemma 4, the small one. The default: nothing else this size "
+        "follows an instruction as closely, and cleanup is all instruction.",
+    "ggml-org/gemma-4-E4B-it-GGUF":
+        "The same model one size up. A little more accurate, about twice the "
+        "weights and twice the wait.",
+    "ggml-org/gemma-3-4b-it-GGUF":
+        "The previous Gemma. Still good, and the smallest of the Gemmas here.",
+    "ggml-org/SmolLM3-3B-GGUF":
+        "Hugging Face's own small model, for a machine the Gemmas crowd.",
+    "ggml-org/Qwen3.5-0.8B-GGUF":
+        "The smallest of them, for a machine nothing else fits on. It thinks "
+        "before it answers unless Thinking below is off.",
+}
+
 # Turbo at q5_0 is smaller than `small` and better than it, which makes the
-# usual "start small" advice point at the same file as "start good".
+# usual "start small" advice point at the same file as "start good". It is
+# large-v3 with the decoder cut from 32 layers to 4: several times faster, at
+# one to two points of word error in English and about two and a half in the
+# other languages.
 SUGGESTED_WHISPER = "ggml-large-v3-turbo-q5_0.bin"
+# Those two and a half points back, for twice the file and several times the
+# work per second. Only suggested where there is a card to do the work and
+# memory to hold it, because that is where the trade stops costing anything a
+# person waiting for a dictation would notice.
+ACCURATE_WHISPER = "ggml-large-v3-q5_0.bin"
+# What to point at instead on a machine the turbo model would crowd. Same
+# quantisation ladder, one rung down in size and in accuracy.
+SMALL_MACHINE_WHISPER = "ggml-small-q5_1.bin"
+# Under this much system memory, a 600 MB model plus the rest of a desktop is
+# already tight, so the suggestion drops to the smaller one. Over the other,
+# the accurate model is the one to point at.
+SMALL_MACHINE = 4 * GB
+# Fifteen and not sixteen: what the machine reports is what is left after the
+# firmware and the graphics have taken their reservations out of it, and a
+# 16 GB machine answers about 15.4. A threshold written at the number on the
+# box is one no machine sold as that size ever reaches.
+ROOMY_MACHINE = 15 * GB
+# What a model may take of this machine's memory before it is called too big:
+# half of it, less a gigabyte for the context and the runtime around the
+# weights. A rule of thumb rather than a measurement, and deliberately a
+# cautious one, because the failure it is guarding against is a machine that
+# swaps itself to a standstill rather than a model that refuses to load.
+MEMORY_SHARE = 0.5
+MEMORY_OVERHEAD = GB
+# What is left to offer on a machine too small for the sum above to leave
+# anything. Enough for the smallest whisper models and for a sub-billion
+# cleanup model, which is what such a machine can run.
+MEMORY_FLOOR = GB // 2
+# What total_memory() read the one time it asked. None until it has.
+_MEMORY = None
 
 
 class LocalError(Exception):
@@ -277,6 +397,102 @@ def _wanted_assets(program):
     return (f"bin-ubuntu-{arch}.tar.gz",)
 
 
+def _managed_whisper(program, tag=""):
+    """Whether this machine is one Dikte publishes its own whisper-server for.
+
+    Linux x86_64 with a Vulkan loader on it, and no version asked for by hand:
+    a pinned version is upstream's to answer.
+    """
+    return (not tag and program is WHISPER and sys.platform == "linux"
+            and platform.machine().lower() in ("x86_64", "amd64")
+            and _has_vulkan())
+
+
+def _managed_asset(refresh=False):
+    """The Vulkan whisper-server Dikte builds itself, or None.
+
+    Taken only when the archive's digest is the reviewed one. Anything else,
+    a release that is not there yet, a GitHub that cannot be reached, a file
+    that is not the reviewed bytes, leaves upstream's processor build as the
+    answer, and the install record says which of the two landed.
+    """
+    try:
+        _, assets = hub.release(DIKTE_REPO, MANAGED_WHISPER_RELEASE,
+                                refresh=refresh)
+    except hub.HubError:
+        return None
+    return next((a for a in assets
+                 if a.name.endswith(MANAGED_WHISPER_VULKAN)
+                 and a.sha256 == MANAGED_WHISPER_SHA256), None)
+
+
+def _matching_asset(program, assets):
+    """The archive this machine wants out of one release's files, or None."""
+    for ending in _wanted_assets(program):
+        item = next((a for a in assets if a.name.endswith(ending)), None)
+        if item:
+            return item
+    return None
+
+
+def _pick_asset(program, tag="", refresh=False):
+    """(tag, Item) for the release archive to install. Item is None when there
+    is none for this machine.
+
+    Dikte's own Vulkan whisper-server comes before upstream's where this
+    machine is one it is built for, because whisper.cpp publishes no Vulkan
+    archive for Linux at all.
+
+    A named tag is taken as given. For the newest, what GitHub answers is not
+    always where the builds are: llama.cpp's latest release is a version marker
+    carrying a single nightly-tag.txt, which names the tag the archives are
+    actually attached to, and those are prereleases that "latest" never points
+    at. The pointer is followed when it is there, and when it is not, the newest
+    release that does carry a build for this machine is taken instead.
+    """
+    if _managed_whisper(program, tag):
+        item = _managed_asset(refresh=refresh)
+        if item:
+            return MANAGED_WHISPER_VERSION, item
+    named = bool(tag) and tag != "latest"
+    missing = None
+    try:
+        tag, assets = hub.release(program.repo, tag or "latest", refresh=refresh)
+    except hub.HubError as exc:
+        # A release carrying no files at all is the case the search below exists
+        # for, not a reason to stop before it: the build for this machine may be
+        # attached to a prerelease that "latest" never points at. The failure is
+        # kept rather than dropped, because an unreachable GitHub arrives here
+        # the same way and that one is the message the caller wants.
+        if named:
+            raise
+        missing, assets = exc, []
+    item = _matching_asset(program, assets)
+    if item or named:
+        return tag, item
+    # Best effort from here on: a machine this project publishes nothing for is
+    # not a failed lookup, and the caller's message about that is the useful
+    # one. Whatever goes wrong while looking further leaves it standing.
+    try:
+        pointer = next((a for a in assets if a.name == NIGHTLY_TAG), None)
+        if pointer:
+            nightly = hub.text(pointer.url).strip()
+            if nightly:
+                found, assets = hub.release(program.repo, nightly, refresh=refresh)
+                item = _matching_asset(program, assets)
+                if item:
+                    return found, item
+        for found, assets in hub.releases(program.repo, refresh=refresh):
+            item = _matching_asset(program, assets)
+            if item:
+                return found, item
+    except hub.HubError:
+        pass
+    if missing is not None:
+        raise missing
+    return tag, None
+
+
 def _install_record(program):
     return BIN_DIR / program.name / "installed.json"
 
@@ -300,6 +516,21 @@ def installed_version(program):
     return _read_record(program).get("tag") or ""
 
 
+def vulkan_missing(program):
+    """Whether what Dikte installed is the processor build on a machine the
+    Vulkan one was fetched for.
+
+    The Vulkan whisper-server is a release of Dikte's own, published by hand
+    once the reviewed archive is built, and the install falls back to the
+    upstream processor build whenever that release, the file in it, or its
+    reviewed digest is not there. Nothing is wrong with the fallback except
+    that it is invisible: a graphics card sitting idle looks exactly like a
+    graphics card being used.
+    """
+    return (bool(installed_program(program))
+            and _read_record(program).get("backend") == "processor")
+
+
 def program_path(program, custom=""):
     """Which copy of the program to run, or "" when there is none.
 
@@ -372,18 +603,15 @@ def install_program(program, tag="", on_progress=None, should_stop=None,
 
     `tag` is empty for whatever the project released last, which is the point:
     a version pinned in Dikte's source would mean a release of Dikte every time
-    whisper.cpp has one.
+    whisper.cpp has one. The Linux Vulkan whisper-server is the exception, and
+    _pick_asset says why.
     """
+    managed = _managed_whisper(program, tag)
     try:
-        tag, assets = hub.release(program.repo, tag or "latest", refresh=refresh)
+        tag, item = _pick_asset(program, tag, refresh=refresh)
     except hub.HubError as exc:
         raise LocalError(str(exc)) from exc
 
-    item = None
-    for ending in _wanted_assets(program):
-        item = next((a for a in assets if a.name.endswith(ending)), None)
-        if item:
-            break
     if item is None:
         # Nothing to download and nothing to install for you: whisper.cpp
         # publishes no macOS binary, and Homebrew's whisper-cpp is configured
@@ -447,8 +675,15 @@ def install_program(program, tag="", on_progress=None, should_stop=None,
         # Found under the sibling, run from the final directory.
         binary = into / binary.relative_to(fresh)
         # Written last, so the record never points at anything half-made.
-        _install_record(program).write_text(
-            json.dumps({"tag": tag, "binary": str(binary)}), encoding="utf-8")
+        record = {"tag": tag, "binary": str(binary)}
+        if managed:
+            # Which of the two builds this machine ended up with. Only written
+            # where both were on offer, so an install that never had the
+            # choice is not made to look like a fallback.
+            record["backend"] = (
+                "vulkan" if item.name.endswith(MANAGED_WHISPER_VULKAN)
+                else "processor")
+        _install_record(program).write_text(json.dumps(record), encoding="utf-8")
     except OSError as exc:
         raise LocalError(t("Could not install {name}: {error}",
                            name=program.name, error=exc)) from exc
@@ -479,9 +714,212 @@ def _drop_old_versions(program, keep):
         pass
 
 
+# --- what this machine can run --------------------------------------------
+
+
+def total_memory():
+    """Bytes of memory on this machine, or 0 when it cannot be read.
+
+    Zero is a real answer and not a failure: every caller treats an unknown
+    machine as one big enough for whatever it is looking at, because a wrong
+    "too big" is worse advice than none.
+
+    Read once and kept. The memory in a machine does not change while Dikte
+    runs, and a list of thirty rows asks this question seventy times: on the
+    Mac path below, where the answer comes from a program rather than a
+    library call, that was seventy processes started on the interface thread
+    every time a list was drawn.
+    """
+    global _MEMORY
+    if _MEMORY is None:
+        _MEMORY = max(_read_memory(), 0)
+    return _MEMORY
+
+
+def _read_memory():
+    """What the system says, which on a bad day is a negative number.
+
+    sysconf answers -1 for a limit it holds to be indeterminate, and CPython
+    hands that straight back rather than raising, so the product below can
+    come out negative. The caller floors it at zero, which is the answer for
+    a machine nothing could be read from: a 64 GB workstation whose sysconf
+    shrugged was otherwise being told every model past 512 MB was too big
+    for it.
+    """
+    try:
+        return os.sysconf("SC_PAGE_SIZE") * os.sysconf("SC_PHYS_PAGES")
+    except (AttributeError, ValueError, OSError):
+        pass
+    if sys.platform == "darwin":
+        # Not every build of Python on a Mac has SC_PHYS_PAGES in its sysconf
+        # table, and this is the number the system itself is asked for.
+        try:
+            out = subprocess.run(["sysctl", "-n", "hw.memsize"], check=True,
+                                 capture_output=True, text=True, timeout=5)
+            return int(out.stdout.strip())
+        except (OSError, ValueError, subprocess.SubprocessError):
+            return 0
+    if sys.platform != "win32":
+        return 0
+
+    class Status(ctypes.Structure):
+        _fields_ = [("dwLength", ctypes.c_ulong),
+                    ("dwMemoryLoad", ctypes.c_ulong),
+                    ("ullTotalPhys", ctypes.c_ulonglong),
+                    ("ullAvailPhys", ctypes.c_ulonglong),
+                    ("ullTotalPageFile", ctypes.c_ulonglong),
+                    ("ullAvailPageFile", ctypes.c_ulonglong),
+                    ("ullTotalVirtual", ctypes.c_ulonglong),
+                    ("ullAvailVirtual", ctypes.c_ulonglong),
+                    ("ullAvailExtendedVirtual", ctypes.c_ulonglong)]
+
+    try:
+        status = Status()
+        status.dwLength = ctypes.sizeof(Status)
+        if ctypes.windll.kernel32.GlobalMemoryStatusEx(ctypes.byref(status)):
+            return int(status.ullTotalPhys)
+    except (AttributeError, OSError, ValueError):
+        pass
+    return 0
+
+
+def accelerator():
+    """The graphics interface this machine offers, or "".
+
+    The machine's half of the answer only. Whether a card is actually reached
+    also depends on which build landed, and the program line above says that:
+    a processor build ignores the card whatever is installed here. What this
+    is for is the other half, which nothing else on the window says at all.
+    """
+    if sys.platform == "darwin":
+        return "Metal"
+    return "Vulkan" if _has_vulkan() else ""
+
+
+def memory_budget(memory=None):
+    """What a model may weigh on this machine, or 0 when that is unknown.
+
+    Floored rather than allowed to reach zero: on a 2 GB machine the share
+    less the overhead is nothing at all, and a budget of nothing is the same
+    number this returns for a machine it could not read, which would turn the
+    tightest machine there is into the one where everything is offered.
+    """
+    memory = total_memory() if memory is None else memory
+    if not memory:
+        return 0
+    return max(int(memory * MEMORY_SHARE) - MEMORY_OVERHEAD, MEMORY_FLOOR)
+
+
+def fits(size, memory=None):
+    """Whether a model of this size is worth offering on this machine."""
+    budget = memory_budget(memory)
+    return not budget or size <= budget
+
+
+def suggested_whisper(memory=None, graphics=None):
+    """The whisper model to point at here, by name.
+
+    Three machines. One with no room, which gets the model that leaves some.
+    One with a card and memory to spare, which gets the accurate model, because
+    the several times the work it is per second is several times a fraction of
+    a second there. Everything in between gets turbo, which is the answer
+    almost every time somebody asks.
+
+    A Vulkan or Metal loader is not proof of a fast card, so the accurate model
+    waits on the memory as well: a machine with 16 GB in it and a driver
+    installed is one that will not notice either way.
+    """
+    memory = total_memory() if memory is None else memory
+    graphics = accelerator() if graphics is None else graphics
+    if memory and memory < SMALL_MACHINE:
+        return SMALL_MACHINE_WHISPER
+    if graphics and memory >= ROOMY_MACHINE:
+        return ACCURATE_WHISPER
+    return SUGGESTED_WHISPER
+
+
+def suggested_llm(memory=None):
+    """The suggested cleanup repositories, the ones that fit here first.
+
+    The order they are written in is the order they are worth having. What
+    this changes is only which of them a machine that cannot hold the best one
+    is shown first, and nothing is dropped: a model that does not fit today
+    fits once something else is closed.
+    """
+    return sorted(SUGGESTED_LLM,
+                  key=lambda repo: not fits(SUGGESTED_LLM_SIZE.get(repo, 0),
+                                            memory))
+
+
+def recommended(items, want="", memory=None):
+    """The one row out of `items` worth pointing at here, or "".
+
+    `want` is taken when it is on offer and fits. Without it, which is the
+    cleanup list, the smallest file that does is taken: q4 is where these
+    lists start, and every rung above it is roughly twice the memory and twice
+    the wait for a difference neither dictation nor cleanup can see. The
+    16-bit weights are left out for the same reason, twice over.
+    """
+    fitting = [i for i in items if fits(i.size, memory)]
+    if want and any(i.name == want for i in fitting):
+        return want
+    usable = [i for i in fitting
+              if not any(mark in i.name.lower() for mark in SIXTEEN_BIT)]
+    return min(usable, key=lambda i: i.size).name if usable else ""
+
+
 # --- the models -----------------------------------------------------------
 
 
+def bit_depth(name):
+    """The bits per weight the file name says, or 0 when it says nothing."""
+    lowered = name.lower()
+    for mark, bits in BIT_DEPTHS:
+        if mark in lowered:
+            return bits
+    return 0
+
+
+def whisper_family(name):
+    """The model a whisper file belongs to: ggml-small.en-q5_1.bin is `small`.
+
+    The list arrives sorted by size and nothing else, which interleaves the
+    families: `large-v3-turbo-q5_0` lands between the two `medium`
+    quantisations, half a screen from the turbo model it is a copy of. Grouping
+    is what puts the choice between models above the choice of quantisation,
+    which is the order somebody actually makes them in.
+    """
+    stem = name
+    if stem.startswith(WHISPER_PREFIX):
+        stem = stem[len(WHISPER_PREFIX):]
+    if stem.endswith(WHISPER_SUFFIX):
+        stem = stem[:-len(WHISPER_SUFFIX)]
+    head, _, last = stem.rpartition("-")
+    # q5_0, q5_1, q8_0. `turbo` is the other thing a last chunk can be, and it
+    # is part of the model's name rather than a quantisation of it.
+    if head and last.startswith("q") and last[1:].replace("_", "").isdigit():
+        stem = head
+    return stem[:-len(ENGLISH_ONLY)] if stem.endswith(ENGLISH_ONLY) else stem
+
+
+def whisper_groups(items):
+    """[(family, [Item])] for a whisper list: one group per model.
+
+    Groups by how big the model gets rather than by a ladder written down
+    here, so a family published next year sorts itself. Inside one, the
+    multilingual files come before the English-only ones and the small
+    quantisations before the large.
+    """
+    groups = {}
+    for item in items:
+        groups.setdefault(whisper_family(item.name), []).append(item)
+    ordered = sorted(groups.items(),
+                     key=lambda pair: (max(i.size for i in pair[1]), pair[0]))
+    return [(family, sorted(files,
+                            key=lambda i: (ENGLISH_ONLY in i.name, i.size)))
+            for family, files in ordered]
+
+
 def whisper_models(refresh=False):
     """[hub.Item] for every whisper model on offer, smallest first."""
     try:
@@ -494,10 +932,23 @@ def whisper_models(refresh=False):
     return sorted(models, key=lambda f: f.size)
 
 
+def can_clean(repo):
+    """Whether a repository could hold a model cleanup can be started on.
+
+    By name, because the alternative is a file listing per repository and the
+    list is forty of them. It catches the kinds that are never a cleanup model
+    rather than the ones that are too big, which the file sizes answer exactly
+    once a publisher is chosen.
+    """
+    lowered = repo.lower()
+    return not any(mark.lower() in lowered for mark in LLM_REPO_SKIP)
+
+
 def llm_repos(refresh=False):
     """Repository ids for the GGUF models on offer, suggestions first."""
     try:
-        found = [r.id for r in hub.repos(author=LLM_AUTHOR, refresh=refresh)]
+        found = [r.id for r in hub.repos(author=LLM_AUTHOR, refresh=refresh)
+                 if can_clean(r.id)]
     except hub.HubError:
         # A menu rather than a catalogue: with nothing to show, the suggestions
         # are still worth showing, and whatever is wrong with the network will
@@ -505,6 +956,14 @@ def llm_repos(refresh=False):
         found = []
     if not found:
         return list(SUGGESTED_LLM)
+    # Gemma publishes its base models under the instruction-tuned one's name
+    # with the `-it` taken out, so the two sit next to each other in the list
+    # and the wrong one answers a cleanup prompt by carrying on writing the
+    # transcript. Dropped only where the tuned sibling is here to drop it for.
+    tuned = set(found)
+    found = [r for r in found
+             if not r.endswith("-GGUF")
+             or r[:-len("-GGUF")] + "-it-GGUF" not in tuned]
     first = [r for r in SUGGESTED_LLM if r in found]
     return first + [r for r in found if r not in first]
 
@@ -662,6 +1121,15 @@ class Server:
         # The pid this instance last wrote to its pid file, so _forget never
         # removes a file some other Dikte wrote after us.
         self._pid = 0
+        # The idle unload. `_idle` is the window in seconds, zero meaning the
+        # model stays loaded until something else stops it; `_used` is when the
+        # address was last handed out or a request last finished; `_busy` counts
+        # the requests still in flight. The count is there because a file being
+        # transcribed is one address lookup and then minutes of work, which to a
+        # clock started at the lookup looks exactly like a model nobody wants.
+        self._idle = 0.0
+        self._used = 0.0
+        self._busy = 0
 
     # ---- settings --------------------------------------------------------
 
@@ -679,6 +1147,23 @@ class Server:
         with self._lock:
             return dict(self._settings)
 
+    def set_idle(self, seconds):
+        """How long a loaded model may sit unused before the memory goes back.
+
+        Deliberately not one of the settings above: those describe the server
+        that is running, and changing one has to restart it. This describes how
+        long to keep it, which the server it is applied to never needs to know.
+        Zero keeps the model until something else stops it.
+        """
+        with self._lock:
+            self._idle = max(0.0, float(seconds))
+
+    @property
+    def idle(self):
+        """The window `set_idle` was last given, in seconds."""
+        with self._lock:
+            return self._idle
+
     def _settings_key(self):
         """What a running server would have to be restarted for."""
         return json.dumps(self._settings, sort_keys=True, default=str)
@@ -718,13 +1203,83 @@ class Server:
             proc, port, log = self._launch(settings)
             with self._lock:
                 self._proc, self._port, self._log, self._key = proc, port, log, key
+                # Only the clock. The count is not this launch's to reset: a
+                # caller that took a hold and then asked for the address, which
+                # is what the local cleanup does, would have it wiped here and
+                # spend the whole request unprotected.
+                self._used = time.monotonic()
+            threading.Thread(target=self._watch, args=(proc,), daemon=True).start()
             return self.base_url()
 
     def _current_url(self):
+        """The address of a server running the current settings, or "".
+
+        Asking counts as using it. Everything that asks is about to send a
+        request, and the idle watcher reads the same clock, so the stamp has to
+        be set here rather than where the answer comes back.
+        """
         with self._lock:
             up = self._proc is not None and self._proc.poll() is None
-            return (f"http://{HOST}:{self._port}/v1"
-                    if up and self._key == self._settings_key() else "")
+            if not (up and self._key == self._settings_key()):
+                return ""
+            self._used = time.monotonic()
+            return f"http://{HOST}:{self._port}/v1"
+
+    @contextlib.contextmanager
+    def busy(self):
+        """Hold the model for the length of one request.
+
+        A dictation is over a second after the address was handed out, but a
+        file is minutes of it, and an hour of meeting is longer still. Without
+        the count the watcher would unload the model out from under the request
+        that started it.
+        """
+        with self._lock:
+            self._busy += 1
+        try:
+            yield
+        finally:
+            with self._lock:
+                # Nothing else moves the count, so every hold that was taken is
+                # given back here and it stays balanced across a restart. A hold
+                # outliving the server it was taken against only keeps the next
+                # one loaded a moment longer, which is the safe way round.
+                self._busy -= 1
+                self._used = time.monotonic()
+
+    def _idle_now(self, proc):
+        """Whether `proc` is still ours and has been sitting unused long enough."""
+        with self._lock:
+            if self._proc is not proc or not self._idle or self._busy:
+                return False
+            return time.monotonic() - self._used >= self._idle
+
+    def _watch(self, proc):
+        """Give the memory back when nothing has asked anything for a while.
+
+        One thread per launch, holding the process it was started for, so that a
+        server stopped and started again is watched by the new thread alone and
+        this one leaves on the first pass that finds its own process gone.
+
+        Started whatever the window is, zero included: turning the unload on in
+        Settings has to reach a model that is already loaded, and a thread that
+        wakes every few seconds to read one number is cheaper than the machinery
+        for starting one later.
+        """
+        while True:
+            time.sleep(IDLE_CHECK_SECONDS)
+            with self._lock:
+                if self._proc is not proc:
+                    return           # stopped, or replaced by a later launch
+            if not self._idle_now(proc):
+                continue
+            with self._starting:
+                # Asked once more under the lock a start has to take. An address
+                # handed out while this thread waited its turn stamps _used, and
+                # the request behind it must not arrive at a server killed here.
+                if self._idle_now(proc):
+                    self._stop_now()
+                    return
 
     def _launch(self, settings):
         args = self._build(settings)        # raises LocalError when unusable
@@ -831,6 +1386,31 @@ class Server:
         except subprocess.TimeoutExpired:
             pass
 
+    def unload(self):
+        """Stop the server unless it is in the middle of something.
+
+        The same rule the idle watcher goes by, taken by hand from the menu, and
+        it says no for the same reason: the memory is worth having back, but not
+        at the price of the dictation waiting on it. A model still being read in
+        counts as in the middle of something too, and that is why the lock is
+        asked for rather than waited on: this runs on the interface's own
+        thread, and a start holds _starting for as long as the load takes, which
+        for a large model on a cold cache is most of a minute. True when nothing
+        is loaded any more, either way.
+        """
+        if not self._starting.acquire(blocking=False):
+            return False
+        try:
+            with self._lock:
+                if self._proc is None:
+                    return True
+                if self._busy:
+                    return False
+            self._stop_now()
+            return True
+        finally:
+            self._starting.release()
+
     def stop(self):
         # Taking _starting means a stop cannot slide past a launch in flight:
         # serve() finishes registering its child first, and the child is then

dikte/hub.py

@@ -121,6 +121,13 @@ def _digest(value):
     return value.split(":", 1)[1] if value.startswith("sha256:") else value
 
