Merge master and correct local acceleration reporting

Reject failed Whisper GPU attempts, identify Llama devices from model buffers, and keep historical diagnostics independent of current settings. Report loaded backends without inferring build support, with regression coverage for each case.

tests/support.py

@@ -112,6 +112,10 @@ 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")
 
@@ -691,6 +827,40 @@ class TranscribeHere(DikteTest):
             api.transcribe_segments(LOCAL, self.wav)
         self.assertEqual(multipart_fields(calls[0])["model"], "ggml-base.bin")
 
+    # ---- the detected language --------------------------------------------
+
+    def test_auto_mode_asks_whisper_for_the_detected_language(self):
+        # The -nlp the server was started with is switched back on for this one
+        # request, so whisper's verbose_json reports what it heard.
+        reply = {"text": " Merhaba dünya. ", "detected_language": "turkish"}
+        with fake_urlopen(reply) as calls:
+            text, code = api.transcribe_detected(LOCAL, self.wav, language="auto")
+        fields = multipart_fields(calls[0])
+        self.assertEqual(fields["response_format"], "verbose_json")
+        self.assertEqual(fields["no_language_probabilities"], "false")
+        self.assertNotIn("language", fields)
+        self.assertEqual(text, "Merhaba dünya.")
+        self.assertEqual(code, "tr")
+
+    def test_a_fixed_language_reports_no_detection(self):
+        with fake_urlopen({"text": "hello"}) as calls:
+            text, code = api.transcribe_detected(LOCAL, self.wav, language="tr")
+        self.assertNotIn("no_language_probabilities", multipart_fields(calls[0]))
+        self.assertEqual(text, "hello")
+        self.assertEqual(code, "")
+
+    def test_a_detected_language_without_a_code_stays_unknown(self):
+        with fake_urlopen({"text": "hello", "detected_language": "somali"}):
+            _text, code = api.transcribe_detected(LOCAL, self.wav, language="auto")
+        self.assertEqual(code, "")
+
+    def test_a_hosted_auto_run_transcribes_without_detection(self):
+        with fake_urlopen({"text": "hi"}) as calls:
+            text, code = api.transcribe_detected(OPENAI, self.wav, language="auto")
+        self.assertNotIn("no_language_probabilities", multipart_fields(calls[0]))
+        self.assertEqual(text, "hi")
+        self.assertEqual(code, "")
+
 
 class Stopping(unittest.TestCase):
     """The Stop button, from the far end: a request already blocked on a reply.

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)
@@ -741,6 +746,13 @@ class Replies(DikteTest):
         self.assertEqual(code, 0)
         self.assertEqual(out.strip(), "Book it for Thursday.")
 
+    def test_the_json_answer_carries_the_detected_language(self):
+        code, out, _ = self.run_verb(
+            ["--json", "record"],
+            {"ok": True, "text": "Selam", "speech_language": "tr"})
+        self.assertEqual(code, 0)
+        self.assertEqual(json.loads(out)["speech_language"], "tr")
+
     def test_a_dictation_that_failed(self):
         code, out, err = self.run_verb(["stop", "--wait"],
                                        {"ok": False, "error": "No speech detected"})
@@ -804,18 +816,19 @@ class LocalModels(DikteTest):
         line = self.status({"whisper": self.entry(
             backend="CPU", device="CPU", available=["CPU"])})
         self.assertIn("loaded on the processor", line)
-        self.assertIn("this build carries none", line)
+        self.assertIn("only the CPU backend was loaded", line)
 
-    def test_the_downloaded_build_is_told_where_a_working_one_comes_from(self):
+    def test_a_download_is_not_assumed_to_lack_gpu_support(self):
         line = self.status({"whisper": self.entry(
             backend="CPU", device="CPU", available=["CPU"], downloaded=True)})
-        self.assertIn("downloaded build has no GPU backend", line)
-        self.assertIn("whisper-server on your system", line)
+        self.assertIn("only the CPU backend was loaded", line)
+        self.assertIn("driver errors", line)
+        self.assertNotIn("has no GPU backend", line)
 
     def test_a_card_the_build_could_have_used_says_something_else(self):
         line = self.status({"whisper": self.entry(
             backend="CPU", device="CPU", available=["CUDA", "CPU"])})
-        self.assertIn("none was found", line)
+        self.assertIn("could not be used", line)
         self.assertNotIn("carries none", line)
 