 
+def _assets(data):
+    return [Item(a.get("name") or "", a.get("browser_download_url") or "",
+                 int(a.get("size") or 0), _digest(a.get("digest")))
+            for a in (data.get("assets") or [])
+            if a.get("browser_download_url")]
+
+
 def release(repo, tag="latest", refresh=False):
     """(tag, [Item]) for one GitHub release, newest when no tag is given."""
     where = "latest" if tag in ("", "latest") else f"tags/{tag}"
@@ -128,10 +135,47 @@ def release(repo, tag="latest", refresh=False):
                   f"{GITHUB_API}/repos/{repo}/releases/{where}", refresh=refresh)
     if not isinstance(data, dict) or not data.get("assets"):
         raise HubError(t("{repo} has no downloadable release.", repo=repo))
-    assets = [Item(a.get("name") or "", a.get("browser_download_url") or "",
-                   int(a.get("size") or 0), _digest(a.get("digest")))
-              for a in data["assets"] if a.get("browser_download_url")]
-    return data.get("tag_name") or tag, assets
+    return data.get("tag_name") or tag, _assets(data)
+
+
+def releases(repo, limit=20, refresh=False):
+    """[(tag, [Item])] for the recent releases, newest first, with their files.
+
+    "latest" is one release and this is the list behind it, prereleases
+    included: a project that attaches its builds to a prerelease is invisible
+    to release() above, and its newest usable build is in here.
+    """
+    data = _fetch(f"gh-list-{repo}-{limit}",
+                  f"{GITHUB_API}/repos/{repo}/releases?per_page={limit}",
+                  refresh=refresh)
+    if not isinstance(data, list):
+        raise HubError(t("{repo} has no downloadable release.", repo=repo))
+    out = []
+    for entry in data:
+        tag, items = entry.get("tag_name") or "", _assets(entry)
+        if tag and items:
+            out.append((tag, items))
+    return out
+
+
+def text(url, limit=4096, timeout=20):
+    """A small text file from a release, as a string.
+
+    Not cached and not checksummed, because what it carries is a pointer: a few
+    bytes naming the release the actual archives are attached to, read once on
+    the way to a download that is checked in full.
+    """
+    request = urllib.request.Request(url, headers={"User-Agent": USER_AGENT})
+    try:
+        with urllib.request.urlopen(request, timeout=timeout) as response:
+            return response.read(limit).decode("utf-8", "replace")
+    except urllib.error.HTTPError as exc:
+        exc.close()
+        raise HubError(t("{url} answered HTTP {code}.",
+                         url=urllib.parse.urlsplit(url).netloc, code=exc.code)) from exc
+    except (urllib.error.URLError, OSError, ValueError) as exc:
+        raise HubError(t("Could not reach {url}: {error}",
+                         url=urllib.parse.urlsplit(url).netloc, error=exc)) from exc
 
 
 def newest_release(repo, refresh=False):

dikte/i18n.py

@@ -189,8 +189,10 @@ TR = {
     "Restore the previous clipboard after pasting":
         "Yapıştırdıktan sonra eski pano içeriğini geri koy",
     "Indicator screen": "Gösterge ekranı",
-    "Follow the mouse pointer": "Fare imlecini takip et",
+    "Follow the active screen": "Etkin ekranı takip et",
     "{name} (not connected)": "{name} (bağlı değil)",
+    "Move it when the active screen changes":
+        "Etkin ekran değiştiğinde göstergeyi de taşı",
     "Indicator corner": "Gösterge köşesi",
     "bottom-left": "sol-alt",
     "bottom-right": "sağ-alt",
@@ -228,6 +230,11 @@ TR = {
     "Transcript cleanup": "Transkripti temizleme",
     "API key": "API anahtarı",
     "Model": "Model",
+    "Audio file model": "Ses dosyası modeli",
+    "The model a timestamped audio file (subtitles) is sent to. Not every model on "
+    "OpenRouter returns segment times; empty means openai/whisper-1.":
+        "Zaman damgalı bir ses dosyasının (altyazı) gönderildiği model. OpenRouter'daki her "
+        "model segment zamanı döndürmez; boşsa openai/whisper-1 kullanılır.",
     "Provider": "Sağlayıcı",
     "sk-… (falls back to OPENAI_API_KEY)": "sk-… (boşsa OPENAI_API_KEY kullanılır)",
     "gsk_… (falls back to GROQ_API_KEY)": "gsk_… (boşsa GROQ_API_KEY kullanılır)",
@@ -775,20 +782,120 @@ TR = {
     "On this machine": "Bu makinede",
     "Use the graphics card": "Ekran kartını kullan",
     "Load the model when Dikte starts": "Modeli Dikte açılırken yükle",
+    "Models on this machine": "Bu makinedeki modeller",
+    "Unload a model that is sitting unused": "Kullanılmayan modeli bellekten çıkar",
+    "A loaded model holds its memory whether anything is using it or "
+    "not: over a gigabyte for whisper, several for an LLM. Unloading "
+    "gives that back to the rest of the desktop, and the next "
+    "dictation loads it again at the cost of the seconds that takes.":
+        "Yüklü bir model, kullanılsa da kullanılmasa da belleği tutar: whisper "
+        "için bir gigabaytın üzerinde, bir LLM için birkaç gigabayt. Bellekten "
+        "çıkarmak bunu masaüstünün geri kalanına iade eder, sonraki dikte de "
+        "modeli birkaç saniye bekleyerek yeniden yükler.",
+    " minute": " dakika",
+    " minutes": " dakika",
+    "After": "Şu kadar sonra",
+    "Unload the model": "Modeli bellekten çıkar",
+    "Unload the models": "Modelleri bellekten çıkar",
+    "No model loaded": "Yüklü model yok",
+    "A model is loading or answering right now. Try again in a "
+    "moment.":
+        "Bir model şu anda yükleniyor ya da cevap veriyor. Az sonra tekrar "
+        "deneyin.",
     "Local whisper": "Yerel whisper",
     "Local model": "Yerel model",
     "Not installed.": "Kurulu değil.",
     "Installed on the system: {path}": "Sistemde kurulu: {path}",
+    "Download again": "Yeniden indir",
     "Downloaded, version {version}.": "İndirildi, sürüm {version}.",
+    "Downloaded, version {version}. There was no Vulkan build, "
+    "so this one runs on the processor.":
+        "İndirildi, sürüm {version}. Vulkan sürümü yoktu, bu sürüm işlemcide çalışıyor.",
     "Fetching the model list…": "Model listesi çekiliyor…",
     "Downloading…": "İndiriliyor…",
+    "Starting the download…": "İndirme başlatılıyor…",
     "Downloading: {done} of {total}{share}": "İndiriliyor: {done} / {total}{share}",
     "Download stopped.": "İndirme durduruldu.",
     "Ready: {name}.": "Hazır: {name}.",
     "Nothing downloaded yet.": "Henüz bir şey indirilmedi.",
     "{name} has not been downloaded yet.": "{name} henüz indirilmedi.",
+    "{name} is here, but the program above is not. Download it first.":
+        "{name} burada, ama yukarıdaki program değil. Önce onu indirin.",
+    "{name} is not on this machine and this publisher does not offer it. "
+    "Choose another model, or another publisher.":
+        "{name} bu makinede yok ve bu yayıncı da sunmuyor. Başka bir model, "
+        "ya da başka bir yayıncı seçin.",
     "downloaded": "indirildi",
     "not downloaded": "indirilmedi",
+    "recommended": "önerilen",
+    "{bits}-bit": "{bits} bit",
+    "English only": "yalnızca İngilizce",
+    "All": "Tümü",
+    "Everything ggml-org publishes, including the models that are too big to "
+    "run here and the ones that are not for cleaning up text.":
+        "ggml-org'un yayımladığı her şey; burada çalıştırılamayacak kadar "
+        "büyük olanlar ve metin temizlemek için olmayanlar dahil.",
+    "Google Gemma 4, the small one. The default: nothing else this size "
+    "follows an instruction as closely, and cleanup is all instruction.":
+        "Google Gemma 4'ün küçüğü. Varsayılan: bu boyutta verilen yönergeyi "
+        "bu kadar iyi izleyen başka bir model yok, temizleme de baştan sona "
+        "yönerge demek.",
+    "The same model one size up. A little more accurate, about twice the "
+    "weights and twice the wait.":
+        "Aynı modelin bir boy büyüğü. Biraz daha isabetli, yaklaşık iki katı "
+        "ağırlık ve iki katı bekleyiş.",
+    "The previous Gemma. Still good, and the smallest of the Gemmas here.":
+        "Bir önceki Gemma. Hâlâ iyi ve buradaki Gemma'ların en küçüğü.",
+    "Hugging Face's own small model, for a machine the Gemmas crowd.":
+        "Hugging Face'in kendi küçük modeli; Gemma'ların sıkıştırdığı bir "
+        "makine için.",
+    "The smallest of them, for a machine nothing else fits on. It thinks "
+    "before it answers unless Thinking below is off.":
+        "En küçükleri; başka hiçbir şeyin sığmadığı bir makine için. "
+        "Aşağıdaki Düşünme kapalı değilse cevaplamadan önce düşünür.",
+    "too big for this machine": "bu makine için fazla büyük",
+    "This machine": "Bu makine",
+    "Graphics: {name}.": "Ekran kartı: {name}.",
+    "No graphics interface found, so this runs on the processor.":
+        "Ekran kartı arayüzü bulunamadı, bu yüzden işlemcide çalışıyor.",
+    "Memory: {size}.": "Bellek: {size}.",
+    "A model may take half of this memory, less a gigabyte for the context "
+    "around the weights. Anything past that is marked too big; it may still "
+    "load, on a machine with nothing else open.":
+        "Bir model bu belleğin yarısını, ağırlıkların çevresindeki bağlam için "
+        "bir gigabayt düşülerek kullanabilir. Bunu aşan modeller fazla büyük "
+        "diye işaretlenir; başka hiçbir şeyin açık olmadığı bir makinede yine "
+        "de yüklenebilirler.",
+    "Recommended for this machine": "Bu makine için önerilen",
+    "Everything this publisher offers": "Bu yayıncının sunduğu her şey",
+    "Already on this machine": "Bu makinede zaten var",
+    "Chosen, but not downloaded": "Seçili, ama indirilmedi",
+    "{repo} publishes nothing that can be run here. Its models are split "
+    "across files, larger than {cap}, or pieces of a model rather than one. "
+    "Choose another publisher.":
+        "{repo} burada çalıştırılabilecek bir şey yayımlamıyor. Modelleri "
+        "birden çok dosyaya bölünmüş, {cap} boyutundan büyük ya da modelin "
+        "kendisi değil parçaları. Başka bir yayıncı seçin.",
+    "large-v3 makes the fewest mistakes and is the slowest of them. "
+    "large-v3-turbo is that model with a four layer decoder in place of a "
+    "thirty-two layer one: several times faster, at one to two points of word "
+    "error in English and about two and a half in the other languages. Below "
+    "those, every step down the list trades accuracy for size, and the .en "
+    "models are trained on English alone.":
+        "En az hatayı large-v3 yapar, en yavaşı da odur. large-v3-turbo, aynı "
+        "modelin otuz iki katmanlı çözücüsü yerine dört katmanlı bir çözücü "
+        "konmuş hâli: birkaç kat hızlı, karşılığında İngilizcede bir iki "
+        "puan, diğer dillerde yaklaşık iki buçuk puan kelime hatası. Bunların "
+        "altında listede her basamak, doğruluğu boyuta değişir; .en modelleri "
+        "ise yalnızca İngilizce ile eğitilmiştir.",
+    "Cleanup is punctuation, capitals and filler words, so what these are "
+    "picked on is following an instruction rather than knowing anything. "
+    "Start at a q4 file; the 16-bit ones are several times the memory for a "
+    "difference this job cannot see.":
+        "Temizleme; noktalama, büyük harf ve dolgu sözcükleri demek, yani bu "
+        "modeller bir şey bilmelerine değil verilen yönergeyi izlemelerine "
+        "göre seçilir. Bir q4 dosyasından başlayın; 16 bitlik olanlar, bu işin "
+        "göremeyeceği bir fark için kat kat bellek ister.",
     "Delete model": "Modeli sil",
     "Delete {name} from this machine?": "{name} bu makineden silinsin mi?",
     "Runs on this machine, on llama.cpp.": "Bu makinede, llama.cpp üzerinde çalışır.",
@@ -872,6 +979,10 @@ TR = {
     "“Off”.":
         "Temizleme modeli bütün yanıtını düşünmeye harcadı. Düşünme'yi "
         "“Kapalı” yap.",
+    "The cleanup model was cut off before it finished.":
+        "Temizleme modeli bitiremeden kesildi.",
+    "The model was cut off before it finished.":
+        "Model bitiremeden kesildi.",
 
     # --- this pass's new messages ---------------------------------------
     "Audio recorder stopped before receiving sound":

dikte/overlay.py

@@ -1,6 +1,7 @@
 """The small recording indicator that appears in a screen corner without taking focus."""
 
 import math
+import os
 import sys
 
 from PyQt6.QtCore import Qt, QTimer, QRectF, QPointF
@@ -15,6 +16,7 @@ MIN_WIDTH = 210
 MAX_WIDTH = 460
 MARGIN = 28
 GAP = 10        # between two indicators sharing a corner
+FOLLOW_EVERY = 8   # ticks between two looks for the pointer: about four a second
 
 BG = QColor(22, 24, 29, 238)
 BORDER = QColor(255, 255, 255, 28)
@@ -37,16 +39,68 @@ STATE_COLORS = {"recording": REC, "asking": ASK, "meeting": REC, "busy": BUSY,
 LIVE = ("recording", "asking", "meeting")
 
 
+# KWin's interface, kept once one has been built. See _compositor_screen.
+_kwin = None
+
+
+def _compositor_screen():
+    """The screen KWin says the session is on, or None where nothing says.
+
+    Wayland tells a client where the pointer is only while it is over one of
+    that client's own windows, and the indicator is never under the pointer, so
+    QCursor.pos() answers with a stale point or, when the pointer has never
+    been over a window of ours, with the origin. Either way the indicator lands
+    in the corner of whichever screen holds 0,0 instead of the one being worked
+    on, and on a two-monitor desk that is the wrong screen most of the time.
+    KWin does know, and it names outputs the way Qt names screens, by
+    connector, natively and through XWayland alike. No other Wayland desktop
+    answers this, so the rest are left with the pointer, which is right on X11
+    and wrong on Wayland exactly as before.
+
+    What it answers with is the active output, which is the one under the
+    pointer only where Plasma is set to let the active screen follow the mouse.
+    Under the default, click to focus, it is the focused window's screen, so
+    the indicator lands where the typing is going rather than where the mouse
+    was left. Which is why nothing here, and nothing in the settings window,
+    promises the pointer.
+    """
+    global _kwin
+    if _kwin is None or not _kwin.isValid():
+        # Which also leaves macOS and Windows out, where nothing sets it and
+        # the pointer can be asked where it is like anywhere else.
+        desktop = os.environ.get("XDG_CURRENT_DESKTOP", "").lower()
+        if "kde" not in desktop and "plasma" not in desktop:
+            return None
+        try:
+            from PyQt6.QtDBus import QDBusConnection, QDBusInterface
+            _kwin = QDBusInterface("org.kde.KWin", "/KWin", "org.kde.KWin",
+                                   QDBusConnection.sessionBus())
+        except Exception:
+            return None
+        if not _kwin.isValid():
+            return None
+        # A compositor busy enough not to answer in a fifth of a second is one
+        # the indicator should stop waiting for, not one it should freeze with.
+        _kwin.setTimeout(200)
+    answer = _kwin.call("activeOutputName").arguments()
+    name = answer[0] if answer else ""
+    return next((item for item in QApplication.screens() if item.name() == name),
+                None)
+
+
 class Overlay(QWidget):
     """One indicator. Give it `below` and it stacks on top of that one instead
     of covering it, which is what lets a dictation and a command to the agent be
     under way at the same time and still both be visible."""
 
     def __init__(self, corner="bottom-left", below=None, dismissable=False,
-                 screen_name=""):
+                 screen_name="", follow_pointer=False):
         super().__init__(None)
         self.corner = corner
         self.screen_name = screen_name
+        # Whether it goes on following the pointer once it is up, rather than
+        # settling on the screen it appeared on.
+        self.follow_pointer = follow_pointer
         self.below = below
         # A job that can run for ten minutes should not have to be watched for
         # ten minutes. Clicking such an indicator puts the progress away; the
@@ -65,6 +119,8 @@ class Overlay(QWidget):
         self.seconds = 0.0
         self._phase = 0.0
         self._concealed = True
+        self._shown_on = ""   # the screen it was last put on, by name
+        self._looks = 0       # ticks since the pointer was last looked for
 
         flags = (
             Qt.WindowType.FramelessWindowHint
@@ -252,16 +308,52 @@ class Overlay(QWidget):
                         min(MAX_WIDTH, metrics.horizontalAdvance(self.message) + extra))
         self.resize(width, HEIGHT)
 
-    def _reposition(self):
-        # The screen the settings name, or, when none is named or it is not
-        # plugged in right now, where the user actually is. Names are connector
-        # names on X11 and model names on macOS, where two identical monitors
-        # can share one; the first then wins.
-        screen = next(
+    def _screen(self):
+        """The screen this indicator belongs on right now.
+
+        The one the settings name, or, when none is named or it is not plugged
+        in right now, where the user actually is. Names are connector names on
+        X11 and model names on macOS, where two identical monitors can share
+        one; the first then wins.
+
+        One stacking on another belongs on that one's screen and nowhere else.
+        Asked for itself it would answer where the user is now, which is not
+        where the ribbon it stacks on was put a minute ago, and the pair would
+        end up a monitor apart with this one raised over nothing.
+        """
+        if self.below is not None and self.below.showing:
+            under = next((item for item in QApplication.screens()
+                          if item.name() == self.below._shown_on), None)
+            if under is not None:
+                return under
+        named = next(
             (item for item in QApplication.screens() if item.name() == self.screen_name),
             None,
         )
-        screen = screen or QApplication.screenAt(QCursor.pos()) or QApplication.primaryScreen()
+        return (named or _compositor_screen()
+                or QApplication.screenAt(QCursor.pos())
+                or QApplication.primaryScreen())
+
+    def _wandered_off(self):
+        """Whether the pointer has left the screen the indicator is on.
+
+        Only asked while it is following, and only every few ticks: the answer
+        costs a word with the compositor, and a hand moving a mouse across a
+        desk is slow next to a 33 ms ribbon. Every tick for one that stacks on
+        another, where the answer is free and waiting a third of a second for
+        it would leave the pair split over two monitors for that long.
+        """
+        if not self.follow_pointer or self.screen_name:
+            return False
+        if self.below is None or not self.below.showing:
+            self._looks = (self._looks + 1) % FOLLOW_EVERY
+            if self._looks:
+                return False
+        return self._screen().name() != self._shown_on
+
+    def _reposition(self):
+        screen = self._screen()
+        self._shown_on = screen.name()
         area = screen.availableGeometry()
         left = "left" in self.corner
         top = "top" in self.corner
@@ -277,8 +369,10 @@ class Overlay(QWidget):
     def _tick(self):
         self._phase += 0.12
         # The one underneath can come and go while this one is up; drop back to
-        # the corner when it does rather than leaving a gap where it was.
-        if self.below is not None and self.below.showing != self._stacked:
+        # the corner when it does rather than leaving a gap where it was. And
+        # the screen under the pointer can change while it is up too.
+        moved = self.below is not None and self.below.showing != self._stacked
+        if moved or self._wandered_off():
             self._reposition()
         if self.state in LIVE and not self.paused:
             # keep the ribbon moving even through a pause in speech

dikte/settings_ui.py

@@ -6,7 +6,7 @@ import shutil
 import sys
 import threading
 
-from PyQt6.QtCore import QEvent, QObject, QRect, Qt, QUrl, pyqtSignal
+from PyQt6.QtCore import QEvent, QObject, QRect, Qt, QTimer, QUrl, pyqtSignal
 from PyQt6.QtGui import QDesktopServices, QGuiApplication, QKeySequence, QShortcut
 from PyQt6.QtWidgets import (
     QAbstractItemView, QAbstractSpinBox, QCheckBox, QComboBox, QDialog,
@@ -181,18 +181,34 @@ class WrappedLabel(QLabel):
         super().setText(text)
         self._fit()
 
+    def showEvent(self, event):
+        # Text set while the window was still being built was measured against
+        # nothing; this is the first moment the width means anything.
+        super().showEvent(event)
+        self._fit()
+
     def resizeEvent(self, event):
         super().resizeEvent(event)
         self._fit()
 
     def _fit(self):
+        # A label the layout has not placed yet is a handful of pixels wide,
+        # and wrapping a sentence against that width invents a hundred lines.
+        # The minimum set from it does not stay a minimum either: QLabel folds
+        # it into its own cached size hints and clears that cache only when the
+        # text changes, so the row stands thousands of pixels tall and carries
+        # the model box and everything under it off the bottom of the window
+        # until another publisher is picked. Nothing to measure against yet
+        # means nothing to claim yet, and the show and resize above come back
+        # for it.
+        if not self.isVisible() or self.width() <= 0:
+            return
         # Measured off the font rather than asked of the label, whose own answer
         # is floored by the minimum set here a moment ago and so only ever grows.
-        if self.width() > 0:
-            wrap = Qt.TextFlag.TextWordWrap | Qt.TextFlag.TextWrapAnywhere
-            box = QRect(0, 0, self.width(), 0)
-            self.setMinimumHeight(
-                self.fontMetrics().boundingRect(box, wrap, self.text()).height())
+        wrap = Qt.TextFlag.TextWordWrap | Qt.TextFlag.TextWrapAnywhere
+        box = QRect(0, 0, self.width(), 0)
+        self.setMinimumHeight(
+            self.fontMetrics().boundingRect(box, wrap, self.text()).height())
 
 
 class WheelGuard(QObject):
@@ -247,6 +263,20 @@ class LocalModelBox(QGroupBox):
         self._pending = False
         self._stop = False
         self._wanted = ""              # the model to select once a list arrives
+        self._chosen_in = ""           # the publisher the selected model is from
+        # Whether a list for the publisher on screen has come back. An empty
+        # box before one has is a box nobody has asked anything yet, and the
+        # two read the same without this.
+        self._answered = False
+        # What the last publisher listing held, so that the switch beside the
+        # box can be flipped without asking for it again.
+        self._found_repos = []
+        # Typing or arrowing through the publisher box changes its text a
+        # character at a time, and each of those would otherwise be a request.
+        self._later = QTimer(self)
+        self._later.setSingleShot(True)
+        self._later.setInterval(400)
+        self._later.timeout.connect(self._later_fetch)
 
         form = QFormLayout(self)
 
@@ -256,15 +286,55 @@ class LocalModelBox(QGroupBox):
         form.addRow(t("Program"), self._side_by_side(self.program_label,
                                                      self.install_button))
 
+        # What the model rows are judged against, said out loud. Without it,
+        # "too big for this machine" and the recommendation above the list are
+        # a verdict with no visible reason behind them.
+        self.machine_label = WrappedLabel()
+        self.machine_label.setToolTip(
+            t("A model may take half of this memory, less a gigabyte for the "
+              "context around the weights. Anything past that is marked too "
+              "big; it may still load, on a machine with nothing else open."))
+        form.addRow(t("This machine"), self.machine_label)
+        self._show_machine()
+
         if self._repos is not None:
             self.repo = QComboBox()
             self.repo.setEditable(True)
             self.repo.setToolTip(t("A Hugging Face repository of GGUF files. The "
                                    "list is fetched; any other one can be typed in."))
             self.repo.currentTextChanged.connect(self._repo_changed)
-            form.addRow(t("Publisher"), self.repo)
+            # Forty repository ids is not a choice anybody can make. The few
+            # that were picked for this job are what the box holds until
+            # somebody asks for the rest.
+            self.every_repo = QCheckBox(t("All"))
+            self.every_repo.setToolTip(
+                t("Everything ggml-org publishes, including the models that "
+                  "are too big to run here and the ones that are not for "
+                  "cleaning up text."))
+            self.every_repo.toggled.connect(self._every_repo_changed)
+            form.addRow(t("Publisher"),
+                        self._side_by_side(self.repo, self.every_repo))
+            # A repository id names the publisher, the parameter count and the
+            # shape of the weights, and says nothing about whether it is the
+            # one to click.
+            self.repo_note = WrappedLabel()
+            form.addRow("", self.repo_note)
 
         self.model = QComboBox()
+        self.model.setToolTip(
+            t("large-v3 makes the fewest mistakes and is the slowest of them. "
+              "large-v3-turbo is that model with a four layer decoder in place "
+              "of a thirty-two layer one: several times faster, at one to two "
+              "points of word error in English and about two and a half in "
+              "the other languages. Below those, every step down the list "
+              "trades accuracy for size, and the .en models are trained on "
+              "English alone.")
+            if program is ggml.WHISPER else
+            t("Cleanup is punctuation, capitals and filler words, so what "
+              "these are picked on is following an instruction rather than "
+              "knowing anything. Start at a q4 file; the 16-bit ones are "
+              "several times the memory for a difference this job cannot "
+              "see."))
         self.download_button = QPushButton(t("Download"))
         self.download_button.clicked.connect(self._download)
         self.delete_button = QPushButton(t("Delete"))
@@ -327,14 +397,16 @@ class LocalModelBox(QGroupBox):
         """
         self._wanted = model
         self._pending = True
+        self._answered = False
         self._show_program()
+        self._chosen_in = ""
         if self._repos is not None:
+            suggested = ggml.suggested_llm()
+            self._chosen_in = repo or suggested[0]
             self.repo.blockSignals(True)
-            self.repo.clear()
-            self.repo.addItems(list(ggml.SUGGESTED_LLM))
-            self.repo.setCurrentText(repo or ggml.SUGGESTED_LLM[0])
+            self.repo.setCurrentText(self._chosen_in)
             self.repo.blockSignals(False)
-            self._fit_popup(self.repo)
+            self._fill_repos_box(suggested)
         self._fill_models([])
 
     def showEvent(self, event):
@@ -349,88 +421,263 @@ class LocalModelBox(QGroupBox):
         path = ggml.program_path(self.program)
         if not path:
             self.program_label.setText(t("Not installed."))
+            self.install_button.setText(t("Download"))
             self.install_button.setVisible(True)
             return
-        self.install_button.setVisible(not ggml.installed_program(self.program)
-                                       and not ggml.system_program(self.program))
+        # A copy that is here is not a copy that is right. whisper.cpp releases
+        # every few weeks, and a graphics card installed after Dikte was
+        # changes which build this machine should be running; the button was
+        # hidden the moment anything landed, and nothing else on this window
+        # asks for the download again.
+        self.install_button.setText(t("Download again")
+                                    if ggml.installed_program(self.program)
+                                    else t("Download"))
+        self.install_button.setVisible(not ggml.system_program(self.program))
         if ggml.system_program(self.program):
             # Worth saying which one is running: a distribution package is built
             # for this machine and may reach the graphics card, while the
             # released binaries carry processor backends only.
             self.program_label.setText(t("Installed on the system: {path}", path=path))
+        elif ggml.vulkan_missing(self.program):
+            # The download landed the processor build where the graphics card
+            # one belongs, and nothing else on this window would say so.
+            self.program_label.setText(
+                t("Downloaded, version {version}. There was no Vulkan build, "
+                  "so this one runs on the processor.",
+                  version=ggml.installed_version(self.program) or "?"))
         else:
             self.program_label.setText(
                 t("Downloaded, version {version}.",
                   version=ggml.installed_version(self.program) or "?"))
 
+    def _show_machine(self):
+        where = ggml.accelerator()
+        memory = ggml.total_memory()
+        parts = [t("Graphics: {name}.", name=where) if where else
+                 t("No graphics interface found, so this runs on the processor.")]
+        if memory:
+            parts.append(t("Memory: {size}.", size=ggml.human_size(memory)))
+        self.machine_label.setText(" ".join(parts))
+
     # ---- the lists -------------------------------------------------------
 
     def _fill_repos(self, current):
         def work():
-            self._listed.emit([("repos", ggml.llm_repos())], "")
+            self._listed.emit([("repos", ggml.llm_repos(), "")], "")
 
         threading.Thread(target=work, daemon=True).start()
 
+    def _fill_repos_box(self, found):
+        """The publishers, with the suggested ones kept apart from the rest.
+
+        Forty repositories in one run is a list nobody reads to the end of, and
+        the few worth starting from are lost in it. A separator rather than a
+        heading, because this box is typed into as well as chosen from and a
+        heading would land in the field as though it were a repository.
+        """
+        self._found_repos = found
+        current = self.repo.currentText()
+        # Every suggestion, whether or not it came back in the listing: that
+        # listing is the forty repositories touched most recently, and a
+        # publisher that has not been updated in a season falls off it while
+        # still being the one to point at.
+        first = list(ggml.suggested_llm())
+        rest = [r for r in found if r not in first]
+        if not self.every_repo.isChecked():
+            # The one being used stays on offer whatever the switch says, so
+            # that a repository somebody typed in is not dropped out from
+            # under them by the next fetch.
+            rest = [r for r in rest if r == current]
+        self.repo.blockSignals(True)
+        self.repo.clear()
+        self.repo.addItems(first)
+        if first and rest:
+            self.repo.insertSeparator(self.repo.count())
+        self.repo.addItems(rest)
+        self.repo.setCurrentText(current)
+        self.repo.blockSignals(False)
+        self._fit_popup(self.repo)
+        self._show_repo_note()
+
     def _repo_changed(self):
+        self._show_repo_note()
+        if not self._downloading:
+            self._later.start()
+
+    def _show_repo_note(self):
+        note = ggml.SUGGESTED_LLM_NOTE.get(self.repository(), "")
+        self.repo_note.setText(t(note) if note else "")
+
+    def _every_repo_changed(self):
+        self._fill_repos_box(self._found_repos)
+
+    def _later_fetch(self):
+        # A download that started inside the wait was not there to be seen when
+        # the timer went off, and rebuilding the rows underneath one is exactly
+        # what the guard above is for.
         if not self._downloading:
             self._fetch_models(self.repository())
 
     def _fetch_models(self, repo=""):
+        self._answered = False
         self.status.setText(t("Fetching the model list…"))
 
         def work():
             try:
                 found = self._models(repo) if self._repos is not None else self._models()
-                self._listed.emit([("models", found)], "")
+                self._listed.emit([("models", found, repo)], "")
             except ggml.LocalError as exc:
-                self._listed.emit([], str(exc))
+                self._listed.emit([("models", [], repo)], str(exc))
 
         threading.Thread(target=work, daemon=True).start()
 
     def _on_listed(self, payload, error):
+        kind, found, repo = payload[0] if payload else ("repos", [], "")
+        # A publisher changed while its predecessor's list was still on the way
+        # would otherwise be answered with the wrong models, whichever request
+        # happened to come back last.
+        if kind == "models" and repo != self.repository():
+            return
         if error:
-            self.status.setText(error)
+            # The list is the publisher's, so a failed one leaves the box no
+            # longer showing this publisher's models: emptying it is what keeps
+            # the two boxes saying the same thing. The message goes on after,
+            # because filling the box writes a status of its own.
+            if kind == "models":
+                self._fill_models([])
             self._refresh_buttons()
+            self.status.setText(error)
             return
-        kind, found = payload[0]
         if kind == "repos":
-            current = self.repo.currentText()
-            self.repo.blockSignals(True)
-            self.repo.clear()
-            self.repo.addItems(found)
-            self.repo.setCurrentText(current)
-            self.repo.blockSignals(False)
-            self._fit_popup(self.repo)
+            self._fill_repos_box(found)
             return
+        self._answered = True
         self._fill_models(found)
 
+    def _sections(self, items, best):
+        """[(heading, [Item])] for the rows to show, in the order to show them.
+
+        The list arrives sorted by size and nothing else, which for whisper
+        interleaves the models: `large-v3-turbo-q5_0` lands between the two
+        `medium` quantisations, half a screen away from the turbo model it is a
+        copy of. Grouping puts the choice of model above the choice of
+        quantisation, and the row this machine should take goes on top, where
+        somebody who does not want to make either choice can stop reading.
+        """
+        if not items:
+            return []
+        groups = (ggml.whisper_groups(items) if self.program is ggml.WHISPER
+                  else [("", items)])
+        # A publisher with one file on offer is not a choice, and a row of its
+        # own above the only row there is would be the same model twice.
+        top = [i for i in items if i.name == best] if len(items) > 1 else []
+        if not top:
+            return groups
+        if len(groups) == 1 and not groups[0][0]:
+            groups = [(t("Everything this publisher offers"), groups[0][1])]
+        return [(t("Recommended for this machine"), top)] + groups
+
+    def _suggested(self):
+        """The name to prefer when it is on offer, or "" for whatever fits."""
+        if self.program is not ggml.WHISPER:
+            return ""
+        # A Vulkan loader on the machine is not a card in play when what was
+        # installed is the processor build: recommending the accurate model
+        # off the loader alone would put a 1 GB model on a processor and the
+        # wait for it in front of somebody who asked for a sentence.
+        return ggml.suggested_whisper(
+            graphics="" if ggml.vulkan_missing(self.program) else None)
+
+    def _add_heading(self, text):
+        """A row that names the group under it and cannot be chosen."""
+        self.model.addItem(text)
+        row = self.model.count() - 1
+        font = self.model.font()
+        font.setBold(True)
+        self.model.setItemData(row, font, Qt.ItemDataRole.FontRole)
+        listing = self.model.model()
+        entry = listing.item(row) if hasattr(listing, "item") else None
+        if entry is not None:
+            entry.setEnabled(False)
+
+    def _add_model(self, name, item, best):
+        """One row: the file, what it weighs, and whether it is worth taking."""
+        here = ggml.have_model(self._model_path(name))
+        if here:
+            marks = [t("downloaded")]
+        elif item is None:
+            # Chosen but neither here nor on offer: the file was deleted from
+            # underneath, or the settings came from another machine.
+            marks = [t("not downloaded")]
+        else:
+            marks = [ggml.human_size(item.size)]
+        # `q5_1`, `Q4_K_M`, `MXFP4`, `BF16`: four spellings of the same thing
+        # in one list, and the number is the whole of what any of them says. A
+        # whisper file with no mark at all is the full 16-bit model, which is
+        # the one convention here that a name does not carry.
+        bits = ggml.bit_depth(name) or (16 if self.program is ggml.WHISPER
+                                        else 0)
+        if bits:
+            marks.append(t("{bits}-bit", bits=bits))
+        if ggml.ENGLISH_ONLY in name:
+            marks.append(t("English only"))
+        # The verdicts last, after everything the row is: what to do about the
+        # row rather than what it holds.
+        if item is not None and not here and not ggml.fits(item.size):
+            marks.append(t("too big for this machine"))
+        if name == best:
+            marks.append(t("recommended"))
+        self.model.addItem(f"{name}  ({', '.join(marks)})", name)
+        self.model.setItemData(self.model.count() - 1, item,
+                               Qt.ItemDataRole.UserRole + 1)
+
+    def _first_model(self):
+        """The first row that is a model rather than a heading."""
+        for row in range(self.model.count()):
+            if self.model.itemData(row):
+                return row
+        return -1
+
     def _fill_models(self, items):
-        """One row per model, saying what it weighs and whether it is here."""
-        wanted = self._wanted or self.selected()
-        here = [name for name in (self._model_path(i.name).name for i in items)]
+        """One row per model, grouped, saying what it weighs and where it is."""
+        # The selection is only worth carrying over within the publisher it was
+        # made in. Carried across one, a model this repository does not publish
+        # would be added back as "not downloaded" and selected again, and
+        # changing the publisher would leave the model box looking untouched.
+        same = self._repos is None or self.repository() == self._chosen_in
+        wanted = self._wanted or (self.selected() if same else "")
+        best = ggml.recommended(items, self._suggested()) if items else ""
         self.model.blockSignals(True)
         self.model.clear()
-        for item, name in zip(items, here):
-            mark = (t("downloaded") if ggml.have_model(self._model_path(item.name))
-                    else ggml.human_size(item.size))
-            self.model.addItem(f"{name}  ({mark})", name)
-            self.model.setItemData(self.model.count() - 1, item, Qt.ItemDataRole.UserRole + 1)
+        listed = set()
+        for heading, group in self._sections(items, best):
+            if heading:
+                self._add_heading(heading)
+            for item in group:
+                name = self._model_path(item.name).name
+                self._add_model(name, item, best)
+                listed.add(name)
         # A model that was downloaded and then dropped from the list upstream is
-        # still on this disk and still works, so it stays on offer.
-        for name in self._on_disk():
-            if self.model.findData(name) < 0:
-                self.model.addItem(f"{name}  ({t('downloaded')})", name)
-        # And one that is chosen but not here, because the file was deleted from
-        # underneath or the settings came from another machine, stays chosen:
-        # Save reads this box, and a row missing here would quietly empty the
-        # setting rather than showing that the model needs downloading again.
-        if wanted and self.model.findData(wanted) < 0:
-            self.model.addItem(f"{wanted}  ({t('not downloaded')})", wanted)
+        # still on this disk and still works, so it stays on offer. So does one
+        # that is chosen but not here: Save reads this box, and a row missing
+        # here would quietly empty the setting rather than showing that the
+        # model needs downloading again.
+        extras = [(t("Already on this machine"),
+                   [name for name in self._on_disk() if name not in listed])]
+        if wanted and wanted not in listed \
+                and not ggml.have_model(self._model_path(wanted)):
+            extras.append((t("Chosen, but not downloaded"), [wanted]))
+        for heading, names in extras:
+            if names and listed:
+                self._add_heading(heading)
+            for name in names:
+                self._add_model(name, None, best)
         index = self.model.findData(wanted)
-        self.model.setCurrentIndex(max(index, 0))
+        self.model.setCurrentIndex(index if index >= 0 else self._first_model())
         self.model.blockSignals(False)
         self._fit_popup(self.model)
         self._wanted = ""
+        self._chosen_in = self.repository()
         self._model_changed()
 
     def _on_disk(self):
@@ -459,6 +706,9 @@ class LocalModelBox(QGroupBox):
         self._show_program()
         if error:
             self.program_label.setText(error)
+        # The model line says whether the program is here, so installing one
+        # changes what it should read.
+        self._refresh_buttons()
         self.changed.emit()
 
     def _current_item(self):
@@ -467,12 +717,21 @@ class LocalModelBox(QGroupBox):
     def _download(self):
         if self._downloading:
             self._stop = True
+            # The flag is only read between blocks, and the wait for the server
+            # to answer is not between blocks: a click during it changes
+            # nothing on screen for as long as the connection takes.
+            self.status.setText(t("Stopping…"))
             return
         item = self._current_item()
         if item is None:
             return
         self._downloading, self._stop = True, False
         self._refresh_buttons()
+        # Opening the connection can take ten or twenty seconds, and the first
+        # byte counts are what the line below would otherwise wait for. Left
+        # saying "not downloaded yet" beside a button that now reads Stop, a
+        # download that started looks like a click that did not register.
+        self.status.setText(t("Starting the download…"))
 
         def work():
             try:
@@ -518,10 +777,17 @@ class LocalModelBox(QGroupBox):
 
     def _fill_models_from_current(self):
         """Redraw the rows without asking anybody anything again."""
-        items = [self.model.itemData(i, Qt.ItemDataRole.UserRole + 1)
-                 for i in range(self.model.count())]
+        # By name, because the recommended model has a row of its own at the
+        # top as well as one in its group, and reading the rows back twice
+        # would double it in the list every time a download finished.
+        items, seen = [], set()
+        for row in range(self.model.count()):
+            item = self.model.itemData(row, Qt.ItemDataRole.UserRole + 1)
+            if item is not None and item.name not in seen:
+                seen.add(item.name)
+                items.append(item)
         self._wanted = self.selected()
-        self._fill_models([i for i in items if i is not None])
+        self._fill_models(items)
 
     def _delete(self):
         name = self.selected()
@@ -545,15 +811,41 @@ class LocalModelBox(QGroupBox):
     def _refresh_buttons(self):
         name = self.selected()
         here = bool(name) and ggml.have_model(self._model_path(name))
+        # A row carries what it takes to fetch it. The ones that do not are the
+        # models found on this disk and the one the settings name but the list
+        # does not offer: there is nothing to press Download for on those, and
+        # a button that can only do nothing is worse than one that is out.
+        item = self._current_item()
         self.delete_button.setEnabled(here and not self._downloading)
         self.download_button.setText(t("Stop") if self._downloading else t("Download"))
-        self.download_button.setEnabled(self._downloading or (bool(name) and not here))
+        self.download_button.setEnabled(self._downloading or (item is not None
+                                                              and not here))
         if self._downloading:
             return
-        if not name:
+        if not name and self._repos is not None and self._answered \
+                and self._first_model() < 0:
+            # An empty box under a publisher that answered perfectly well: what
+            # it publishes is split across files, past the size cap, or a
+            # projector or draft head rather than a model of its own. Said
+            # nowhere, it read as though the click had not registered.
+            self.status.setText(
+                t("{repo} publishes nothing that can be run here. Its models "
+                  "are split across files, larger than {cap}, or pieces of a "
+                  "model rather than one. Choose another publisher.",
+                  repo=self.repository(), cap=ggml.human_size(ggml.GGUF_MAX_BYTES)))
+        elif not name:
             self.status.setText(t("Nothing downloaded yet."))
+        elif here and not ggml.program_path(self.program):
+            # The model alone runs nothing, and "Ready" over a missing program
+            # reads as though it does.
+            self.status.setText(t("{name} is here, but the program above is "
+                                  "not. Download it first.", name=name))
         elif here:
             self.status.setText(t("Ready: {name}.", name=name))
+        elif item is None:
+            self.status.setText(t("{name} is not on this machine and this "
+                                  "publisher does not offer it. Choose another "
+                                  "model, or another publisher.", name=name))
         else:
             self.status.setText(t("{name} has not been downloaded yet.", name=name))
 
@@ -801,7 +1093,11 @@ class SettingsWindow(QDialog):
         form = QFormLayout(page)
 
         self.indicator_screen = QComboBox()
-        self.indicator_screen.addItem(t("Follow the mouse pointer"), "")
+        # The active screen rather than the pointer, for the reason in
+        # overlay._compositor_screen: it is what a compositor will answer for,
+        # and on Plasma the two are one screen only where the active screen is
+        # set to follow the mouse.
+        self.indicator_screen.addItem(t("Follow the active screen"), "")
         for screen in QGuiApplication.screens():
             # The native resolution, so that a scaled 4K screen reads
             # 3840 × 2160 and not the 1920 × 1080 Qt sees through the scale.
@@ -815,12 +1111,26 @@ class SettingsWindow(QDialog):
             )
         form.addRow(t("Indicator screen"), self.indicator_screen)
 
+        # Only the screen it appeared on is decided when it appears; this is
+        # what makes it keep up with a session that moves to another one
+        # mid-recording. The active screen and not the pointer, because that is
+        # what a compositor will answer for: on Plasma the two are the same
+        # screen only where the active screen is set to follow the mouse, and
+        # otherwise it is the focused window that decides. Nothing to offer
+        # when a screen is named above, since that name is the whole answer.
+        self.follow_pointer = QCheckBox(t("Move it when the active screen changes"))
+        self.indicator_screen.currentIndexChanged.connect(self._sync_follow_pointer)
+        form.addRow("", self.follow_pointer)
+
         self.corner = QComboBox()
         for value in CORNERS:
             self.corner.addItem(t(value), value)
         form.addRow(t("Indicator corner"), self.corner)
         return page
 
+    def _sync_follow_pointer(self):
+        self.follow_pointer.setEnabled(not self.indicator_screen.currentData())
+
     def _api_tab(self):
         page = QWidget()
         outer = QVBoxLayout(page)
@@ -864,6 +1174,15 @@ class SettingsWindow(QDialog):
         self.transcribe_model_row = self._row(self.transcribe_model,
                                               self.refresh_transcribe_models)
         stt_form.addRow(t("Model"), self.transcribe_model_row)
+        # OpenRouter only: which of its models a timestamped run asks for.
+        self.file_model = QComboBox()
+        self.file_model.setEditable(True)
+        self.file_model.lineEdit().setPlaceholderText(api.OPENROUTER_FILE_MODEL)
+        self.file_model.setToolTip(
+            t("The model a timestamped audio file (subtitles) is sent to. Not every model "
+              "on OpenRouter returns segment times; empty means openai/whisper-1."))
+        self.file_model_row = self._row(self.file_model)
+        stt_form.addRow(t("Audio file model"), self.file_model_row)
         # A spanning row: in the narrow field column a wrapped label gets a
         # height that fits one line, and the rest of the text is cut off.
         self.transcribe_status = QLabel("")
@@ -1006,9 +1325,47 @@ class SettingsWindow(QDialog):
         orr_form.addRow(self.local_llm_options)
 
         outer.addWidget(orr)
+
+        # One box for both servers rather than a row inside each: what is being
+        # decided is whether this machine keeps gigabytes tied up between
+        # dictations, and that is not a question anybody wants to answer once
+        # per model.
+        self.local_box = QGroupBox(t("Models on this machine"))
+        local_form = QFormLayout(self.local_box)
+        self.local_idle_unload = QCheckBox(t("Unload a model that is sitting unused"))
+        self.local_idle_unload.setToolTip(
+            t("A loaded model holds its memory whether anything is using it or "
+              "not: over a gigabyte for whisper, several for an LLM. Unloading "
+              "gives that back to the rest of the desktop, and the next "
+              "dictation loads it again at the cost of the seconds that takes."))
+        self.local_idle_minutes = QSpinBox()
+        self.local_idle_minutes.setRange(1, 720)
+        self.local_idle_minutes.valueChanged.connect(self._idle_suffix)
+        self._idle_suffix(self.local_idle_minutes.value())
+        self.local_idle_unload.toggled.connect(self.local_idle_minutes.setEnabled)
+        local_form.addRow("", self.local_idle_unload)
+        local_form.addRow(t("After"), self.local_idle_minutes)
+        outer.addWidget(self.local_box)
+
         outer.addStretch(1)
         return page
 
+    def _idle_suffix(self, minutes):
+        """The spin box's own noun, since its lowest value is one of them.
+
+        Turkish is handed both and translates them the same: a number there is
+        followed by the singular however many it counts.
+        """
+        self.local_idle_minutes.setSuffix(
+            t(" minute") if minutes == 1 else t(" minutes"))
+
+    def _refresh_local_box(self):
+        """The idle unload is only on screen when something here runs locally."""
+        self.local_box.setVisible(
+            (self.transcribe_provider.currentData() or "local") == "local"
+            or (self.cleanup_provider.currentData() or "openrouter") == "local"
+        )
+
     def _prompt_tab(self):
         page = QWidget()
         layout = QVBoxLayout(page)
@@ -1730,6 +2087,8 @@ class SettingsWindow(QDialog):
         if screen_name and self.indicator_screen.findData(screen_name) < 0:
             self.indicator_screen.addItem(t("{name} (not connected)", name=screen_name), screen_name)
         self._select_data(self.indicator_screen, screen_name)
+        self.follow_pointer.setChecked(conf["overlay_follows_pointer"])
+        self._sync_follow_pointer()
         self._select_data(self.corner, conf["overlay_corner"])
         self.max_seconds.setValue(conf["max_seconds"])
         self.skip_silent.setChecked(conf["skip_silent"])
@@ -1747,6 +2106,7 @@ class SettingsWindow(QDialog):
         self._shown_provider = ""
         self._select_data(self.transcribe_provider, conf["transcribe_provider"])
         self._provider_changed()  # selecting index 0 fires no signal
+        self.file_model.setCurrentText(conf["openrouter_file_model"])
         self.local_gpu.setChecked(conf["local_gpu"])
         self.local_preload.setChecked(conf["local_preload"])
         self.local_threads.setValue(int(conf["local_threads"]))
@@ -1772,6 +2132,9 @@ class SettingsWindow(QDialog):
         self.local_llm_preload.setChecked(conf["local_llm_preload"])
         self._select_data(self.local_llm_reasoning, conf["local_llm_reasoning"])
         self.local_llm.load(conf["local_llm_model"], conf["local_llm_repo"])
+        self.local_idle_unload.setChecked(conf["local_idle_unload"])
+        self.local_idle_minutes.setValue(int(conf["local_idle_minutes"]))
+        self.local_idle_minutes.setEnabled(conf["local_idle_unload"])
         # The defaults as they read NOW, kept for the save comparison: after a
         # language switch the boxes still hold the old language's default, and
         # comparing against the new one would store that text as a custom
@@ -1849,6 +2212,9 @@ class SettingsWindow(QDialog):
         conf["paste_shortcut"] = self.paste_shortcut.currentText().strip()
         conf["restore_clipboard"] = self.restore_clipboard.isChecked()
         conf["overlay_screen"] = self.indicator_screen.currentData() or ""
+        # Read even while it is greyed out, so that naming a screen and taking
+        # the name back again does not clear a preference nobody touched.
+        conf["overlay_follows_pointer"] = self.follow_pointer.isChecked()
         conf["overlay_corner"] = self.corner.currentData() or "bottom-left"
         conf["max_seconds"] = self.max_seconds.value()
         conf["skip_silent"] = self.skip_silent.isChecked()
@@ -1864,6 +2230,7 @@ class SettingsWindow(QDialog):
         for name, who in cfg.TRANSCRIBERS.items():
             conf[who.key] = self._key_fields[name].text().strip()
             conf[who.model] = self._models[name].strip() or cfg.DEFAULTS[who.model]
+        conf["openrouter_file_model"] = self.file_model.currentText().strip()
         conf["gemini_api_key"] = self.gemini_key.text().strip()
         conf["opencode_api_key"] = self.opencode_key.text().strip()
         conf["local_model"] = self.local_whisper.selected()
@@ -1899,6 +2266,8 @@ class SettingsWindow(QDialog):
         conf["local_llm_gpu"] = self.local_llm_gpu.isChecked()
         conf["local_llm_preload"] = self.local_llm_preload.isChecked()
         conf["local_llm_reasoning"] = self.local_llm_reasoning.currentData() or ""
+        conf["local_idle_unload"] = self.local_idle_unload.isChecked()
+        conf["local_idle_minutes"] = self.local_idle_minutes.value()
 
         # Store an empty prompt when it matches a default: the one it was
         # loaded with, or today's (a Reset click in a session that switched
@@ -2037,16 +2406,26 @@ class SettingsWindow(QDialog):
         self._shown_provider = provider
         local = provider == "local"
         self.stt_form.setRowVisible(self.transcribe_model_row, not local)
+        self.stt_form.setRowVisible(self.file_model_row, provider == "openrouter")
         self.stt_form.setRowVisible(self.transcribe_status, not local)
         self.stt_form.setRowVisible(self.local_whisper, local)
         self.stt_form.setRowVisible(self.local_options, local)
+        self._refresh_local_box()
         if local:
             return
         self.transcribe_model.clear()
         self.transcribe_model.addItems(TRANSCRIBE_MODELS[provider])
         self.transcribe_model.setCurrentText(self._models[provider])
+        if provider == "openrouter":
+            self._fill_file_models(TRANSCRIBE_MODELS[provider])
         self.transcribe_status.setText("")
 
+    def _fill_file_models(self, models):
+        current = self.file_model.currentText()
+        self.file_model.clear()
+        self.file_model.addItems(models)
+        self.file_model.setCurrentText(current)
+
     def _load_transcribe_models(self):
         """The model list of whichever provider is selected."""
         provider = self.transcribe_provider.currentData() or "openai"
@@ -2075,6 +2454,8 @@ class SettingsWindow(QDialog):
         self.transcribe_model.clear()
         self.transcribe_model.addItems(models)
         self.transcribe_model.setCurrentText(current)
+        if self._shown_provider == "openrouter":
+            self._fill_file_models(models)
         self.transcribe_status.setText(t("{count} models loaded.", count=len(models)))
 
     def _load_models(self):
@@ -2538,6 +2919,7 @@ class SettingsWindow(QDialog):
                                         provider != "local")
         self.cleanup_form.setRowVisible(self.local_llm, provider == "local")
         self.cleanup_form.setRowVisible(self.local_llm_options, provider == "local")
+        self._refresh_local_box()
         binary = cleanup.executable(provider)
         found = shutil.which(binary) if binary else ""
         if provider == "local":

packaging/whisper-vulkan/Dockerfile.build

@@ -0,0 +1,44 @@
+FROM ubuntu@sha256:2edbbc5dc405e9612ba3584ce95480277e3eb374407b5505fe26f17df77c7dbc
+
+ARG DEBIAN_FRONTEND=noninteractive
+ARG CMAKE_VERSION=3.31.6
+ARG CMAKE_SHA256=5a1133ff103c71eb5120e2cc3de922733e7d8a26a98ae716397e8676adb367bf
+
+COPY lunarg-signing-key-pub.asc /tmp/lunarg.asc
+
+RUN set -eux; \
+    test "$(sha256sum /tmp/lunarg.asc | cut -d' ' -f1)" = aa1c3c29673140e77f0d6a9aaeed5d9b5621e305ead51c59fae4458bbb4df92b; \
+    apt-get update; \
+    apt-get install --no-install-recommends -y \
+      build-essential=12.9ubuntu3 \
+      ca-certificates \
+      curl \
+      file \
+      git \
+      gnupg \
+      ninja-build=1.10.1-1 \
+      patchelf=0.14.3-1 \
+      python3 \
+      xz-utils; \
+    install -d -m 0755 /usr/share/keyrings; \
+    gpg --dearmor -o /usr/share/keyrings/lunarg.gpg /tmp/lunarg.asc; \
+    printf '%s\n' 'deb [signed-by=/usr/share/keyrings/lunarg.gpg] https://packages.lunarg.com/vulkan jammy main' \
+      > /etc/apt/sources.list.d/lunarg-vulkan.list; \
+    apt-get update; \
+    apt-get install --no-install-recommends -y \
+      libvulkan-dev=1.4.313.0~rc1-1lunarg22.04-1 \
+      vulkan-headers=1.4.313.0~rc1-1lunarg22.04-1 \
+      shaderc=2025.2~rc1-1lunarg22.04-1 \
+      spirv-headers=1.6.1+1.4.313.0~rc1-1lunarg22.04-1; \
+    curl --fail --location --retry 3 \
+      "https://github.com/Kitware/CMake/releases/download/v${CMAKE_VERSION}/cmake-${CMAKE_VERSION}-linux-x86_64.tar.gz" \
+      -o /tmp/cmake.tar.gz; \
+    test "$(sha256sum /tmp/cmake.tar.gz | cut -d' ' -f1)" = "$CMAKE_SHA256"; \
+    tar -xzf /tmp/cmake.tar.gz --strip-components=1 -C /usr/local; \
+    rm -rf /var/lib/apt/lists/* /tmp/cmake.tar.gz /tmp/lunarg.asc; \
+    cmake --version; \
+    glslc --version; \
+    test -f /usr/include/vulkan/vulkan.h; \
+    test -f /usr/share/cmake/SPIRV-Headers/SPIRV-HeadersConfig.cmake
+
+WORKDIR /work

packaging/whisper-vulkan/Dockerfile.runtime-cpu

@@ -0,0 +1,6 @@
+FROM ubuntu@sha256:2edbbc5dc405e9612ba3584ce95480277e3eb374407b5505fe26f17df77c7dbc
+ARG DEBIAN_FRONTEND=noninteractive
+RUN apt-get update \
+ && apt-get install --no-install-recommends -y ca-certificates curl libstdc++6 \
+ && rm -rf /var/lib/apt/lists/*
+WORKDIR /bundle

packaging/whisper-vulkan/Dockerfile.runtime-noicd

@@ -0,0 +1,9 @@
+FROM ubuntu@sha256:2edbbc5dc405e9612ba3584ce95480277e3eb374407b5505fe26f17df77c7dbc
+ARG DEBIAN_FRONTEND=noninteractive
+# The loader and nothing behind it: the machine that has libvulkan because
+# something else pulled it in, and no driver to go with it.
+RUN apt-get update \
+ && apt-get install --no-install-recommends -y \
+      ca-certificates curl libstdc++6 libvulkan1 \
+ && rm -rf /var/lib/apt/lists/*
+WORKDIR /bundle

packaging/whisper-vulkan/Dockerfile.runtime-vulkan

@@ -0,0 +1,7 @@
+FROM ubuntu@sha256:2edbbc5dc405e9612ba3584ce95480277e3eb374407b5505fe26f17df77c7dbc
+ARG DEBIAN_FRONTEND=noninteractive
+RUN apt-get update \
+ && apt-get install --no-install-recommends -y \
+      ca-certificates curl libstdc++6 libvulkan1 mesa-vulkan-drivers vulkan-tools \
+ && rm -rf /var/lib/apt/lists/*
+WORKDIR /bundle

packaging/whisper-vulkan/README.md

@@ -0,0 +1,49 @@
+# The Vulkan whisper-server bundle
+
+whisper.cpp publishes a CPU-only archive for Linux, so the graphics card on a
+Linux machine is out of reach through the Download button. This directory
+builds the archive upstream does not: `whisper-server` with a dynamic Vulkan
+backend next to the CPU ones, for x86_64, against the Ubuntu 22.04 runtime
+contract.
+
+It is published as a release of Dikte's own, `whisper.cpp-v<version>`, marked
+as a prerelease and kept off Latest so that neither the update check nor the
+download page picks it up. `dikte/ggml.py` fetches it by tag and installs it
+only when the archive's digest is the reviewed one; anything else falls back
+to upstream's CPU archive, and the settings window says when it did.
+
+## Publishing a new bundle
+
+1. Enable GitHub's immutable releases setting for the repository, and give the
+   `dependency-release` environment a required reviewer. Both are repository
+   settings, not something this workflow can do for itself.
+2. Run **whisper.cpp Vulkan bundle** on `master` with the new version and its
+   peeled commit, `expected_sha256` empty and `publish: false`. The run builds
+   the archive and reports its digest; without a reviewed digest it refuses to
+   publish, which is what the first run is for.
+3. Review that digest against a build of your own, then run the workflow again
+   with the same version and commit, `expected_sha256` set to it, and
+   `publish: true`. Approve the environment when it asks.
+4. Write the same version, tag and digest into `MANAGED_WHISPER_RELEASE`,
+   `MANAGED_WHISPER_VERSION` and `MANAGED_WHISPER_SHA256` in `dikte/ggml.py`,
+   and into `REVIEWED_WHISPER_VERSION` and `REVIEWED_WHISPER_SHA256` in the
+   workflow. `tests/test_packaging.py` holds the two sides together.
+5. Ship a Dikte release. Until one goes out, nobody's Dikte knows the new
+   bundle exists.
+
+## What this costs, and what it does not promise
+
+The digest lives in Dikte's source, so a backend update is a Dikte release.
+Linux x86_64 machines with a Vulkan loader stay on the pinned whisper.cpp
+version until step 5 happens, while every other platform follows upstream's
+newest release on its own. That is the deliberate trade: an executable Dikte
+downloads is not allowed to change without a reviewed digest behind it.
+
+The build is deterministic between two runs of the same builder, not across
+time. The base image, the CMake tarball, the LunarG packages and the direct
+apt packages are pinned by digest or version, but the Ubuntu and LunarG
+repository metadata behind them is not, and LunarG drops superseded packages.
+A rebuild months later can fail to resolve, or resolve to something that
+produces a different digest. Treat the published archive as the artifact, not
+as something reproducible on demand: a version bump means building,
+validating, reviewing the new digest and updating the pinned tuple together.

packaging/whisper-vulkan/build-package.sh

@@ -0,0 +1,128 @@
+#!/usr/bin/env bash
+set -euo pipefail
+shopt -s nullglob
+
+: "${SOURCE_DIR:=/src}"
+: "${OUT_DIR:=/work/out}"
+: "${WHISPER_VERSION:=1.9.3}"
+: "${WHISPER_COMMIT:=371b5a7561823ab2bb32142d2751e35e7534727b}"
+: "${SOURCE_DATE_EPOCH:=1787219223}"
+
+export SOURCE_DATE_EPOCH TZ=UTC LC_ALL=C LANG=C
+asset=whisper-bin-ubuntu-vulkan-x64
+build=/work/build
+source_copy=/work/source
+root="$OUT_DIR/root/$asset"
+
+rm -rf "$build" "$source_copy" "$OUT_DIR"
+mkdir -p "$build" "$root/LICENSES"
+# Upstream configures bindings/javascript/package.json in the source directory.
+# Build a private copy so the checked-out, verified source remains untouched.
+cp -a "$SOURCE_DIR" "$source_copy"
+chmod -R u+w "$source_copy"
+git config --global --add safe.directory "$source_copy"
+
+cmake -S "$source_copy" -B "$build" -G Ninja \
+  -DCMAKE_BUILD_TYPE=Release \
+  -DCMAKE_BUILD_RPATH='$ORIGIN' \
+  -DCMAKE_INSTALL_RPATH='$ORIGIN' \
+  -DCMAKE_BUILD_WITH_INSTALL_RPATH=ON \
+  -DCMAKE_C_FLAGS="-ffile-prefix-map=$source_copy=. -fdebug-prefix-map=$source_copy=. -fmacro-prefix-map=$source_copy=." \
+  -DCMAKE_CXX_FLAGS="-ffile-prefix-map=$source_copy=. -fdebug-prefix-map=$source_copy=. -fmacro-prefix-map=$source_copy=." \
+  -DBUILD_SHARED_LIBS=ON \
+  -DGGML_BACKEND_DL=ON \
+  -DGGML_CPU_ALL_VARIANTS=ON \
+  -DGGML_NATIVE=OFF \
+  -DGGML_CCACHE=OFF \
+  -DGGML_OPENMP=OFF \
+  -DGGML_VULKAN=ON \
+  -DWHISPER_BUILD_EXAMPLES=ON \
+  -DWHISPER_BUILD_SERVER=ON \
+  -DWHISPER_BUILD_TESTS=OFF \
+  -DWHISPER_BUILD_IS_DEV=OFF \
+  -DWHISPER_CURL=OFF \
+  -DWHISPER_SDL2=OFF \
+  -DWHISPER_COMMON_FFMPEG=OFF \
+  -DWHISPER_BUILD_COMMIT="$WHISPER_COMMIT" \
+  -DWHISPER_BUILD_NUMBER=0
+cmake --build "$build" --target whisper-server --parallel "$(nproc)"
+
+# Package an allowlist, not everything examples/ happens to build in the future.
+cp -a "$build/bin/whisper-server" "$root/"
+cp -a "$build/bin"/libwhisper.so* "$root/"
+cp -a "$build/bin"/libggml.so* "$root/"
+cp -a "$build/bin"/libggml-base.so* "$root/"
+cp -a "$build/bin"/libggml-cpu*.so* "$root/"
+cp -a "$build/bin"/libggml-vulkan.so* "$root/"
+
+# Strip real ELF files only; preserve the SONAME symlink chains.
+while IFS= read -r -d '' file; do
+  if file "$file" | grep -q ELF; then
+    strip --strip-unneeded "$file"
+    patchelf --set-rpath '$ORIGIN' "$file"
+  fi
+done < <(find "$root" -type f -print0)
+
+cp "$SOURCE_DIR/LICENSE" "$root/LICENSES/whisper.cpp-MIT.txt"
+cp /packaging/licenses/cpp-httplib-MIT.txt "$root/LICENSES/"
+cp /packaging/licenses/nlohmann-json-MIT.txt "$root/LICENSES/"
+
+cat > "$root/BUILD-INFO.json" <<EOF
+{
+  "asset": "$asset.tar.gz",
+  "source": "https://github.com/ggml-org/whisper.cpp",
+  "source_version": "v$WHISPER_VERSION",
+  "source_commit": "$WHISPER_COMMIT",
+  "source_date_epoch": $SOURCE_DATE_EPOCH,
+  "build_platform": "ubuntu-22.04-x86_64",
+  "base_image": "ubuntu@sha256:2edbbc5dc405e9612ba3584ce95480277e3eb374407b5505fe26f17df77c7dbc",
+  "cmake": "3.31.6",
+  "cmake_flags": [
+    "BUILD_SHARED_LIBS=ON",
+    "C/CXX_FILE_PREFIX_MAP=/work/source=.",
+    "GGML_BACKEND_DL=ON",
+    "GGML_CPU_ALL_VARIANTS=ON",
+    "GGML_NATIVE=OFF",
+    "GGML_CCACHE=OFF",
+    "GGML_OPENMP=OFF",
+    "GGML_VULKAN=ON",
+    "WHISPER_BUILD_EXAMPLES=ON",
+    "WHISPER_BUILD_SERVER=ON",
+    "WHISPER_BUILD_TESTS=OFF",
+    "WHISPER_BUILD_IS_DEV=OFF",
+    "WHISPER_CURL=OFF",
+    "WHISPER_SDL2=OFF",
+    "WHISPER_COMMON_FFMPEG=OFF"
+  ],
+  "runtime_contract": {
+    "minimum_glibc": "2.34",
+    "minimum_glibcxx": "3.4.30",
+    "required": ["x86_64 Linux", "glibc", "libstdc++.so.6", "libgcc_s.so.1"],
+    "optional_gpu": ["libvulkan.so.1", "a working Vulkan ICD"],
+    "cpu_fallback": "dynamic CPU backends are included; -ng forces CPU"
+  }
+}
+EOF
+
+# A deterministic CycloneDX sidecar generated from the files actually shipped.
+ROOT="$root" VERSION="$WHISPER_VERSION" COMMIT="$WHISPER_COMMIT" EPOCH="$SOURCE_DATE_EPOCH" \
+python3 /packaging/make-sbom.py > "$root/$asset.cdx.json"
+
+(
+  cd "$root"
+  find . -type f ! -name SHA256SUMS -print0 \
+    | sort -z \
+    | xargs -0 sha256sum
+) > "$root/SHA256SUMS"
+
+mkdir -p "$OUT_DIR"
+tar --sort=name --owner=0 --group=0 --numeric-owner \
+  --mtime="@$SOURCE_DATE_EPOCH" \
+  --pax-option=delete=atime,delete=ctime \
+  -C "$OUT_DIR/root" -cf - "$asset" \
+  | gzip -n -9 > "$OUT_DIR/$asset.tar.gz"
+(
+  cd "$OUT_DIR"
+  sha256sum "$asset.tar.gz" > "$asset.tar.gz.sha256"
+)
+cp "$root/$asset.cdx.json" "$OUT_DIR/$asset.cdx.json"

packaging/whisper-vulkan/licenses/cpp-httplib-MIT.txt

@@ -0,0 +1,21 @@
+The MIT License (MIT)
+
+Copyright (c) 2017 yhirose
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

packaging/whisper-vulkan/licenses/nlohmann-json-MIT.txt

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2013-2022 Niels Lohmann
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

packaging/whisper-vulkan/lunarg-signing-key-pub.asc

@@ -0,0 +1,31 @@
+-----BEGIN PGP PUBLIC KEY BLOCK-----
+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+=nXAe
+-----END PGP PUBLIC KEY BLOCK-----

packaging/whisper-vulkan/make-sbom.py

@@ -0,0 +1,116 @@
+#!/usr/bin/env python3
+import datetime
+import hashlib
+import json
+import os
+import uuid
+from pathlib import Path
+
+root = Path(os.environ["ROOT"])
+version = os.environ["VERSION"]
+commit = os.environ["COMMIT"]
+epoch = int(os.environ["EPOCH"])
+asset = "whisper-bin-ubuntu-vulkan-x64"
+sbom_path = root / f"{asset}.cdx.json"
+
+def digest(path):
+    h = hashlib.sha256()
+    with path.open("rb") as stream:
+        for block in iter(lambda: stream.read(1024 * 1024), b""):
+            h.update(block)
+    return h.hexdigest()
+
+files = []
+for path in sorted(root.rglob("*")):
+    if path != sbom_path and path.is_file() and not path.is_symlink():
+        rel = path.relative_to(root).as_posix()
+        files.append({
+            "type": "file",
+            "bom-ref": f"file:{rel}",
+            "name": rel,
+            "hashes": [{"alg": "SHA-256", "content": digest(path)}],
+        })
+
+ts = datetime.datetime.fromtimestamp(
+    epoch, datetime.timezone.utc,
+).isoformat().replace("+00:00", "Z")
+root_ref = f"pkg:github/ggml-org/whisper.cpp@{version}?commit={commit}"
+ggml_ref = "pkg:github/ggml-org/[email protected]"
+httplib_ref = "pkg:github/yhirose/[email protected]"
+json_ref = "pkg:github/nlohmann/[email protected]"
+
+sbom = {
+    "bomFormat": "CycloneDX",
+    "specVersion": "1.6",
+    "serialNumber": f"urn:uuid:{uuid.uuid5(uuid.NAMESPACE_URL, root_ref)}",
+    "version": 1,
+    "metadata": {
+        "timestamp": ts,
+        "tools": {"components": [
+            {"type": "application", "name": "make-sbom.py", "version": "1"},
+            {"type": "application", "name": "CMake", "version": "3.31.6"},
+            {"type": "application", "name": "glslc", "version": "2025.2"},
+        ]},
+        "component": {
+            "type": "application",
+            "bom-ref": root_ref,
+            "group": "ggml-org",
+            "name": "whisper-server",
+            "version": version,
+            "purl": root_ref,
+            "licenses": [{"expression": "MIT"}],
+            "externalReferences": [{
+                "type": "vcs",
+                "url": f"https://github.com/ggml-org/whisper.cpp/tree/{commit}",
+            }],
+            "properties": [
+                {"name": "dikte:asset-name", "value": f"{asset}.tar.gz"},
+                {"name": "dikte:source-commit", "value": commit},
+                {"name": "dikte:runtime:glibc-minimum", "value": "2.34"},
+                {"name": "dikte:runtime:glibcxx-minimum", "value": "3.4.30"},
+                {"name": "dikte:runtime:vulkan-loader", "value": "optional; libvulkan.so.1"},
+            ],
+        },
+    },
+    "components": [
+        {
+            "type": "library",
+            "bom-ref": ggml_ref,
+            "group": "ggml-org",
+            "name": "ggml",
+            "version": "0.20.2",
+            "purl": ggml_ref,
+            "licenses": [{"expression": "MIT"}],
+            "properties": [{
+                "name": "dikte:source",
+                "value": "vendored by the pinned whisper.cpp commit",
+            }],
+        },
+        {
+            "type": "library",
+            "bom-ref": httplib_ref,
+            "group": "yhirose",
+            "name": "cpp-httplib",
+            "version": "0.20.0",
+            "purl": httplib_ref,
+            "licenses": [{"expression": "MIT"}],
+        },
+        {
+            "type": "library",
+            "bom-ref": json_ref,
+            "group": "nlohmann",
+            "name": "json",
+            "version": "3.11.2",
+            "purl": json_ref,
+            "licenses": [{"expression": "MIT"}],
+        },
+        *files,
+    ],
+    "dependencies": [{
+        "ref": root_ref,
+        "dependsOn": [ggml_ref, httplib_ref, json_ref]
+                     + [item["bom-ref"] for item in files],
+    }],
+}
+json.dump(sbom, fp=os.sys.stdout, indent=2, sort_keys=True)
+print()

packaging/whisper-vulkan/smoke-runtime.sh

@@ -0,0 +1,76 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+mode=${1:?usage: smoke-runtime.sh cpu|noicd|vulkan}
+: "${OUT_DIR:=work/out}"
+: "${FIXTURE_SOURCE:=vendor/whisper.cpp}"
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+OUT_DIR="$(realpath "$OUT_DIR")"
+FIXTURE_SOURCE="$(realpath "$FIXTURE_SOURCE")"
+asset=whisper-bin-ubuntu-vulkan-x64
+case "$mode" in
+  cpu) dockerfile=Dockerfile.runtime-cpu; image=dikte-whisper-runtime-cpu:spike ;;
+  noicd) dockerfile=Dockerfile.runtime-noicd; image=dikte-whisper-runtime-noicd:spike ;;
+  vulkan) dockerfile=Dockerfile.runtime-vulkan; image=dikte-whisper-runtime-vulkan:spike ;;
+  *) echo "unknown mode: $mode" >&2; exit 2 ;;
+esac
+
+tmp=$(mktemp -d)
+trap 'rm -rf "$tmp"' EXIT
+tar -xzf "$OUT_DIR/$asset.tar.gz" -C "$tmp"
+docker build --pull=false -f "$SCRIPT_DIR/$dockerfile" -t "$image" "$SCRIPT_DIR"
+
+args=("/bundle/$asset/whisper-server" -m /fixtures/model.bin
+      --host 127.0.0.1 --port 8080
+      --inference-path /v1/audio/transcriptions -l auto -sns -nlp)
+env_args=()
+# No -ng anywhere: Dikte passes it only when its GPU setting is off, so the
+# run that has to survive a missing loader or a missing device is this one,
+# where the backend registry actually goes looking for them.
+if [[ "$mode" == vulkan ]]; then
+  env_args=(-e LIBGL_ALWAYS_SOFTWARE=1
+            -e VK_ICD_FILENAMES=/usr/share/vulkan/icd.d/lvp_icd.x86_64.json)
+fi
+
+docker run --rm --name "dikte-whisper-$mode-smoke" \
+  -e SMOKE_MODE="$mode" \
+  "${env_args[@]}" \
+  -v "$tmp/$asset:/bundle/$asset:ro" \
+  -v "$FIXTURE_SOURCE/models/for-tests-ggml-base.en.bin:/fixtures/model.bin:ro" \
+  -v "$FIXTURE_SOURCE/samples/jfk.wav:/fixtures/jfk.wav:ro" \
+  "$image" bash -ec '
+    if [ "$SMOKE_MODE" = cpu ] && ldconfig -p | grep -q libvulkan.so.1; then
+      echo "CPU smoke image unexpectedly has a Vulkan loader" >&2
+      exit 1
+    fi
+    if [ "$SMOKE_MODE" = noicd ]; then
+      if ! ldconfig -p | grep -q libvulkan.so.1; then
+        echo "no-ICD smoke image has no Vulkan loader to load" >&2
+        exit 1
+      fi
+      if compgen -G "/usr/share/vulkan/icd.d/*.json" >/dev/null; then
+        echo "no-ICD smoke image has a driver after all" >&2
+        exit 1
+      fi
+    fi
+    "$@" >/tmp/server.log 2>&1 &
+    pid=$!
+    trap "kill $pid 2>/dev/null || true" EXIT
+    for _ in $(seq 1 120); do
+      kill -0 "$pid" 2>/dev/null || { cat /tmp/server.log; exit 1; }
+      if curl --silent --show-error --fail --max-time 180 \
+          -F file=@/fixtures/jfk.wav -F response_format=json \
+          http://127.0.0.1:8080/v1/audio/transcriptions >/tmp/response.json; then
+        grep -q "\"text\"" /tmp/response.json
+        if [ "$SMOKE_MODE" = vulkan ]; then
+          grep -q "loaded Vulkan backend" /tmp/server.log
+        fi
+        cat /tmp/response.json
+        cat /tmp/server.log
+        exit 0
+      fi
+      sleep 1
+    done
+    cat /tmp/server.log
+    exit 1
+  ' bash "${args[@]}"

packaging/whisper-vulkan/validate-package.sh

@@ -0,0 +1,145 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+: "${OUT_DIR:=work/out}"
+: "${SOURCE_DIR:=whisper.cpp}"
+asset=whisper-bin-ubuntu-vulkan-x64
+archive="$OUT_DIR/$asset.tar.gz"
+tmp=$(mktemp -d)
+trap 'rm -rf "$tmp"' EXIT
+
+test -s "$archive"
+(cd "$OUT_DIR" && sha256sum --check "$asset.tar.gz.sha256")
+ARCHIVE="$archive" ASSET="$asset" python3 - <<'PY'
+import os
+import posixpath
+import tarfile
+
+archive = os.environ["ARCHIVE"]
+asset = os.environ["ASSET"]
+
+
+def under_root(name):
+    normalized = posixpath.normpath(name)
+    return (not posixpath.isabs(normalized)
+            and normalized != ".."
+            and not normalized.startswith("../")
+            and normalized.split("/", 1)[0] == asset)
+
+
+with tarfile.open(archive, "r:gz") as bundle:
+    for member in bundle:
+        if not under_root(member.name):
+            raise SystemExit(f"unsafe archive member: {member.name}")
+        if member.isdev() or member.isfifo():
+            raise SystemExit(f"special archive member: {member.name}")
+        if not (member.isdir() or member.isfile()
+                or member.issym() or member.islnk()):
+            raise SystemExit(f"unsupported archive member: {member.name}")
+        if member.issym():
+            target = posixpath.join(posixpath.dirname(member.name),
+                                    member.linkname)
+            if not under_root(target):
+                raise SystemExit(f"unsafe symlink: {member.name}")
+        if member.islnk() and not under_root(member.linkname):
+            raise SystemExit(f"unsafe hardlink: {member.name}")
+PY
+tar -xzf "$archive" -C "$tmp"
+root="$tmp/$asset"
+
+test -x "$root/whisper-server"
+test -f "$root/libwhisper.so"
+test -f "$root/libggml.so"
+test -f "$root/libggml-base.so"
+test -f "$root/libggml-vulkan.so"
+compgen -G "$root/libggml-cpu-*.so" >/dev/null
+test -f "$root/LICENSES/whisper.cpp-MIT.txt"
+test -f "$root/LICENSES/cpp-httplib-MIT.txt"
+test -f "$root/LICENSES/nlohmann-json-MIT.txt"
+(cd "$root" && sha256sum --check SHA256SUMS)
+
+# All shipped ELF objects must be relocatable and must not remember /work.
+while IFS= read -r -d '' file; do
+  file "$file" | grep -q ELF || continue
+  dynamic=$(readelf -d "$file")
+  if ! grep -Fq 'Library runpath: [$ORIGIN]' <<<"$dynamic"; then
+    echo "runpath is not \$ORIGIN in $file" >&2
+    exit 1
+  fi
+  if grep -Eq '/(home|tmp|work)/' <<<"$dynamic"; then
+    echo "build path remains in $file" >&2
+    exit 1
+  fi
+done < <(find "$root" -type f -print0)
+
+# Vulkan remains a plugin dependency. The executable must start without a loader.
+if readelf -d "$root/whisper-server" | grep -q 'libvulkan.so'; then
+  echo "whisper-server links Vulkan instead of loading it as a plugin" >&2
+  exit 1
+fi
+readelf -d "$root/libggml-vulkan.so" | grep -q 'libvulkan.so.1'
+
+# Ubuntu 22.04 establishes the glibc ceiling promised by this artifact.
+ROOT="$root" python3 - <<'PY'
+import os, pathlib, re, subprocess
+root = pathlib.Path(os.environ['ROOT'])
+seen = {'GLIBC': set(), 'GLIBCXX': set(), 'CXXABI': set()}
+external = {
+    'libc.so.6', 'libgcc_s.so.1', 'libm.so.6', 'libstdc++.so.6',
+    'libvulkan.so.1', 'ld-linux-x86-64.so.2',
+}
+for path in root.iterdir():
+    if not path.is_file() or path.is_symlink():
+        continue
+    header = subprocess.run(['readelf', '-h', path], text=True,
+                            stdout=subprocess.PIPE,
+                            stderr=subprocess.DEVNULL).stdout
+    if not header:
+        continue
+    if 'Machine:                           Advanced Micro Devices X86-64' not in header:
+        raise SystemExit(f'wrong ELF architecture: {path.name}')
+    dynamic = subprocess.run(['readelf', '-d', path], text=True,
+                             stdout=subprocess.PIPE,
+                             stderr=subprocess.DEVNULL).stdout
+    needed = re.findall(r'\(NEEDED\).*\[(.*?)\]', dynamic)
+    unexpected = [name for name in needed
+                  if name not in external
+                  and not re.fullmatch(
+                      r'lib(?:whisper|ggml(?:-base)?)\.so\.\d+', name)]
+    if unexpected:
+        raise SystemExit(
+            f'unexpected DT_NEEDED in {path.name}: {unexpected}')
+    if path.name != 'libggml-vulkan.so' and 'libvulkan.so.1' in needed:
+        raise SystemExit(f'Vulkan is not plugin-only in {path.name}')
+    contents = path.read_bytes()
+    for marker in (b'/home/', b'/tmp/', b'/work/'):
+        if marker in contents:
+            raise SystemExit(
+                f'build path {marker!r} remains in {path.name}')
+    text = subprocess.run(['objdump', '-T', path], text=True,
+                          stdout=subprocess.PIPE, stderr=subprocess.DEVNULL).stdout
+    for family in seen:
+        pattern = rf'{family}_([0-9]+(?:\.[0-9]+)+)'
+        seen[family].update(tuple(map(int, version.split('.')))
+                            for version in re.findall(pattern, text))
+assert seen['GLIBC'] and max(seen['GLIBC']) <= (2, 34), max(seen['GLIBC'])
+assert seen['GLIBCXX'] and max(seen['GLIBCXX']) <= (3, 4, 30), max(seen['GLIBCXX'])
+assert seen['CXXABI'] and max(seen['CXXABI']) <= (1, 3, 13), max(seen['CXXABI'])
+for family, versions in seen.items():
+    print(f'maximum {family} symbol:', '.'.join(map(str, max(versions))))
+PY
+
+python3 - "$root/$asset.cdx.json" <<'PY'
+import json, sys
+with open(sys.argv[1], encoding='utf-8') as stream:
+    doc = json.load(stream)
+assert doc['bomFormat'] == 'CycloneDX'
+assert doc['specVersion'] == '1.6'
+assert doc['metadata']['component']['name'] == 'whisper-server'
+assert len(doc['components']) >= 3
+print('SBOM components:', len(doc['components']))
+PY
+
+LD_LIBRARY_PATH='' "$root/whisper-server" --help >/dev/null 2>&1
+
+echo "structure: PASS"

tests/support.py

@@ -24,6 +24,7 @@ from unittest import mock
 
 from dikte import assistant
 from dikte import config as cfg
+from dikte import ggml
 from dikte import i18n
 from dikte import update
 
@@ -95,6 +96,10 @@ class DikteTest(unittest.TestCase):
         i18n.set_language("en")
         self.addCleanup(i18n.set_language, "en")
 
+        # Read once and kept for the life of the process, which across a test
+        # run means one test's machine answering for the next one's.
+        self.patch_attr(ggml, "_MEMORY", None)
+
         # cli.launch_gui replaces this process with the application when no
         # instance is running. A test that reaches it would take the whole run
         # with it and hang, so it fails loudly here instead.

tests/test_api.py

@@ -9,6 +9,7 @@ is blocked on, and a faked urlopen has no socket to cut, so those tests talk to
 a server of their own on the loopback interface.
 """
 
+import contextlib
 import http.server
 import json
 import os
@@ -53,6 +54,16 @@ class TimestampModel(unittest.TestCase):
         self.assertEqual(api.timestamp_model("openai", "gpt-4o-transcribe"),
                          "whisper-1")
 
+    def test_openrouter_takes_the_file_model_that_was_set(self):
+        self.assertEqual(
+            api.timestamp_model("openrouter", "openai/gpt-4o-transcribe",
+                                "openai/whisper-large-v3"),
+            "openai/whisper-large-v3")
+
+    def test_openrouter_with_no_file_model_falls_back_to_whisper(self):
+        self.assertEqual(api.timestamp_model("openrouter", "openai/gpt-4o-transcribe", ""),
+                         "openai/whisper-1")
+
 
 class Explain(DikteTest):
     def error(self, status):
@@ -311,13 +322,25 @@ class TranscribeSegments(DikteTest):
         fields = multipart_fields(calls[0])
         self.assertEqual(fields["model"], "whisper-1")
         self.assertEqual(fields["response_format"], "verbose_json")
-        self.assertEqual(fields["timestamp_granularities[]"], "segment")
+        # Both are asked for: whisper answers with segments, and a model that
+        # does not mark them still answers with word times.
+        body = calls[0].data.decode("utf-8", "replace")
+        for level in ("segment", "word"):
+            self.assertIn(
+                f'name="timestamp_granularities[]"\r\n\r\n{level}\r\n', body)
 
     def test_openrouter_uses_the_namespaced_id(self):
         with fake_urlopen(self.reply([{"start": 0, "end": 1, "text": "hi"}])) as calls:
             api.transcribe_segments(OPENROUTER, self.wav)
         self.assertEqual(multipart_fields(calls[0])["model"], "openai/whisper-1")
 
+    def test_openrouter_asks_for_the_file_model_when_one_is_set(self):
+        target = OPENROUTER._replace(file_model="mistralai/voxtral-mini-transcribe")
+        with fake_urlopen(self.reply([{"start": 0, "end": 1, "text": "hi"}])) as calls:
+            api.transcribe_segments(target, self.wav)
+        self.assertEqual(multipart_fields(calls[0])["model"],
+                         "mistralai/voxtral-mini-transcribe")
+
     def test_groq_stays_on_the_model_it_was_given(self):
         target = GROQ._replace(model="whisper-large-v3")
         with fake_urlopen(self.reply([{"start": 0, "end": 1, "text": "hi"}])) as calls:
@@ -345,6 +368,74 @@ class TranscribeSegments(DikteTest):
             self.assertEqual(api.transcribe_segments(OPENAI, self.wav),
                              [(5.0, 5.0, "hi")])
 
+    def test_a_long_sentence_is_broken_where_it_gets_too_long_to_read(self):
+        words = [{"word": "word", "start": i * 0.2, "end": i * 0.2 + 0.2}
+                 for i in range(60)]
+        cues = api.cues_from_words(words)
+        self.assertGreater(len(cues), 1)
+        for start, end, text in cues:
+            self.assertLessEqual(len(text), api.MAX_CUE_CHARS)
+            self.assertLessEqual(end - start, api.MAX_CUE_SECONDS + 0.2)
+
+    def test_a_pause_between_short_sentences_does_not_join_them(self):
+        cues = api.cues_from_words([
+            {"word": "Yes.", "start": 0.0, "end": 0.3},
+            {"word": "No.", "start": 9.0, "end": 9.3},
+        ])
+        self.assertEqual([(start, text) for start, _, text in cues],
+                         [(0.0, "Yes."), (9.0, "No.")])
+
+    def test_a_cue_too_short_to_read_is_held_until_the_next_one(self):
+        cues = api.cues_from_words([
+            {"word": "Yes.", "start": 0.0, "end": 0.3},
+            {"word": "No.", "start": 9.0, "end": 9.3},
+        ])
+        # The first has the room for it, the last has nothing after it to wait for.
+        self.assertEqual(cues[0][1], api.MIN_CUE_SECONDS)
+        self.assertEqual(cues[1][1], 9.0 + api.MIN_CUE_SECONDS)
+
+    def test_a_list_marker_does_not_end_a_cue_on_its_own(self):
+        cues = api.cues_from_words([
+            {"word": "1.", "start": 0.0, "end": 0.2},
+            {"word": "Antivirus.", "start": 0.4, "end": 1.6},
+        ])
+        self.assertEqual([text for _, _, text in cues], ["1. Antivirus."])
+
+    def test_a_sentence_ending_inside_a_quote_still_ends_the_cue(self):
+        cues = api.cues_from_words([
+            {"word": '"Stop', "start": 0.0, "end": 1.0},
+            {"word": 'there."', "start": 1.1, "end": 2.0},
+            {"word": "Then", "start": 2.2, "end": 2.6},
+        ])
+        self.assertEqual([text for _, _, text in cues],
+                         ['"Stop there."', "Then"])
+
+    def test_word_times_take_over_from_segments_too_long_to_read(self):
+        # What a model that does not mark segments answers with: one entry for
+        # the whole file, and the real timing in the words beside it.
+        reply = {
+            "text": "One. Two.",
+            "segments": [{"start": 0, "end": 60, "text": "One. Two."}],
+            "words": [
+                {"word": "One.", "start": 0.1, "end": 1.5},
+                {"word": "Two.", "start": 1.7, "end": 3.0},
+            ],
+        }
+        with fake_urlopen(reply):
+            self.assertEqual(api.transcribe_segments(OPENAI, self.wav),
+                             [(0.1, 1.5, "One."), (1.7, 3.0, "Two.")])
+
+    def test_whisper_segments_are_left_alone_when_words_come_too(self):
+        reply = {
+            "text": "hi there",
+            "segments": [{"start": 0, "end": 2, "text": "hi there"}],
+            "words": [{"word": "hi", "start": 0.0, "end": 0.5},
+                      {"word": "there", "start": 0.5, "end": 2.0}],
+        }
+        with fake_urlopen(reply):
+            self.assertEqual(api.transcribe_segments(OPENAI, self.wav),
+                             [(0.0, 2.0, "hi there")])
+
     def test_a_model_that_returned_no_segments_still_gives_its_text(self):
         with fake_urlopen(self.reply([], text="the whole thing")):
             self.assertEqual(api.transcribe_segments(OPENAI, self.wav),
@@ -446,6 +537,29 @@ class Cleanup(DikteTest):
         with fake_urlopen(chat_reply("   ")), self.assertRaises(api.ApiError):
             api.cleanup("hello", "k", "m", "p")
 
+    def test_a_reply_cut_off_at_a_ceiling_is_refused_rather_than_pasted(self):
+        # Half a sentence looks like a cleaned-up transcript and is not one. The
+        # caller keeps what it was given, which is the whole dictation.
+        reply = {"choices": [{"message": {"content": "Hello, and then the"},
+                              "finish_reason": "length"}]}
+        with fake_urlopen(reply), self.assertRaises(api.ApiError) as caught:
+            api.cleanup("hello", "k", "m", "p")
+        self.assertIn("cut off", str(caught.exception))
+
+    def test_a_reply_that_stopped_on_its_own_is_kept(self):
+        reply = {"choices": [{"message": {"content": "Hello."},
+                              "finish_reason": "stop"}]}
+        with fake_urlopen(reply):
+            self.assertEqual(api.cleanup("hello", "k", "m", "p"), "Hello.")
+
+    def test_all_thinking_is_named_before_the_ceiling_it_was_cut_at(self):
+        """Both are true at once, and only one of them says what to change."""
+        reply = {"choices": [{"message": {"content": "", "reasoning": "hmm"},
+                              "finish_reason": "length"}]}
+        with fake_urlopen(reply), self.assertRaises(api.ApiError) as caught:
+            api.cleanup("hello", "k", "m", "p")
+        self.assertIn("Thinking", str(caught.exception))
+
     def test_a_rate_limit_is_explained(self):
         with fake_urlopen(http_error(429)), \
                 self.assertRaises(api.ApiError) as caught:
@@ -454,6 +568,14 @@ class Cleanup(DikteTest):
 
 
 class Chat(DikteTest):
+    def test_an_answer_cut_off_at_a_ceiling_is_refused_rather_than_pasted(self):
+        # Half an answer reads like a whole one once it is on the screen.
+        reply = {"choices": [{"message": {"content": "Booked it for the"},
+                              "finish_reason": "length"}]}
+        with fake_urlopen(reply), self.assertRaises(api.ApiError) as caught:
+            api.chat([{"role": "user", "content": "book it"}], "k", "m", "p")
+        self.assertIn("cut off", str(caught.exception))
+
     def test_the_history_is_sent_after_the_system_prompt(self):
         history = [{"role": "user", "content": "book it"},
                    {"role": "assistant", "content": "done"}]
@@ -599,11 +721,14 @@ if __name__ == "__main__":
 class FakeServer:
     """A ggml.Server as far as api.py is concerned."""
 
-    def __init__(self, url="http://127.0.0.1:9999/v1", fails="", log=""):
+    def __init__(self, url="http://127.0.0.1:9999/v1", fails="", log="",
+                 context=8192):
         self.url = url
         self.fails = fails
         self.log = log
         self.starts = 0
+        self.held = 0
+        self.context = context
 
     def serve(self):
         self.starts += 1
@@ -611,9 +736,20 @@ class FakeServer:
             raise ggml.LocalError(self.fails)
         return self.url
 
+    @contextlib.contextmanager
+    def busy(self):
+        self.held += 1
+        try:
+            yield
+        finally:
+            self.held -= 1
+
     def error(self):
         return self.log
 
+    def settings(self):
+        return {"context": self.context}
+
 
 LOCAL = api.Target("local", "Local whisper", "", "", "ggml-base.bin")
 

tests/test_cleanup.py

@@ -411,6 +411,52 @@ class Here(DikteTest):
             cleanup.run("uh, done", self.conf, "the rules")
         self.assertEqual(sent_json(calls[0])["max_tokens"], 512)
 
+    def test_thinking_is_given_room_of_its_own_rather_than_the_answer_s(self):
+        # llama.cpp counts the thinking towards the same ceiling, so a rung that
+        # took its budget out of the answer would leave a short dictation with
+        # nothing to reply with. On a context roomy enough that the clamp the
+        # top rung would otherwise meet is not what is being measured.
+        self.patch_attr(ggml, "llm", FakeServer(context=32768))
+        for rung, room in api.THINKING_ROOM.items():
+            with self.subTest(rung=rung):
+                self.conf["local_llm_reasoning"] = rung
+                with fake_urlopen(chat_reply("Done.")) as calls:
+                    cleanup.run("uh, done", self.conf, "the rules")
+                self.assertEqual(sent_json(calls[0])["max_tokens"], 512 + room)
+
+    def test_each_rung_of_the_ladder_thinks_longer_than_the_one_below(self):
+        rungs = [api.THINKING_ROOM[name] for name in
+                 ("minimal", "low", "medium", "high", "xhigh", "max")]
+        self.assertEqual(rungs, sorted(rungs))
+        self.assertEqual(len(set(rungs)), len(rungs))
+
+    def test_the_models_own_default_is_given_room_to_think_in_too(self):
+        # Nothing is sent, so a template that thinks will think, and the ceiling
+        # has to survive that as well.
+        self.conf["local_llm_reasoning"] = ""
+        with fake_urlopen(chat_reply("Done.")) as calls:
+            cleanup.run("uh, done", self.conf, "the rules")
+        self.assertEqual(sent_json(calls[0])["max_tokens"],
+                         512 + api.DEFAULT_THINKING_ROOM)
+
+    def test_the_ceiling_stays_under_the_context_the_server_was_started_with(self):
+        # Above the context there is no ceiling at all: the runaway would run to
+        # the end of the context instead of stopping where this says.
+        self.patch_attr(ggml, "llm", FakeServer(context=2048))
+        self.conf["local_llm_reasoning"] = "max"
+        with fake_urlopen(chat_reply("Done.")) as calls:
+            cleanup.run("uh, done", self.conf, "the rules")
+        self.assertLess(sent_json(calls[0])["max_tokens"], 2048)
+
+    def test_the_prompt_keeps_its_share_of_a_small_context(self):
+        self.patch_attr(ggml, "llm", FakeServer(context=2048))
+        self.conf["local_llm_reasoning"] = "max"
+        with fake_urlopen(chat_reply("Done.")) as calls:
+            cleanup.run("x" * 2000, self.conf, "the rules")
+        # 2048 less half the characters of prompt and transcript together.
+        self.assertEqual(sent_json(calls[0])["max_tokens"],
+                         2048 - (len("the rules") + 2000) // 2)
+
     def test_a_reply_that_was_all_thinking_names_the_setting_that_fixes_it(self):
         reply = {"choices": [{"message": {"content": "", "reasoning": "hmm"}}]}
         with fake_urlopen(reply), self.assertRaises(api.ApiError) as caught:

tests/test_cli.py

@@ -9,6 +9,7 @@ socket is faked, and everything that runs locally runs for real.
 import contextlib
 import io
 import json
+import sys
 import unittest
 import webbrowser
 from typing import ClassVar
@@ -668,7 +669,11 @@ class WithoutAnInstance(DikteTest):
     def run_verb(self, argv):
         # launch_gui replaces this process with the application, so it never
         # comes back in real use and must not be allowed to here.
+        # `ask` with no text reads what was piped in, and the runner's own
+        # stdin is not that: under pytest it is an object that refuses to be
+        # read at all.
         with mock.patch.object(ipc, "send", return_value=None), \
+                mock.patch.object(sys, "stdin", io.StringIO()), \
                 mock.patch.object(cli, "launch_gui") as launch, \
                 captured() as (out, err):
             code = cli.run(argv)

tests/test_config.py

@@ -220,6 +220,19 @@ class TranscribeTarget(DikteTest):
         self.assertEqual(target.service, "OpenRouter")
         self.assertEqual(target.api_key, "sk-or-test")
         self.assertEqual(target.model, "openai/whisper-1")
+        self.assertEqual(target.file_model, "")
+
+    def test_openrouter_carries_its_file_model(self):
+        conf = self.config(transcribe_provider="openrouter",
+                           openrouter_api_key="sk-or-test",
+                           openrouter_file_model=" openai/whisper-large-v3 ")
+        self.assertEqual(conf.transcribe_target().file_model,
+                         "openai/whisper-large-v3")
+
+    def test_only_openrouter_has_a_file_model(self):
+        conf = self.config(transcribe_provider="openai", openai_api_key="sk-test",
+                           openrouter_file_model="openai/whisper-large-v3")
+        self.assertEqual(conf.transcribe_target().file_model, "")
 
     def test_groq_when_it_is_picked(self):
         conf = self.config(transcribe_provider="groq", groq_api_key="gsk-test",
@@ -676,3 +689,24 @@ class ReadyToRun(DikteTest):
         self.assertEqual(ggml.whisper.settings()["threads"], 4)
         self.assertFalse(ggml.whisper.settings()["gpu"])
         self.assertEqual(ggml.llm.settings()["context"], 4096)
+
+    def test_the_idle_window_is_in_seconds(self):
+        conf = self.config(local_idle_unload=True, local_idle_minutes=15)
+        self.assertEqual(conf.idle_seconds(), 900)
+
+    def test_an_unchecked_box_keeps_the_model(self):
+        conf = self.config(local_idle_unload=False, local_idle_minutes=15)
+        self.assertEqual(conf.idle_seconds(), 0)
+
+    def test_a_window_of_no_minutes_is_still_a_window(self):
+        """The spin box will not go below one; a config edited by hand can."""
+        conf = self.config(local_idle_unload=True, local_idle_minutes=0)
+        self.assertEqual(conf.idle_seconds(), 60)
+
+    def test_both_servers_are_told_the_window(self):
+        conf = self.config(local_idle_unload=True, local_idle_minutes=3)
+        self.addCleanup(ggml.llm.set_idle, 0)
+        self.addCleanup(ggml.whisper.set_idle, 0)
+        conf.apply_local()
+        self.assertEqual(ggml.whisper.idle, 180)
+        self.assertEqual(ggml.llm.idle, 180)

tests/test_ggml.py

@@ -49,6 +49,11 @@ def item(name, data, url="https://example.invalid/f", sha=True):
                     hashlib.sha256(data).hexdigest() if sha else "")
 
 
+def listed(name, size):
+    """A row as a listing hands it over: a name and a size, no bytes."""
+    return hub.Item(name, f"https://example.invalid/{name}", size, "a" * 64)
+
+
 @contextlib.contextmanager
 def serving(release, archive):
     """Answer by what is being asked for rather than by what came before.
@@ -205,6 +210,7 @@ class InstallProgram(Local):
         # These fixtures are Ubuntu release archives.  Keep checking that path
         # on every host, including the Mac that checks the macOS backend.
         self.patch_attr(sys, "platform", "linux")
+        self.patch_attr(ggml.platform, "machine", lambda: "x86_64")
         # Built once, because the release listing has to publish its checksum
         # and a tarball is not the same bytes twice.
         self.archive = tarball({
@@ -221,6 +227,7 @@ class InstallProgram(Local):
 
     def install(self, *names, archive=None):
         self.patch_attr(ggml, "_arch", lambda: "x64")
+        self.patch_attr(ggml, "_has_vulkan", lambda: False)
         blob = self.archive if archive is None else archive
         with serving(self.release(*names, archive=blob), blob) as calls:
             path = ggml.install_program(ggml.WHISPER)
@@ -238,6 +245,162 @@ class InstallProgram(Local):
                                "whisper-bin-ubuntu-x64.tar.gz")
         self.assertTrue(urls[1].endswith("whisper-bin-ubuntu-x64.tar.gz"))
 
+    def test_the_nightly_pointer_is_followed_to_where_the_builds_are(self):
+        """llama.cpp's latest release carries a tag name, not the binaries."""
+        self.patch_attr(ggml, "_arch", lambda: "x64")
+        self.patch_attr(ggml, "_has_vulkan", lambda: False)
+        marker = self.release(ggml.NIGHTLY_TAG)
+        nightly = dict(self.release("llama-b10809-bin-ubuntu-x64.tar.gz"),
+                       tag_name="b10809")
+
+        def opener(request, timeout=None):
+            url = request.full_url
+            if url.endswith("/releases/latest"):
+                return json_body(marker)
+            if url.endswith("/releases/tags/b10809"):
+                return json_body(nightly)
+            if url.endswith(ggml.NIGHTLY_TAG):
+                return body(b"b10809\n")
+            return body(self.archive)
+
+        with mock.patch("urllib.request.urlopen", side_effect=opener):
+            tag, found = ggml._pick_asset(ggml.LLAMA)
+        self.assertEqual(tag, "b10809")
+        self.assertEqual(found.name, "llama-b10809-bin-ubuntu-x64.tar.gz")
+
+    def test_without_a_pointer_the_newest_release_that_has_a_build_is_taken(self):
+        self.patch_attr(ggml, "_arch", lambda: "x64")
+        self.patch_attr(ggml, "_has_vulkan", lambda: False)
+        marker = self.release("source.zip")
+        listing = [dict(self.release("llama-b2-bin-win-cpu-x64.zip"), tag_name="b2"),
+                   dict(self.release("llama-b1-bin-ubuntu-x64.tar.gz"), tag_name="b1")]
+
+        def opener(request, timeout=None):
+            url = request.full_url
+            return json_body(listing if "per_page" in url else marker)
+
+        with mock.patch("urllib.request.urlopen", side_effect=opener):
+            tag, found = ggml._pick_asset(ggml.LLAMA)
+        self.assertEqual(tag, "b1")
+        self.assertEqual(found.name, "llama-b1-bin-ubuntu-x64.tar.gz")
+    def test_linux_x64_with_vulkan_takes_diktes_accelerated_build(self):
+        self.patch_attr(ggml, "_arch", lambda: "x64")
+        self.patch_attr(ggml, "_has_vulkan", lambda: True)
+        listing = self.release("whisper-bin-ubuntu-vulkan-x64.tar.gz")
+        listing["tag_name"] = "whisper.cpp-v1.9.3"
+        managed_sha = hashlib.sha256(self.archive).hexdigest()
+        with mock.patch.object(ggml, "MANAGED_WHISPER_SHA256", managed_sha,
+                               create=True):
+            with fake_urlopen(listing, body(self.archive)) as calls:
+                path = ggml.install_program(ggml.WHISPER)
+        urls = [call.full_url for call in calls]
+        self.assertIn(
+            "/repos/yusufipk/dikte/releases/tags/whisper.cpp-v1.9.3",
+            urls[0],
+        )
+        self.assertTrue(urls[1].endswith(
+            "whisper-bin-ubuntu-vulkan-x64.tar.gz"))
+        self.assertTrue(os.path.isfile(path))
+        self.assertEqual("v1.9.3", ggml.installed_version(ggml.WHISPER))
+        self.assertFalse(ggml.vulkan_missing(ggml.WHISPER))
+
+    def test_an_explicit_whisper_version_still_comes_from_upstream(self):
+        self.patch_attr(ggml, "_arch", lambda: "x64")
+        self.patch_attr(ggml, "_has_vulkan", lambda: True)
+        listing = self.release("whisper-bin-ubuntu-x64.tar.gz")
+        with fake_urlopen(listing, body(self.archive)) as calls:
+            ggml.install_program(ggml.WHISPER, tag="v1.9.1")
+        self.assertIn(
+            "/repos/ggml-org/whisper.cpp/releases/tags/v1.9.1",
+            calls[0].full_url,
+        )
+
+    def test_linux_arm64_keeps_using_the_upstream_cpu_build(self):
+        self.patch_attr(ggml, "_arch", lambda: "arm64")
+        self.patch_attr(ggml.platform, "machine", lambda: "aarch64")
+        self.patch_attr(ggml, "_has_vulkan", lambda: True)
+        listing = self.release("whisper-bin-ubuntu-arm64.tar.gz")
+        with fake_urlopen(listing, body(self.archive)) as calls:
+            ggml.install_program(ggml.WHISPER)
+        self.assertIn(
+            "/repos/ggml-org/whisper.cpp/releases/latest",
+            calls[0].full_url,
+        )
+
+    def test_linux_non_x86_does_not_try_the_managed_x64_build(self):
+        self.patch_attr(ggml, "_has_vulkan", lambda: True)
+        listing = self.release("whisper-bin-ubuntu-arm64.tar.gz")
+        with mock.patch("platform.machine", return_value="ppc64le"):
+            with fake_urlopen(listing, listing) as calls:
+                with self.assertRaises(ggml.LocalError):
+                    ggml.install_program(ggml.WHISPER)
+        self.assertIn(
+            "/repos/ggml-org/whisper.cpp/releases/latest",
+            calls[0].full_url,
+        )
+
+    def test_a_missing_managed_build_falls_back_to_upstream_cpu(self):
+        self.patch_attr(ggml, "_arch", lambda: "x64")
+        self.patch_attr(ggml, "_has_vulkan", lambda: True)
+        managed = self.release("Dikte-1.1.0-x86_64.AppImage")
+        managed["tag_name"] = "whisper.cpp-v1.9.3"
+        upstream = self.release("whisper-bin-ubuntu-x64.tar.gz")
+        with fake_urlopen(managed, upstream, body(self.archive)) as calls:
+            path = ggml.install_program(ggml.WHISPER)
+        urls = [call.full_url for call in calls]
+        self.assertIn(
+            "/repos/yusufipk/dikte/releases/tags/whisper.cpp-v1.9.3",
+            urls[0],
+        )
+        self.assertIn("/repos/ggml-org/whisper.cpp/releases/latest", urls[1])
+        self.assertTrue(urls[2].endswith("whisper-bin-ubuntu-x64.tar.gz"))
+        self.assertTrue(os.path.isfile(path))
+
+    def test_a_managed_build_with_an_unreviewed_digest_falls_back(self):
+        self.patch_attr(ggml, "_has_vulkan", lambda: True)
+        managed = self.release("whisper-bin-ubuntu-vulkan-x64.tar.gz")
+        managed["assets"][0]["digest"] = "sha256:" + "0" * 64
+        upstream = self.release("whisper-bin-ubuntu-x64.tar.gz")
+        with fake_urlopen(managed, upstream, body(self.archive)) as calls:
+            try:
+                path = ggml.install_program(ggml.WHISPER)
+            except ggml.LocalError as exc:
+                self.fail(f"unreviewed digest did not fall back: {exc}")
+        urls = [call.full_url for call in calls]
+        self.assertEqual(3, len(urls))
+        self.assertTrue(urls[2].endswith("whisper-bin-ubuntu-x64.tar.gz"))
+        self.assertTrue(os.path.isfile(path))
+
+    def test_an_unavailable_managed_release_falls_back_to_upstream_cpu(self):
+        self.patch_attr(ggml, "_arch", lambda: "x64")
+        self.patch_attr(ggml, "_has_vulkan", lambda: True)
+        upstream = self.release("whisper-bin-ubuntu-x64.tar.gz")
+        with fake_urlopen(http_error(404), upstream,
+                          body(self.archive)) as calls:
+            path = ggml.install_program(ggml.WHISPER)
+        self.assertEqual(3, len(calls))
+        self.assertTrue(calls[2].full_url.endswith(
+            "whisper-bin-ubuntu-x64.tar.gz"))
+        self.assertTrue(os.path.isfile(path))
+
+    def test_a_fallback_to_the_processor_build_is_there_to_be_shown(self):
+        """Until the Vulkan package is published every download lands the
+        processor build, and a graphics card sitting idle looks exactly like
+        one being used. The window asks this and says so."""
+        self.patch_attr(ggml, "_arch", lambda: "x64")
+        self.patch_attr(ggml, "_has_vulkan", lambda: True)
+        managed = self.release("Dikte-1.1.0-x86_64.AppImage")
+        managed["tag_name"] = "whisper.cpp-v1.9.3"
+        upstream = self.release("whisper-bin-ubuntu-x64.tar.gz")
+        with fake_urlopen(managed, upstream, body(self.archive)):
+            ggml.install_program(ggml.WHISPER)
+        self.assertTrue(ggml.vulkan_missing(ggml.WHISPER))
+
+    def test_a_machine_with_no_vulkan_is_not_told_it_is_missing_one(self):
+        # Nothing was on offer to fall back from, so there is nothing to say.
+        self.install("whisper-bin-ubuntu-x64.tar.gz")
+        self.assertFalse(ggml.vulkan_missing(ggml.WHISPER))
+
     def test_a_release_with_nothing_for_this_machine_says_so(self):
         self.patch_attr(ggml, "_arch", lambda: "x64")
         with fake_urlopen(self.release("whisper-bin-Win32.zip")):
@@ -506,6 +669,53 @@ class Catalogue(Local):
             with self.assertRaises(ggml.LocalError):
                 ggml.whisper_models()
 
+    def test_the_speculative_decoding_heads_are_not_models(self):
+        # They are the small files in a repository, so a list sorted by size
+        # puts them first, where the eye lands and the click goes.
+        tree = GGUF_TREE + [
+            {"type": "file", "path": "dflash-Qwen3-8B-Q8_0.gguf",
+             "size": 1_120_000_000, "lfs": {"oid": "f" * 64}},
+            {"type": "file", "path": "eagle3-gpt-oss-20b-Q8_0.gguf",
+             "size": 920_000_000, "lfs": {"oid": "0" * 64}},
+        ]
+        with fake_urlopen(tree):
+            names = [q.name for q in ggml.llm_quants("ggml-org/x-GGUF")]
+        self.assertEqual(names,
+                         ["gemma-3-4b-it-Q4_K_M.gguf", "gemma-3-4b-it-Q8_0.gguf"])
+
+    def test_a_speech_or_vision_repository_is_not_a_cleanup_publisher(self):
+        listing = [{"id": "ggml-org/parakeet-GGUF"},
+                   {"id": "ggml-org/Qwen3-TTS-12Hz-1.7B-Base-GGUF"},
+                   {"id": "ggml-org/SmolVLM2-256M-Video-Instruct-GGUF"},
+                   {"id": "ggml-org/Qwen3-8B-Base-GGUF"},
+                   {"id": "ggml-org/SmolLM3-3B-GGUF"}]
+        with fake_urlopen(listing):
+            found = ggml.llm_repos()
+        self.assertEqual([r for r in found if r.startswith("ggml-org/Smol")],
+                         ["ggml-org/SmolLM3-3B-GGUF"])
+        self.assertNotIn("ggml-org/parakeet-GGUF", found)
+        self.assertNotIn("ggml-org/Qwen3-8B-Base-GGUF", found)
+
+    def test_a_publisher_is_not_dropped_for_a_word_it_happens_to_contain(self):
+        # The skip marks are matched as plain substrings, and an unanchored
+        # "test-" is also inside "Latest-".
+        self.assertTrue(ggml.can_clean("ggml-org/Qwen3-Latest-GGUF"))
+        self.assertFalse(ggml.can_clean("ggml-org/test-model-router-download"))
+
+    def test_a_base_model_beside_its_tuned_twin_is_dropped(self):
+        # Gemma names the base model after the tuned one with the `-it` taken
+        # out, so the two sit next to each other and the wrong one answers a
+        # cleanup prompt by carrying on writing the transcript.
+        listing = [{"id": "ggml-org/gemma-4-E2B-GGUF"},
+                   {"id": "ggml-org/gemma-4-E2B-it-GGUF"},
+                   {"id": "ggml-org/Qwen3-0.6B-GGUF"}]
+        with fake_urlopen(listing):
+            found = ggml.llm_repos()
+        self.assertNotIn("ggml-org/gemma-4-E2B-GGUF", found)
+        self.assertIn("ggml-org/gemma-4-E2B-it-GGUF", found)
+        # Nothing named it, so nothing says it is the wrong half of a pair.
+        self.assertIn("ggml-org/Qwen3-0.6B-GGUF", found)
+
     def test_what_is_on_disk_is_read_from_disk(self):
         self.assertEqual(ggml.installed_whisper_models(), [])
         path = ggml.whisper_model_path("ggml-base.bin")
@@ -573,7 +783,9 @@ STAND_IN = textwrap.dedent("""
 """)
 
 
-class Servers(Local):
+class ServerCase(Local):
+    """The stand-in server and the fixture around it, with no tests of its own."""
+
     def setUp(self):
         super().setUp()
         self.path("data").mkdir(parents=True, exist_ok=True)
@@ -596,6 +808,8 @@ class Servers(Local):
         self.addCleanup(made.stop)
         return made
 
+
+class Servers(ServerCase):
     def test_a_started_server_hands_back_its_address(self):
         server = self.server()
         url = server.serve()
@@ -827,6 +1041,122 @@ class Servers(Local):
         self.assertFalse(server.sweep())      # and the pid file went with it
 
 
+class IdleUnload(ServerCase):
+    """Giving the memory back when nothing has asked anything for a while."""
+
+    IDLE = 0.3
+
+    def setUp(self):
+        super().setUp()
+        # The real check runs every five seconds against a window of minutes.
+        # Both are scaled down here; what is being tested is the decision, and
+        # nothing in it reads the clock in units of its own.
+        self.patch_attr(ggml, "IDLE_CHECK_SECONDS", 0.05)
+
+    def idle_server(self, seconds=None, **settings):
+        server = self.server(**settings)
+        server.set_idle(self.IDLE if seconds is None else seconds)
+        return server
+
+    def wait_for(self, predicate, timeout=5.0):
+        """True as soon as `predicate` holds, False once the wait runs out."""
+        deadline = time.monotonic() + timeout
+        while time.monotonic() < deadline:
+            if predicate():
+                return True
+            time.sleep(0.02)
+        return False
+
+    def test_a_model_nobody_is_using_is_unloaded(self):
+        server = self.idle_server()
+        server.serve()
+        self.assertTrue(self.wait_for(lambda: not server.running))
+
+    def test_the_default_is_to_keep_it(self):
+        """A server nobody set a window on stays until something stops it."""
+        server = self.server()
+        server.serve()
+        self.assertFalse(self.wait_for(lambda: not server.running, timeout=0.6))
+
+    def test_a_window_of_zero_keeps_it_too(self):
+        server = self.idle_server(0)
+        server.serve()
+        self.assertFalse(self.wait_for(lambda: not server.running, timeout=0.6))
+
+    def test_a_request_in_flight_holds_the_model(self):
+        """A file is one address lookup and then minutes of work: the clock
+        alone would call that idle and unload it mid-transcription."""
+        server = self.idle_server()
+        server.serve()
+        with server.busy():
+            self.assertFalse(
+                self.wait_for(lambda: not server.running, timeout=self.IDLE * 3))
+        self.assertTrue(self.wait_for(lambda: not server.running))
+
+    def test_asking_for_the_address_puts_the_window_back(self):
+        server = self.idle_server()
+        first = server.serve()
+        for _ in range(4):
+            time.sleep(self.IDLE / 2)
+            self.assertEqual(server.serve(), first)   # never restarted
+        self.assertTrue(server.running)
+
+    def test_the_next_request_loads_it_again(self):
+        server = self.idle_server()
+        first = server.serve()
+        self.assertTrue(self.wait_for(lambda: not server.running))
+        second = server.serve()
+        self.assertTrue(server.running)
+        self.assertNotEqual(second, first)      # a new process, a new port
+
+    def test_the_watcher_of_a_stopped_server_does_not_touch_the_next_one(self):
+        server = self.idle_server()
+        server.serve()
+        server.stop()
+        server.set_idle(0)
+        server.serve()
+        self.assertFalse(self.wait_for(lambda: not server.running, timeout=0.6))
+
+    def test_unloading_by_hand_does_not_wait_for_the_window(self):
+        server = self.idle_server(0)
+        server.serve()
+        self.assertTrue(server.unload())
+        self.assertFalse(server.running)
+
+    def test_a_hold_taken_before_the_start_survives_it(self):
+        """The local cleanup takes the hold and only then asks for the address,
+        so the start it triggers must not be what drops the hold."""
+        server = self.idle_server()
+        with server.busy():
+            server.serve()
+            self.assertFalse(
+                self.wait_for(lambda: not server.running, timeout=self.IDLE * 3))
+        self.assertTrue(self.wait_for(lambda: not server.running))
+
+    def test_unloading_is_refused_while_the_model_is_still_loading(self):
+        """It runs on the interface's thread, and a start holds its lock for as
+        long as the load takes: waiting there would freeze the whole window."""
+        server = self.idle_server(0, extra=["--wait", "0.6"])
+        thread = threading.Thread(target=server.serve)
+        thread.start()
+        try:
+            began = time.monotonic()
+            self.assertFalse(server.unload())
+            self.assertLess(time.monotonic() - began, 0.2)
+        finally:
+            thread.join(timeout=10)
+
+    def test_unloading_is_refused_while_a_request_is_in_flight(self):
+        server = self.idle_server(0)
+        server.serve()
+        with server.busy():
+            self.assertFalse(server.unload())
+            self.assertTrue(server.running)
+
+    def test_unloading_nothing_is_not_a_refusal(self):
+        self.assertTrue(self.server().unload())
+
+
 class Arguments(Local):
     """What the two command lines say, since neither program is here to say it."""
 
@@ -1025,3 +1355,205 @@ class WindowsOwnership(Local):
         # from here", and only one of those makes the pid file safe to drop.
         self.image("")
         self.assertIsNone(self.made._is_ours(1234))
+
+
+class Machine(Local):
+    """What this machine can hold, and what that makes worth pointing at."""
+
+    def _sysconf(self, phys_pages, page_size=4096):
+        """Stand where sysconf answers whatever this test wants it to.
+
+        `create` because Windows has no os.sysconf at all, and a patch that
+        insists on the real attribute fails there before the test runs. What
+        the code under test does about that absence is two lines down from
+        what these are checking, and it is checked on its own below.
+        """
+        return mock.patch.object(
+            ggml.os, "sysconf", create=True,
+            side_effect=lambda name: (page_size if name == "SC_PAGE_SIZE"
+                                      else phys_pages))
+
+    def test_the_memory_is_read_the_way_each_system_reports_it(self):
+        # Linux and most Macs answer through sysconf.
+        with self._sysconf(4_194_304):
+            self.assertEqual(ggml.total_memory(), 16 * ggml.GB)
+
+    def test_a_mac_without_the_page_count_is_asked_for_the_number(self):
+        # Not every build of Python on a Mac carries SC_PHYS_PAGES, and a Mac
+        # that answered nothing would be a Mac with none of this on it.
+        def answer(args, **kwargs):
+            self.assertEqual(args, ["sysctl", "-n", "hw.memsize"])
+            return mock.Mock(stdout=f"{32 * ggml.GB}\n")
+
+        with mock.patch.object(ggml.os, "sysconf", create=True,
+                               side_effect=ValueError), \
+                mock.patch.object(sys, "platform", "darwin"), \
+                mock.patch.object(ggml.subprocess, "run", answer):
+            self.assertEqual(ggml.total_memory(), 32 * ggml.GB)
+
+    def test_a_sysconf_that_shrugs_is_an_unknown_machine_and_not_a_tiny_one(self):
+        # sysconf answers -1 for a limit it holds to be indeterminate and
+        # CPython hands that back rather than raising, so the product came out
+        # negative: a 64 GB workstation was told every model past 512 MB was
+        # too big for it, and the machine line read "Memory: -4096 B".
+        with self._sysconf(-1):
+            self.assertEqual(ggml.total_memory(), 0)
+        self.assertTrue(ggml.fits(574 << 20, memory=0))
+
+    def test_the_memory_is_read_once_and_kept(self):
+        # A list of thirty rows asks seventy times, and on the Mac path the
+        # answer comes from a program rather than a library call.
+        calls = []
+        with mock.patch.object(ggml, "_read_memory",
+                               lambda: calls.append(1) or 16 * ggml.GB):
+            self.assertEqual(ggml.total_memory(), 16 * ggml.GB)
+            self.assertEqual(ggml.total_memory(), 16 * ggml.GB)
+        self.assertEqual(len(calls), 1)
+
+    def test_a_system_that_answers_nothing_is_an_unknown_machine(self):
+        with mock.patch.object(ggml.os, "sysconf", create=True,
+                               side_effect=ValueError), \
+                mock.patch.object(sys, "platform", "linux"):
+            self.assertEqual(ggml.total_memory(), 0)
+
+    def test_a_mac_is_taken_to_have_a_graphics_interface(self):
+        with mock.patch.object(sys, "platform", "darwin"):
+            self.assertEqual(ggml.accelerator(), "Metal")
+
+    def test_elsewhere_the_vulkan_loader_is_what_says_so(self):
+        with mock.patch.object(sys, "platform", "linux"), \
+                mock.patch.object(ggml.ctypes.util, "find_library",
+                                  lambda name: "/usr/lib/libvulkan.so.1"):
+            self.assertEqual(ggml.accelerator(), "Vulkan")
+        with mock.patch.object(sys, "platform", "linux"), \
+                mock.patch.object(ggml.ctypes.util, "find_library",
+                                  lambda name: None):
+            self.assertEqual(ggml.accelerator(), "")
+
+    def test_a_model_is_measured_against_half_the_memory(self):
+        self.assertTrue(ggml.fits(2 * ggml.GB, memory=8 * ggml.GB))
+        self.assertFalse(ggml.fits(4 * ggml.GB, memory=8 * ggml.GB))
+
+    def test_a_machine_whose_memory_could_not_be_read_holds_anything(self):
+        # A wrong "too big" is worse advice than none.
+        self.assertTrue(ggml.fits(40 * ggml.GB, memory=0))
+
+    def test_the_smallest_machine_is_not_the_one_where_everything_fits(self):
+        # Half of 2 GB less the gigabyte of overhead is nothing, and a budget
+        # of nothing used to read as the unknown machine above.
+        self.assertFalse(ggml.fits(3 * ggml.GB, memory=2 * ggml.GB))
+
+    def test_a_crowded_machine_is_pointed_at_the_smaller_model(self):
+        self.assertEqual(ggml.suggested_whisper(memory=3 * ggml.GB, graphics=""),
+                         ggml.SMALL_MACHINE_WHISPER)
+
+    def test_a_card_and_the_memory_for_it_are_pointed_at_the_accurate_one(self):
+        self.assertEqual(
+            ggml.suggested_whisper(memory=32 * ggml.GB, graphics="Vulkan"),
+            ggml.ACCURATE_WHISPER)
+
+    def test_memory_without_a_card_is_pointed_at_the_fast_one(self):
+        # Several times the work per second is several times a long wait on a
+        # processor, whatever there is room for.
+        self.assertEqual(
+            ggml.suggested_whisper(memory=32 * ggml.GB, graphics=""),
+            ggml.SUGGESTED_WHISPER)
+
+    def test_a_sixteen_gigabyte_machine_counts_as_a_roomy_one(self):
+        # What a machine reports is what the firmware and the graphics left
+        # of it: 16 GB answers about 15.4, and a threshold written at the
+        # number on the box is one no machine ever reaches.
+        self.assertEqual(
+            ggml.suggested_whisper(memory=int(15.4 * ggml.GB), graphics="Metal"),
+            ggml.ACCURATE_WHISPER)
+
+    def test_the_suggestion_that_fits_is_offered_first(self):
+        first = ggml.suggested_llm(memory=6 * ggml.GB)[0]
+        self.assertTrue(ggml.fits(ggml.SUGGESTED_LLM_SIZE[first],
+                                  memory=6 * ggml.GB))
+        # Nothing is dropped: what does not fit today fits once something else
+        # is closed.
+        self.assertEqual(sorted(ggml.suggested_llm(memory=6 * ggml.GB)),
+                         sorted(ggml.SUGGESTED_LLM))
+
+    def test_the_wanted_model_wins_when_there_is_room_for_it(self):
+        items = [listed("ggml-tiny.bin", 70 << 20),
+                 listed("ggml-large-v3-turbo-q5_0.bin", 574 << 20)]
+        self.assertEqual(
+            ggml.recommended(items, "ggml-large-v3-turbo-q5_0.bin",
+                             memory=16 * ggml.GB),
+            "ggml-large-v3-turbo-q5_0.bin")
+
+    def test_a_model_too_big_for_the_machine_is_not_recommended(self):
+        items = [listed("small.gguf", 1 << 30), listed("huge.gguf", 12 * ggml.GB)]
+        self.assertEqual(ggml.recommended(items, "huge.gguf",
+                                          memory=8 * ggml.GB), "small.gguf")
+
+    def test_the_full_precision_weights_are_never_the_recommendation(self):
+        # Twice the memory and twice the wait for a difference this job
+        # cannot see.
+        items = [listed("model-Q4_0.gguf", 2 * ggml.GB),
+                 listed("model-BF16.gguf", 3 * ggml.GB)]
+        self.assertEqual(ggml.recommended(items, memory=32 * ggml.GB),
+                         "model-Q4_0.gguf")
+
+    def test_nothing_is_recommended_when_nothing_fits(self):
+        self.assertEqual(
+            ggml.recommended([listed("huge.gguf", 40 * ggml.GB)],
+                             memory=8 * ggml.GB), "")
+
+
+class Grouping(Local):
+    """One group per model, rather than one long list sorted by size."""
+
+    def test_every_spelling_of_a_quantisation_reads_as_its_number(self):
+        # One list holds q5_1, Q4_K_M, MXFP4 and BF16, and the number is the
+        # whole of what any of them says to somebody choosing a row.
+        self.assertEqual(ggml.bit_depth("ggml-small-q5_1.bin"), 5)
+        self.assertEqual(ggml.bit_depth("SmolLM3-Q4_K_M.gguf"), 4)
+        self.assertEqual(ggml.bit_depth("gpt-oss-20b-MXFP4.gguf"), 4)
+        self.assertEqual(ggml.bit_depth("gemma-4-E2B-it-Q8_0.gguf"), 8)
+        # bf16 is not f16 read badly.
+        self.assertEqual(ggml.bit_depth("gemma-4-E2B-it-BF16.gguf"), 16)
+        self.assertEqual(ggml.bit_depth("mmproj-model-f16.gguf"), 16)
+        # A whisper file with no mark is the full model, and its name is the
+        # one convention here that does not carry the answer.
+        self.assertEqual(ggml.bit_depth("ggml-large-v3-turbo.bin"), 0)
+
+    def test_a_quantisation_belongs_to_the_model_it_is_a_copy_of(self):
+        self.assertEqual(ggml.whisper_family("ggml-small.en-q5_1.bin"), "small")
+        self.assertEqual(ggml.whisper_family("ggml-large-v3-q5_0.bin"),
+                         "large-v3")
+        self.assertEqual(ggml.whisper_family("ggml-large-v3-turbo.bin"),
+                         "large-v3-turbo")
+        self.assertEqual(ggml.whisper_family("ggml-medium.en.bin"), "medium")
+
+    def test_turbo_is_a_model_and_not_a_quantisation(self):
+        # The last chunk of the name is a quantisation for most of the list
+        # and part of the model's name here.
+        self.assertEqual(ggml.whisper_family("ggml-large-v3-turbo-q8_0.bin"),
+                         "large-v3-turbo")
+
+    def test_the_turbo_files_are_not_scattered_through_the_medium_ones(self):
+        # Sorted by size alone, large-v3-turbo-q5_0 lands between the two
+        # medium quantisations, half a screen from the model it is a copy of.
+        models = [listed("ggml-medium-q5_0.bin", 539 << 20),
+                  listed("ggml-large-v3-turbo-q5_0.bin", 574 << 20),
+                  listed("ggml-medium-q8_0.bin", 823 << 20),
+                  listed("ggml-large-v3-turbo.bin", 1624 << 20)]
+        groups = dict(ggml.whisper_groups(models))
+        self.assertEqual([i.name for i in groups["large-v3-turbo"]],
+                         ["ggml-large-v3-turbo-q5_0.bin",
+                          "ggml-large-v3-turbo.bin"])
+        self.assertEqual([i.name for i in groups["medium"]],
+                         ["ggml-medium-q5_0.bin", "ggml-medium-q8_0.bin"])
+
+    def test_the_smallest_model_comes_first_and_the_english_ones_last(self):
+        models = [listed("ggml-small.en-q5_1.bin", 190 << 20),
+                  listed("ggml-small-q5_1.bin", 190 << 20),
+                  listed("ggml-tiny.bin", 77 << 20)]
+        groups = ggml.whisper_groups(models)
+        self.assertEqual([family for family, _ in groups], ["tiny", "small"])
+        self.assertEqual([i.name for _, group in groups for i in group],
+                         ["ggml-tiny.bin", "ggml-small-q5_1.bin",
+                          "ggml-small.en-q5_1.bin"])

tests/test_packaging.py

@@ -0,0 +1,217 @@
+"""The release build that makes Linux Vulkan a one-click install."""
+
+import hashlib
+import io
+import json
+import os
+import pathlib
+import shutil
+import subprocess
+import sys
+import tarfile
+import tempfile
+import unittest
+
+from dikte import ggml
+
+
+ROOT = pathlib.Path(__file__).parents[1]
+PACKAGING = ROOT / "packaging" / "whisper-vulkan"
+WORKFLOW = ROOT / ".github" / "workflows" / "whisper-vulkan.yml"
+
+
+class WhisperVulkanPackaging(unittest.TestCase):
+    @unittest.skipUnless(sys.platform != "win32" and shutil.which("bash"),
+                         "bash syntax check is unavailable")
+    def test_the_release_scripts_parse_as_shell(self):
+        for name in ("build-package.sh", "validate-package.sh",
+                     "smoke-runtime.sh"):
+            script = PACKAGING / name
+            checked = subprocess.run(
+                ["bash", "-n", script], capture_output=True, text=True,
+            )
+            self.assertEqual("", checked.stderr)
+            self.assertEqual(0, checked.returncode)
+
+    def test_the_workflow_builds_validates_smokes_and_publishes(self):
+        workflow = WORKFLOW.read_text(encoding="utf-8")
+        for step in ("Build deterministic archive",
+                     "Verify reviewed archive digest",
+                     "Validate archive and ELF contract",
+                     "CPU fallback smoke test (no Vulkan loader)",
+                     "Vulkan loader present, no device smoke test",
+                     "Vulkan plugin-load smoke test (Mesa llvmpipe)",
+                     "Publish dependency release"):
+            self.assertIn(step, workflow)
+        self.assertNotRegex(workflow, r"uses: [^\n]+@v\d+(?:\s|$)")
+
+    def test_publish_is_safe_for_dikte_and_limited_to_reviewed_master(self):
+        workflow = WORKFLOW.read_text(encoding="utf-8")
+        self.assertGreaterEqual(workflow.count("persist-credentials: false"), 2)
+        self.assertIn("github.ref == 'refs/heads/master'", workflow)
+        self.assertIn("--prerelease", workflow)
+        self.assertIn("--latest=false", workflow)
+        self.assertIn("--verify-tag", workflow)
+        self.assertIn("refusing to replace existing tag", workflow)
+        self.assertIn("^[0-9]+\\.[0-9]+\\.[0-9]+$", workflow)
+        self.assertIn("^[0-9a-f]{40}$", workflow)
+        publish_script = workflow.split("      - name: Publish dependency release", 1)[1]
+        publish_script = publish_script.split("        run: |", 1)[1]
+        self.assertNotIn("${{ inputs.", publish_script)
+
+    def test_bundle_ci_runs_only_for_what_the_bundle_is_built_from(self):
+        """A 45 minute build on a README typo is a tax on every other change.
+
+        What ties ggml.py to the release is checked in this file instead, and
+        this file runs on every pull request in milliseconds."""
+        workflow = WORKFLOW.read_text(encoding="utf-8")
+        trigger = workflow.split("workflow_dispatch:", 1)[0]
+        self.assertIn("- packaging/whisper-vulkan/**", trigger)
+        self.assertIn("- .github/workflows/whisper-vulkan.yml", trigger)
+        for path in ("dikte/ggml.py", "tests/test_ggml.py",
+                     "tests/test_packaging.py", "README.md", "README.tr.md"):
+            self.assertNotIn(f"- {path}", trigger)
+
+    def test_the_smoke_tests_run_what_dikte_runs(self):
+        """-ng is what Dikte passes when its GPU setting is off, and a run
+        with it never asks for a backend at all. The three runs that have to
+        hold are the ones without it: no loader, a loader with nothing behind
+        it, and a working device."""
+        script = (PACKAGING / "smoke-runtime.sh").read_text(encoding="utf-8")
+        code = "\n".join(line for line in script.splitlines()
+                         if not line.lstrip().startswith("#"))
+        self.assertNotIn("-ng", code)
+        for mode in ("cpu)", "noicd)", "vulkan)"):
+            self.assertIn(mode, script)
+        self.assertTrue((PACKAGING / "Dockerfile.runtime-noicd").is_file())
+
+    def test_an_unreviewed_version_is_reported_and_never_published(self):
+        """The digest of a version nobody has reviewed cannot be known before
+        it is built, so the gate cannot be the only way through."""
+        workflow = WORKFLOW.read_text(encoding="utf-8")
+        self.assertIn("expected_sha256", workflow)
+        self.assertIn(
+            "refusing to publish an archive whose digest has not been reviewed",
+            workflow)
+
+    def test_the_shape_of_the_inputs_is_checked_before_they_are_used(self):
+        workflow = WORKFLOW.read_text(encoding="utf-8")
+        self.assertLess(workflow.index("- name: Validate source coordinates"),
+                        workflow.index("- name: Check out pinned whisper.cpp"))
+
+    def test_the_validator_checks_tar_links_before_extraction(self):
+        validator = (PACKAGING / "validate-package.sh").read_text(
+            encoding="utf-8")
+        for check in ("member.issym()", "member.islnk()", "member.isdev()"):
+            self.assertIn(check, validator)
+
+    @unittest.skipUnless(sys.platform == "linux" and shutil.which("bash"),
+                         "Linux packaging test is unavailable")
+    def test_the_validator_rejects_an_escaping_symlink(self):
+        asset = "whisper-bin-ubuntu-vulkan-x64"
+        with tempfile.TemporaryDirectory() as temporary:
+            output = pathlib.Path(temporary)
+            archive = output / f"{asset}.tar.gz"
+            with tarfile.open(archive, "w:gz") as bundle:
+                link = tarfile.TarInfo(f"{asset}/whisper-server")
+                link.type = tarfile.SYMTYPE
+                link.linkname = "/etc/passwd"
+                bundle.addfile(link, io.BytesIO())
+            digest = hashlib.sha256(archive.read_bytes()).hexdigest()
+            (output / f"{asset}.tar.gz.sha256").write_text(
+                f"{digest}  {asset}.tar.gz\n", encoding="utf-8",
+            )
+            checked = subprocess.run(
+                ["bash", PACKAGING / "validate-package.sh"],
+                env=os.environ | {"OUT_DIR": str(output)},
+                capture_output=True, text=True,
+            )
+        self.assertNotEqual(0, checked.returncode)
+        self.assertIn("unsafe symlink", checked.stderr)
+
+    def test_the_validator_checks_elf_architecture_dependencies_and_paths(self):
+        validator = (PACKAGING / "validate-package.sh").read_text(
+            encoding="utf-8")
+        for check in ("Advanced Micro Devices X86-64", "unexpected DT_NEEDED",
+                      "path.read_bytes()"):
+            self.assertIn(check, validator)
+
+    def test_the_builder_and_its_downloads_are_pinned(self):
+        dockerfile = (PACKAGING / "Dockerfile.build").read_text(
+            encoding="utf-8")
+        self.assertRegex(dockerfile, r"FROM ubuntu@sha256:[0-9a-f]{64}")
+        self.assertIn("CMAKE_SHA256=", dockerfile)
+        self.assertIn("libvulkan-dev=", dockerfile)
+        self.assertIn("shaderc=", dockerfile)
+        key = (PACKAGING / "lunarg-signing-key-pub.asc").read_bytes()
+        key = key.replace(b"\r\n", b"\n")
+        self.assertEqual(
+            "aa1c3c29673140e77f0d6a9aaeed5d9b5621e305ead51c59fae4458bbb4df92b",
+            hashlib.sha256(key).hexdigest(),
+        )
+
+    def test_the_bundle_has_portable_dynamic_backends(self):
+        script = (PACKAGING / "build-package.sh").read_text(
+            encoding="utf-8")
+        for flag in ("GGML_BACKEND_DL=ON", "GGML_CPU_ALL_VARIANTS=ON",
+                     "GGML_NATIVE=OFF", "GGML_OPENMP=OFF",
+                     "GGML_VULKAN=ON"):
+            self.assertIn(flag, script)
+        self.assertIn("libggml-cpu*.so", script)
+        self.assertIn("libggml-vulkan.so", script)
+
+    def test_the_dependency_release_matches_the_installer(self):
+        workflow = WORKFLOW.read_text(encoding="utf-8")
+        script = (PACKAGING / "build-package.sh").read_text(
+            encoding="utf-8")
+        self.assertEqual("whisper.cpp-v1.9.3",
+                         ggml.MANAGED_WHISPER_RELEASE)
+        self.assertEqual("v1.9.3", ggml.MANAGED_WHISPER_VERSION)
+        self.assertIn("RELEASE_TAG: whisper.cpp-v${{ inputs.whisper_version }}",
+                      workflow)
+        self.assertIn("WHISPER_VERSION:=1.9.3", script)
+        commit = "371b5a7561823ab2bb32142d2751e35e7534727b"
+        self.assertIn(f"WHISPER_COMMIT:={commit}", script)
+        self.assertIn(commit, workflow)
+        self.assertIn(ggml.MANAGED_WHISPER_VULKAN, workflow)
+        self.assertIn(ggml.MANAGED_WHISPER_SHA256, workflow)
+
+    def test_the_bundle_carries_metadata_and_all_required_licenses(self):
+        script = (PACKAGING / "build-package.sh").read_text(
+            encoding="utf-8")
+        for name in ("BUILD-INFO.json", "SHA256SUMS", ".cdx.json"):
+            self.assertIn(name, script)
+        for name in ("cpp-httplib-MIT.txt", "nlohmann-json-MIT.txt"):
+            self.assertTrue((PACKAGING / "licenses" / name).is_file())
+
+    def _make_test_sbom(self):
+        with tempfile.TemporaryDirectory() as temporary:
+            root = pathlib.Path(temporary)
+            (root / "whisper-server").write_bytes(b"elf")
+            sbom = root / "whisper-bin-ubuntu-vulkan-x64.cdx.json"
+            environment = os.environ | {
+                "ROOT": str(root),
+                "VERSION": "1.9.3",
+                "COMMIT": "371b5a7561823ab2bb32142d2751e35e7534727b",
+                "EPOCH": "1787219223",
+            }
+            with sbom.open("w", encoding="utf-8") as output:
+                subprocess.run(
+                    [sys.executable, PACKAGING / "make-sbom.py"],
+                    env=environment, stdout=output, check=True,
+                )
+            return json.loads(sbom.read_text(encoding="utf-8")), sbom.name
+
+    def test_the_sbom_does_not_record_the_file_being_written(self):
+        document, sbom_name = self._make_test_sbom()
+        names = {component["name"] for component in document["components"]}
+        self.assertNotIn(sbom_name, names)
+
+    def test_the_sbom_lists_ggml(self):
+        document, _ = self._make_test_sbom()
+        names = {component["name"] for component in document["components"]}
+        self.assertIn("ggml", names)
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/test_paths.py

@@ -42,9 +42,12 @@ class Directories(unittest.TestCase):
 
     def test_a_mac_does_not_read_the_xdg_variables(self):
         """A Mac with them set from some other tool still stores in one place."""
-        with mock.patch.dict(os.environ, {"XDG_CONFIG_HOME": "/c"}):
+        # Something no temporary directory can be called: the home this runs
+        # under is a mkdtemp path, and a two-letter needle matched the "/c" in
+        # somebody's TMPDIR rather than the variable being read.
+        with mock.patch.dict(os.environ, {"XDG_CONFIG_HOME": "/xdg-elsewhere"}):
             config_dir, _ = paths.directories("darwin")
-        self.assertNotIn("/c", config_dir.as_posix())
+        self.assertNotIn("xdg-elsewhere", config_dir.as_posix())
 
     def test_windows_keeps_the_models_out_of_the_roaming_profile(self):
         """Settings roam with the account; several gigabytes must not."""

tests/test_ui.py

@@ -6,13 +6,15 @@ save, so a setting added to one half and not the other is silently reset the
 next time anybody presses Save. That is the failure this catches.
 """
 
+import json
 import os
 import sys
+import time
 import unittest
 from typing import ClassVar
 from unittest import mock
 
-from PyQt6.QtCore import QPoint, QPointF, Qt
+from PyQt6.QtCore import QPoint, QPointF, QRect, Qt
 from PyQt6.QtGui import QWheelEvent
 from PyQt6.QtWidgets import QApplication, QMessageBox
 
@@ -21,11 +23,13 @@ from dikte import cleanup
 from dikte import config as cfg
 from dikte import ggml
 from dikte import hotkey
+from dikte import hub
 from dikte import ipc
 from dikte import overlay as overlay_module
 from dikte import paste
 from dikte import settings_ui
 from dikte import update
+from dikte.i18n import t
 from tests.support import DikteTest, only_these_tools
 
 # The harness below replaces this method on the class so that opening a window
@@ -48,6 +52,7 @@ CHANGED = {
     "restore_clipboard": True,
     "overlay_corner": "top-right",
     "overlay_screen": "DP-1",
+    "overlay_follows_pointer": True,
     "max_seconds": 120,
     "skip_silent": False,
     "silence_db": -42.0,
@@ -78,6 +83,8 @@ CHANGED = {
     "local_llm_gpu": False,
     "local_llm_preload": True,
     "local_llm_reasoning": "low",
+    "local_idle_unload": False,
+    "local_idle_minutes": 45,
     "cleanup_prompt": "Only fix the punctuation.",
     "file_cleanup_prompt": "Keep the stamps where they are.",
     "transcribe_prompt": "Paraşüt, OpenFrame",
@@ -239,6 +246,39 @@ class Settings(DikteTest):
         label.resize(2000, line)
         self.assertLessEqual(label.minimumHeight(), line)
 
+    def test_a_label_written_before_the_layout_places_it_claims_nothing(self):
+        # The publisher note is written while the settings window is still
+        # being built, when the label is a handful of pixels wide. Wrapped
+        # against that width the sentence became a hundred lines, and the
+        # minimum taken from it did not stay a minimum: QLabel folds it into
+        # its own cached size hints and clears that cache only when the text
+        # changes. The group box stood thousands of pixels tall, with the
+        # model box and everything under it off the bottom of the window,
+        # until another publisher was picked.
+        label = settings_ui.WrappedLabel()
+        self.addCleanup(label.deleteLater)
+        line = label.fontMetrics().height()
+        label.resize(8, line)
+        label.setText("Google Gemma 4, the small one. The default: nothing "
+                      "else this size follows an instruction as closely, and "
+                      "cleanup is all instruction.")
+        self.assertEqual(label.minimumHeight(), 0)
+        # Placed and shown, which is the first width worth measuring against.
+        # The room the wrapping needs is claimed then, and it is the lines the
+        # sentence takes at this width rather than at the last one. Counted
+        # off the font rather than written down here, because how many lines
+        # 400 pixels hold is a different answer on every machine.
+        label.resize(400, line)
+        label.show()
+        wrap = Qt.TextFlag.TextWordWrap | Qt.TextFlag.TextWrapAnywhere
+        needed = label.fontMetrics().boundingRect(
+            QRect(0, 0, 400, 0), wrap, label.text()).height()
+        self.assertGreater(needed, line)      # or the sentence never wrapped
+        self.assertEqual(label.minimumHeight(), needed)
+        # And the label's own hints are the wrapping at this width too, not
+        # the hundred lines the eight pixel one asked for.
+        self.assertLessEqual(label.sizeHint().height(), 3 * needed)
+
     def test_saving_without_touching_anything_changes_nothing(self):
         """Every widget has to load what is stored, or Save writes its default
         over it. This says so for the whole table at once."""
@@ -504,6 +544,20 @@ class Settings(DikteTest):
         self.assertEqual(conf["transcribe_model"], "gpt-4o-transcribe")
         self.assertEqual(conf["groq_transcribe_model"], "whisper-large-v3")
 
+    def test_the_file_model_is_saved_and_only_shown_for_openrouter(self):
+        self.write_config({"transcribe_provider": "openrouter",
+                           "openrouter_file_model": "openai/whisper-large-v3"})
+        conf = cfg.Config()
+        window = self.window(conf)
+        self.assertEqual(window.file_model.currentText(), "openai/whisper-large-v3")
+        self.assertTrue(window.stt_form.isRowVisible(window.file_model_row))
+        window.file_model.setCurrentText(" deepgram/nova-3 ")
+        window._save()
+        self.assertEqual(conf["openrouter_file_model"], "deepgram/nova-3")
+        window.transcribe_provider.setCurrentIndex(
+            window.transcribe_provider.findData("openai"))
+        self.assertFalse(window.stt_form.isRowVisible(window.file_model_row))
+
     def test_the_provider_box_offers_every_provider_config_knows(self):
         window = self.window(cfg.Config())
         offered = [window.transcribe_provider.itemData(i)
@@ -1051,6 +1105,121 @@ class Overlay(DikteTest):
         screen_at.assert_not_called()
         self.assertEqual(widget.pos(), QPoint(1948, 995))
 
+    def _screen(self, name, area):
+        screen = mock.Mock()
+        screen.name.return_value = name
+        screen.availableGeometry.return_value = area
+        return screen
+
+    def _kwin(self, *answer):
+        kwin = mock.Mock()
+        kwin.isValid.return_value = True
+        kwin.call.return_value.arguments.return_value = list(answer)
+        return kwin
+
+    def test_the_compositor_says_which_screen_the_pointer_is_on(self):
+        """Wayland tells a client where the pointer is only while it is over one
+        of that client's own windows, so QCursor.pos() comes back at the origin
+        and every indicator lands on whichever screen holds it. KWin knows."""
+        screens = [self._screen("DP-1", settings_ui.QRect(0, 0, 1920, 1080)),
+                   self._screen("DP-2", settings_ui.QRect(1920, 0, 1920, 1080))]
+        widget = self.overlay()
+        with mock.patch.object(overlay_module, "_kwin", self._kwin("DP-2")), \
+                mock.patch.object(QApplication, "screens", return_value=screens), \
+                mock.patch.object(QApplication, "screenAt") as screen_at:
+            widget._reposition()
+        screen_at.assert_not_called()
+        self.assertEqual(widget.pos(), QPoint(1948, 995))
+
+    def test_the_pointer_decides_when_the_compositor_will_not_say(self):
+        """Every desktop but Plasma, and Plasma while KWin is being replaced."""
+        screens = [self._screen("DP-1", settings_ui.QRect(0, 0, 1920, 1080))]
+        widget = self.overlay()
+        with mock.patch.object(overlay_module, "_kwin", self._kwin()), \
+                mock.patch.object(QApplication, "screens", return_value=screens), \
+                mock.patch.object(QApplication, "screenAt",
+                                  return_value=screens[0]) as screen_at:
+            widget._reposition()
+        screen_at.assert_called()
+        self.assertEqual(widget.pos(), QPoint(28, 995))
+
+    def _two_screens(self):
+        return [self._screen("DP-1", settings_ui.QRect(0, 0, 1920, 1080)),
+                self._screen("DP-2", settings_ui.QRect(1920, 0, 1920, 1080))]
+
+    def _ticks_on(self, widget, screens, kwin):
+        """Run the ribbon long enough for one look at where the pointer is."""
+        with mock.patch.object(overlay_module, "_kwin", kwin), \
+                mock.patch.object(QApplication, "screens", return_value=screens), \
+                mock.patch.object(QApplication, "screenAt", return_value=screens[0]):
+            for _ in range(overlay_module.FOLLOW_EVERY):
+                widget._tick()
+
+    def test_it_can_be_told_to_keep_up_with_the_pointer(self):
+        """The screen it started on is not always the screen you end up on."""
+        screens = self._two_screens()
+        kwin = self._kwin("DP-2")
+        widget = self.overlay(follow_pointer=True)
+        with mock.patch.object(overlay_module, "_kwin", kwin), \
+                mock.patch.object(QApplication, "screens", return_value=screens):
+            widget.show_recording()
+        self.assertEqual(widget.pos(), QPoint(1948, 995))
+        kwin.call.return_value.arguments.return_value = ["DP-1"]
+        self._ticks_on(widget, screens, kwin)
+        self.assertEqual(widget.pos(), QPoint(28, 995))
+
+    def test_it_stays_where_it_appeared_unless_it_was_told_otherwise(self):
+        """Left off, because an indicator that jumps desks mid-sentence is one
+        more thing moving while you are trying to talk."""
+        screens = self._two_screens()
+        kwin = self._kwin("DP-2")
+        widget = self.overlay()
+        with mock.patch.object(overlay_module, "_kwin", kwin), \
+                mock.patch.object(QApplication, "screens", return_value=screens):
+            widget.show_recording()
+        kwin.call.return_value.arguments.return_value = ["DP-1"]
+        self._ticks_on(widget, screens, kwin)
+        self.assertEqual(widget.pos(), QPoint(1948, 995))
+
+    def test_a_named_screen_is_never_left_for_the_pointer(self):
+        """Naming one is the whole answer; following it would undo the naming."""
+        screens = self._two_screens()
+        kwin = self._kwin("DP-2")
+        widget = self.overlay(screen_name="DP-1", follow_pointer=True)
+        with mock.patch.object(QApplication, "screens", return_value=screens):
+            widget.show_recording()
+        self._ticks_on(widget, screens, kwin)
+        kwin.call.assert_not_called()
+        self.assertEqual(widget.pos(), QPoint(28, 995))
+
+    def test_the_one_on_top_goes_where_the_one_underneath_is(self):
+        """Asking for itself would put the pair on two monitors, with this one
+        raised over a ribbon that is not underneath it."""
+        screens = self._two_screens()
+        kwin = self._kwin("DP-2")
+        first = self.overlay()
+        with mock.patch.object(overlay_module, "_kwin", kwin), \
+                mock.patch.object(QApplication, "screens", return_value=screens):
+            first.show_recording()
+            kwin.call.return_value.arguments.return_value = ["DP-1"]
+            second = self.overlay(below=first)
+            second.show_busy("Asking Claude…")
+        self.assertEqual(first.pos(), QPoint(1948, 995))
+        self.assertEqual(second.pos(), QPoint(1948, 929))
+
+    def test_the_compositor_is_asked_only_now_and_then(self):
+        """Every tick would be thirty conversations a second about a hand
+        moving a mouse."""
+        screens = self._two_screens()
+        kwin = self._kwin("DP-2")
+        widget = self.overlay(follow_pointer=True)
+        with mock.patch.object(overlay_module, "_kwin", kwin), \
+                mock.patch.object(QApplication, "screens", return_value=screens):
+            widget.show_recording()
+        kwin.call.reset_mock()
+        self._ticks_on(widget, screens, kwin)
+        self.assertEqual(kwin.call.call_count, 1)
+
     def test_a_warning_and_an_error_both_show(self):
         widget = self.overlay()
         widget.show_warning("cleanup failed")
@@ -1202,6 +1371,28 @@ class LocalModels(DikteTest):
         self.assertIn("10", box.program_label.text())
         self.assertIn("20", box.status.text())
 
+    def test_a_download_says_something_before_the_first_byte(self):
+        # Opening the connection takes ten or twenty seconds, and the byte
+        # counts only start after it. The line underneath still read "has not
+        # been downloaded yet" beside a button that now said Stop, so a
+        # download that had started looked like a click that had not landed.
+        box = self.window(cfg.Config()).local_llm
+        box.load("", "ggml-org/SmolLM3-3B-GGUF")
+        box.repo.blockSignals(True)
+        box.repo.setCurrentText("ggml-org/SmolLM3-3B-GGUF")
+        box.repo.blockSignals(False)
+        box._on_listed([("models", [self._item("SmolLM3-Q4_K_M.gguf")],
+                         "ggml-org/SmolLM3-3B-GGUF")], "")
+        with mock.patch.object(settings_ui.threading, "Thread"):
+            box._download()
+        self.assertIn("Starting", box.status.text())
+        # And the same again for the stop, which is read between blocks and so
+        # not read at all while the connection is still being opened.
+        with mock.patch.object(settings_ui.threading, "Thread"):
+            box._download()
+        self.assertTrue(box._stop)
+        self.assertIn("Stopping", box.status.text())
+
     def test_a_long_model_name_is_not_cut_in_half(self):
         # The list under a combo box takes the box's width and elides what does
         # not fit, in the middle: "ggml-org/Qwen....7B-Base-GGUF".
@@ -1214,6 +1405,319 @@ class LocalModels(DikteTest):
                      for row in range(box.repo.count()))
         self.assertGreaterEqual(view.minimumWidth(), widest)
 
+    @staticmethod
+    def _item(name, size=1 << 20):
+        return hub.Item(name, f"https://example.invalid/{name}", size, "")
+
+    @staticmethod
+    def _rows(box):
+        """Every row's text, headings included."""
+        return [box.model.itemText(row) for row in range(box.model.count())]
+
+    @staticmethod
+    def _repos(box):
+        return [box.repo.itemText(row) for row in range(box.repo.count())]
+
+    @staticmethod
+    def _roomy():
+        """Stand on a machine with room for every suggestion.
+
+        The order the publishers come in follows the memory, so a test that
+        reads it has to say which machine it is standing on. A build runner
+        with 7 GB in it puts the two Gemma 4 rows last and is right to.
+        """
+        return mock.patch.object(ggml, "total_memory", return_value=64 << 30)
+
+    @staticmethod
+    def _offered(box):
+        """The model names in the box, headings and duplicates left out."""
+        names = []
+        for row in range(box.model.count()):
+            name = box.model.itemData(row)
+            if name and name not in names:
+                names.append(name)
+        return names
+
+    def test_a_row_with_nothing_to_fetch_does_not_offer_a_download(self):
+        # The model the settings name is not in the list any more, so its row
+        # was rebuilt from the name alone and carries no file to fetch. The
+        # button stayed lit and the press did nothing at all.
+        box = self.window(self.config(local_llm_model="gone.gguf")).local_llm
+        box.load("gone.gguf", "ggml-org/SmolLM3-3B-GGUF")
+        self.assertEqual(box.selected(), "gone.gguf")
+        self.assertFalse(box.download_button.isEnabled())
+        self.assertIn("gone.gguf", box.status.text())
+        self.assertIn("publisher", box.status.text())
+
+    def test_a_model_without_its_program_does_not_say_it_is_ready(self):
+        # The model runs on the program above it, and "Ready" over a missing
+        # one is what had people asking why nothing transcribed.
+        box = self.window(cfg.Config()).local_whisper
+        path = ggml.whisper_model_path("ggml-small.bin")
+        path.parent.mkdir(parents=True, exist_ok=True)
+        path.write_bytes(b"not really a model")
+        box.load("ggml-small.bin")
+        self.assertFalse(ggml.program_path(ggml.WHISPER))
+        self.assertNotIn("Ready", box.status.text())
+        self.assertIn("program", box.status.text())
+
+    def test_changing_the_publisher_changes_the_model(self):
+        # The model chosen under the old publisher is not published by the new
+        # one. Carried over, it was added back as "not downloaded" and selected
+        # again, and the box looked as though the change had not taken.
+        box = self.window(self.config(local_llm_model="gemma-3-4b-it-Q4_K_M.gguf",
+                                      local_llm_repo="ggml-org/gemma-3-4b-it-GGUF")).local_llm
+        box.load("gemma-3-4b-it-Q4_K_M.gguf", "ggml-org/gemma-3-4b-it-GGUF")
+        box.repo.blockSignals(True)
+        box.repo.setCurrentText("ggml-org/SmolLM3-3B-GGUF")
+        box.repo.blockSignals(False)
+        box._on_listed([("models", [self._item("SmolLM3-Q4_K_M.gguf")],
+                         "ggml-org/SmolLM3-3B-GGUF")], "")
+        self.assertEqual(box.selected(), "SmolLM3-Q4_K_M.gguf")
+        self.assertEqual(self._offered(box), ["SmolLM3-Q4_K_M.gguf"])
+
+    def test_a_list_for_a_publisher_that_is_no_longer_chosen_is_dropped(self):
+        # Every change starts its own request, and they do not come back in the
+        # order they went out.
+        box = self.window(cfg.Config()).local_llm
+        box.load("", "ggml-org/SmolLM3-3B-GGUF")
+        box.repo.blockSignals(True)
+        box.repo.setCurrentText("ggml-org/SmolLM3-3B-GGUF")
+        box.repo.blockSignals(False)
+        box._on_listed([("models", [self._item("SmolLM3-Q4_K_M.gguf")],
+                         "ggml-org/SmolLM3-3B-GGUF")], "")
+        box._on_listed([("models", [self._item("gemma-3-4b-it-Q4_K_M.gguf")],
+                         "ggml-org/gemma-3-4b-it-GGUF")], "")
+        self.assertEqual(box.selected(), "SmolLM3-Q4_K_M.gguf")
+
+    def test_the_publisher_box_is_not_asked_on_every_keystroke(self):
+        box = self.window(cfg.Config()).local_llm
+        with mock.patch.object(box, "_fetch_models") as fetch:
+            for text in ("g", "gg", "ggm", "ggml-org/SmolLM3-3B-GGUF"):
+                box.repo.setCurrentText(text)
+            fetch.assert_not_called()
+            box._later.setInterval(0)
+            box._later.start()
+            _app.processEvents()
+            time.sleep(0.05)
+            _app.processEvents()
+        self.assertEqual(fetch.call_count, 1)
+    def test_the_models_are_grouped_by_the_model_rather_than_by_size(self):
+        # Sorted by size alone, the turbo files land between the two medium
+        # ones, half a screen from the model they are a copy of.
+        box = self.window(cfg.Config()).local_whisper
+        with mock.patch.object(ggml, "total_memory", return_value=8 << 30), \
+                mock.patch.object(ggml, "accelerator", return_value=""):
+            box._on_listed([("models", [
+                self._item("ggml-medium-q5_0.bin", 539 << 20),
+                self._item("ggml-large-v3-turbo-q5_0.bin", 574 << 20),
+                self._item("ggml-medium-q8_0.bin", 823 << 20),
+                self._item("ggml-large-v3-turbo.bin", 1624 << 20),
+            ], "")], "")
+        rows = self._rows(box)
+        # The two medium files under one heading, the two turbo ones under
+        # theirs, and the model rather than the file deciding the order.
+        self.assertEqual(rows[rows.index("medium"):],
+                         ["medium",
+                          "ggml-medium-q5_0.bin  (539.0 MB, 5-bit)",
+                          "ggml-medium-q8_0.bin  (823.0 MB, 8-bit)",
+                          "large-v3-turbo",
+                          "ggml-large-v3-turbo-q5_0.bin  "
+                          "(574.0 MB, 5-bit, recommended)",
+                          "ggml-large-v3-turbo.bin  (1.6 GB, 16-bit)"])
+        # A heading is not a model, and nothing can be saved from one.
+        self.assertIsNone(box.model.itemData(rows.index("medium")))
+
+    def test_the_row_for_this_machine_is_on_top_and_says_so(self):
+        box = self.window(cfg.Config()).local_whisper
+        with mock.patch.object(ggml, "total_memory", return_value=8 << 30), \
+                mock.patch.object(ggml, "accelerator", return_value=""):
+            box._on_listed([("models", [
+                self._item("ggml-tiny.bin", 77 << 20),
+                self._item("ggml-large-v3-turbo-q5_0.bin", 574 << 20),
+            ], "")], "")
+        self.assertEqual(box.selected(), "ggml-large-v3-turbo-q5_0.bin")
+        self.assertEqual(box.model.itemData(1), "ggml-large-v3-turbo-q5_0.bin")
+        self.assertIn(t("recommended"), box.model.itemText(1))
+
+    def test_a_model_the_memory_cannot_hold_says_so_on_its_row(self):
+        box = self.window(cfg.Config()).local_llm
+        box.repo.blockSignals(True)
+        box.repo.setCurrentText("ggml-org/x-GGUF")
+        box.repo.blockSignals(False)
+        with mock.patch.object(ggml, "total_memory", return_value=8 << 30):
+            box._on_listed([("models", [
+                self._item("small-Q4_0.gguf", 1 << 30),
+                self._item("huge-Q8_0.gguf", 12 << 30),
+            ], "ggml-org/x-GGUF")], "")
+        rows = {box.model.itemData(row): box.model.itemText(row)
+                for row in range(box.model.count())}
+        self.assertNotIn(t("too big for this machine"), rows["small-Q4_0.gguf"])
+        self.assertIn(t("too big for this machine"), rows["huge-Q8_0.gguf"])
+
+    def test_a_recommended_row_is_not_listed_twice_after_a_download(self):
+        # It has a row of its own on top as well as one in its group, and
+        # reading the rows back the way a finished download does was doubling
+        # it in the list every time.
+        box = self.window(cfg.Config()).local_whisper
+        with self._roomy():
+            box._on_listed([("models", [
+                self._item("ggml-tiny.bin", 77 << 20),
+                self._item("ggml-large-v3-turbo-q5_0.bin", 574 << 20),
+            ], "")], "")
+            before = self._offered(box)
+            box._fill_models_from_current()
+        self.assertEqual(self._offered(box), before)
+        names = [box.model.itemData(row) for row in range(box.model.count())]
+        self.assertEqual(len([n for n in names if n]), len(before) + 1)
+
+    def test_a_processor_build_is_not_recommended_the_accurate_model(self):
+        # The Vulkan loader is on the machine but what was installed is the
+        # processor build, so there is no card in play whatever the loader
+        # says, and a 1 GB model on a processor is a wait somebody is sitting
+        # through with a sentence half typed.
+        binary = self.path("bin/whisper/v1.9.3/whisper-server")
+        binary.parent.mkdir(parents=True)
+        binary.write_text("")
+        binary.chmod(0o755)
+        self.path("bin/whisper/installed.json").write_text(json.dumps(
+            {"tag": "v1.9.3", "binary": str(binary), "backend": "processor"}))
+        self.patch_attr(ggml.shutil, "which", lambda name: None)
+        box = self.window(cfg.Config()).local_whisper
+        with mock.patch.object(ggml, "total_memory", return_value=32 << 30), \
+                mock.patch.object(ggml, "accelerator", return_value="Vulkan"):
+            self.assertEqual(box._suggested(), ggml.SUGGESTED_WHISPER)
+
+    def test_a_publisher_with_nothing_to_offer_says_why(self):
+        # Half of what ggml-org publishes is split across files or past the
+        # size cap, and an empty box read as though the click had not landed.
+        box = self.window(cfg.Config()).local_llm
+        box.repo.blockSignals(True)
+        box.repo.setCurrentText("ggml-org/gpt-oss-120b-GGUF")
+        box.repo.blockSignals(False)
+        box._on_listed([("models", [], "ggml-org/gpt-oss-120b-GGUF")], "")
+        self.assertIn("ggml-org/gpt-oss-120b-GGUF", box.status.text())
+        self.assertIn("publisher", box.status.text())
+
+    def test_an_empty_box_nobody_has_asked_yet_is_not_a_publisher_fault(self):
+        box = self.window(cfg.Config()).local_llm
+        box.load("", "ggml-org/SmolLM3-3B-GGUF")
+        self.assertNotIn("publisher", box.status.text())
+
+    def test_only_the_suggested_publishers_are_offered_to_start_with(self):
+        # Forty repository ids is not a choice anybody can make.
+        box = self.window(cfg.Config()).local_llm
+        with self._roomy():
+            box._on_listed([("repos", [ggml.SUGGESTED_LLM[0],
+                                       "ggml-org/something-else-GGUF"], "")], "")
+            self.assertEqual(self._repos(box), list(ggml.SUGGESTED_LLM))
+
+    def test_a_suggestion_missing_from_the_listing_is_still_offered(self):
+        # The listing is the forty repositories touched most recently, and a
+        # publisher that has not been updated in a season falls off it while
+        # still being the one to point at.
+        box = self.window(cfg.Config()).local_llm
+        box._on_listed([("repos", ["ggml-org/something-else-GGUF"], "")], "")
+        self.assertIn(ggml.SUGGESTED_LLM[0], self._repos(box))
+
+    def test_the_switch_brings_the_rest_and_keeps_them_apart(self):
+        box = self.window(cfg.Config()).local_llm
+        with self._roomy():
+            box._on_listed([("repos", [ggml.SUGGESTED_LLM[0],
+                                       "ggml-org/something-else-GGUF"], "")], "")
+            box.every_repo.setChecked(True)
+            rows = self._repos(box)
+        self.assertEqual(rows[:len(ggml.SUGGESTED_LLM)],
+                         list(ggml.SUGGESTED_LLM))
+        # A separator rather than a heading: the box is typed into as well as
+        # chosen from, and a heading would land in the field as a repository.
+        self.assertEqual(rows[len(ggml.SUGGESTED_LLM)], "")
+        self.assertEqual(rows[-1], "ggml-org/something-else-GGUF")
+
+    def test_a_publisher_typed_in_is_not_dropped_by_the_next_fetch(self):
+        box = self.window(cfg.Config()).local_llm
+        box.repo.blockSignals(True)
+        box.repo.setCurrentText("ggml-org/something-else-GGUF")
+        box.repo.blockSignals(False)
+        box._on_listed([("repos", [ggml.SUGGESTED_LLM[0],
+                                   "ggml-org/something-else-GGUF"], "")], "")
+        self.assertFalse(box.every_repo.isChecked())
+        self.assertIn("ggml-org/something-else-GGUF", self._repos(box))
+        self.assertEqual(box.repository(), "ggml-org/something-else-GGUF")
+
+    def test_the_chosen_publisher_is_said_in_words(self):
+        # A repository id names the publisher, the parameter count and the
+        # shape of the weights, and none of that says whether to click it.
+        box = self.window(cfg.Config()).local_llm
+        box.repo.setCurrentText(ggml.SUGGESTED_LLM[0])
+        self.assertTrue(box.repo_note.text())
+        box.repo.setCurrentText("ggml-org/nobody-wrote-a-note-GGUF")
+        self.assertEqual(box.repo_note.text(), "")
+
+    def test_the_box_says_what_this_machine_will_run_on(self):
+        box = self.window(cfg.Config()).local_whisper
+        with mock.patch.object(ggml, "accelerator", return_value="Vulkan"), \
+                mock.patch.object(ggml, "total_memory", return_value=32 << 30):
+            box._show_machine()
+        self.assertIn("Vulkan", box.machine_label.text())
+        self.assertIn("32.0 GB", box.machine_label.text())
+
+    def test_a_machine_with_no_card_is_told_it_is_on_the_processor(self):
+        box = self.window(cfg.Config()).local_whisper
+        with mock.patch.object(ggml, "accelerator", return_value=""), \
+                mock.patch.object(ggml, "total_memory", return_value=8 << 30):
+            box._show_machine()
+        self.assertIn("processor", box.machine_label.text())
+
+    def test_a_processor_build_where_the_vulkan_one_belongs_says_so(self):
+        # The Vulkan whisper-server is published by hand, and until it is
+        # there the download lands upstream's processor build. Said nowhere,
+        # an idle graphics card looks exactly like one that is being used.
+        binary = self.path("bin/whisper/v1.9.3/whisper-server")
+        binary.parent.mkdir(parents=True)
+        binary.write_text("")
+        binary.chmod(0o755)
+        self.path("bin/whisper/installed.json").write_text(json.dumps(
+            {"tag": "v1.9.3", "binary": str(binary), "backend": "processor"}))
+        # A whisper-server on this machine's PATH would win over the download.
+        self.patch_attr(ggml.shutil, "which", lambda name: None)
+        label = self.window(cfg.Config()).local_whisper.program_label.text()
+        self.assertIn("v1.9.3", label)
+        self.assertIn("Vulkan", label)
+
+    def test_an_ordinary_install_is_reported_without_a_word_about_vulkan(self):
+        binary = self.path("bin/whisper/v1.9.3/whisper-server")
+        binary.parent.mkdir(parents=True)
+        binary.write_text("")
+        binary.chmod(0o755)
+        self.path("bin/whisper/installed.json").write_text(json.dumps(
+            {"tag": "v1.9.3", "binary": str(binary)}))
+        self.patch_attr(ggml.shutil, "which", lambda name: None)
+        label = self.window(cfg.Config()).local_whisper.program_label.text()
+        self.assertIn("v1.9.3", label)
+        self.assertNotIn("Vulkan", label)
+
+    def test_a_downloaded_program_can_still_be_asked_for_again(self):
+        # The button used to disappear the moment anything landed, which left
+        # no way to pick up a newer whisper.cpp, or the Vulkan build on a
+        # machine whose driver was installed after Dikte was.
+        binary = self.path("bin/whisper/v1.9.3/whisper-server")
+        binary.parent.mkdir(parents=True)
+        binary.write_text("")
+        binary.chmod(0o755)
+        self.path("bin/whisper/installed.json").write_text(json.dumps(
+            {"tag": "v1.9.3", "binary": str(binary)}))
+        self.patch_attr(ggml.shutil, "which", lambda name: None)
+        box = self.window(cfg.Config()).local_whisper
+        self.assertTrue(box.install_button.isVisibleTo(box))
+        self.assertEqual(box.install_button.text(), t("Download again"))
+
+    def test_a_system_copy_is_not_offered_for_download(self):
+        # Nothing Dikte downloads would be run while one is on the PATH.
+        self.patch_attr(ggml.shutil, "which", lambda name: "/usr/bin/" + name)
+        box = self.window(cfg.Config()).local_whisper
+        self.assertFalse(box.install_button.isVisibleTo(box))
+
     def test_only_the_chosen_transcriber_is_on_screen(self):
         window = self.window(self.config(transcribe_provider="openai"))
         self.assertTrue(window.stt_form.isRowVisible(window.transcribe_model_row))
@@ -1232,6 +1736,26 @@ class LocalModels(DikteTest):
         # Its own thinking box, because the two default to opposite things.
         self.assertFalse(window.cleanup_form.isRowVisible(window.cleanup_reasoning))
 
+    def test_the_idle_unload_is_offered_to_whoever_runs_a_model_here(self):
+        for transcriber, cleaner in (("local", "openrouter"),
+                                     ("openai", "local"),
+                                     ("local", "local")):
+            with self.subTest(transcriber=transcriber, cleaner=cleaner):
+                window = self.window(self.config(transcribe_provider=transcriber,
+                                                 cleanup_provider=cleaner))
+                self.assertTrue(window.local_box.isVisibleTo(window))
+
+    def test_a_machine_that_runs_neither_is_not_asked_about_memory(self):
+        window = self.window(self.config(transcribe_provider="openai",
+                                         cleanup_provider="openrouter"))
+        self.assertFalse(window.local_box.isVisibleTo(window))
+
+    def test_the_minutes_follow_the_checkbox(self):
+        window = self.window(self.config(local_idle_unload=False))
+        self.assertFalse(window.local_idle_minutes.isEnabled())
+        window.local_idle_unload.setChecked(True)
+        self.assertTrue(window.local_idle_minutes.isEnabled())
+
     def test_each_cleaner_brings_its_own_model_row_and_no_other(self):
         window = self.window(cfg.Config())
         rows = {"openrouter": window.cleanup_model_row,