     def test_a_card_nobody_asked_for_is_not_a_complaint(self):
@@ -861,7 +874,31 @@ class LocalModels(DikteTest):
                  "whisper_backend_init_gpu: no GPU found\n")
         line = self.doctor(transcribe_provider="local", local_gpu=True)
         self.assertIn("last run on the processor", line)
-        self.assertIn("this build carries none", line)
+        self.assertNotIn("this build carries none", line)
+        self.assertNotIn("check the server log", line)
+
+    def test_old_cpu_log_does_not_diagnose_a_new_system_binary(self):
+        self.log("load_backend: loaded CPU backend from /old-download.so\n"
+                 "whisper_backend_init_gpu: no GPU found\n")
+        with mock.patch.object(ggml, "program_path",
+                               return_value="/usr/bin/whisper-server"):
+            line = self.doctor(transcribe_provider="local", local_gpu=True)
+            data = self.doctor(as_json=True, transcribe_provider="local",
+                               local_gpu=True)
+        self.assertIn("last run on the processor", line)
+        self.assertNotIn("carries none", line)
+        self.assertNotIn("gpu_wanted", data["local"]["whisper"])
+        self.assertNotIn("downloaded", data["local"]["whisper"])
+
+    def test_enabling_gpu_does_not_reinterpret_a_past_cpu_run(self):
+        self.log("load_backend: loaded Vulkan backend from /gpu.so\n"
+                 "load_backend: loaded CPU backend from /cpu.so\n"
+                 "whisper_init_with_params_no_state: use gpu = 0\n"
+                 "whisper_backend_init_gpu: no GPU found\n")
+        line = self.doctor(transcribe_provider="local", local_gpu=True)
+        self.assertIn("last run on the processor", line)
+        self.assertNotIn("none was found", line)
+        self.assertNotIn("could not be used", line)
 
     def test_a_run_that_named_no_backend_is_not_read_as_no_run_at_all(self):
         # A log with nothing recognisable in it still says a server started

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",
@@ -261,6 +274,18 @@ class CleanupPrompt(DikteTest):
     def test_no_glossary_means_no_rule_about_one(self):
         self.assertEqual(cfg.Config().cleanup_prompt(), cfg.CLEANUP_PROMPT_EN)
 
+    def test_a_detected_turkish_recording_gets_the_turkish_prompt(self):
+        """Auto mode learns what was heard, and that decides the prompt rather
+        than the interface language."""
+        self.write_config({"ui_language": "en", "transcribe_prompt": "Paraşüt"})
+        conf = cfg.Config()
+        prompt = conf.cleanup_prompt(speech="tr")
+        self.assertEqual(prompt, cfg.CLEANUP_PROMPT_TR
+                         + cfg.GLOSSARY_RULE_TR.format(glossary="Paraşüt"))
+        self.assertIn("KONUŞMACININ KULLANDIĞI İSİM VE TERİMLER", prompt)
+        self.assertIn("NAMES AND TERMS THE SPEAKER USES",
+                      conf.cleanup_prompt(speech="de"))
+
     def test_subtitles_use_their_own_prompt(self):
         conf = cfg.Config()
         self.assertNotEqual(conf.cleanup_prompt(subtitles=True), conf.cleanup_prompt())
@@ -676,3 +701,24 @@ 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")
@@ -558,6 +768,7 @@ load_backend: loaded CPU backend from /usr/lib/ggml/libggml-cpu-haswell.so
 whisper_model_load:          CPU total size =   189.49 MB
 whisper_backend_init_gpu: device 0: Vulkan0 (type: 1)
 whisper_backend_init_gpu: found GPU device 0: Vulkan0 (type: 1, cnt: 0)
+whisper_backend_init_gpu: using Vulkan0 backend
 whisper_backend_init_gpu: failed to initialize Vulkan0 backend
 """
 
@@ -599,6 +810,7 @@ load_backend: loaded Vulkan backend from /opt/llama/libggml-vulkan.so
 load_backend: loaded CPU backend from /opt/llama/libggml-cpu.so
 load_tensors: offloading 28 repeating layers to GPU
 load_tensors: offloaded 29/29 layers to GPU
+load_tensors:      Vulkan0 model buffer size = 2048.00 MiB
 """
 
 LLAMA_CPU = """\
@@ -655,9 +867,9 @@ class WhatItRunsOn(Local):
         self.assertEqual(accel.backend, "CPU")
         self.assertEqual(ggml.accel_kind(accel), "cpu")
 
-    def test_a_build_with_nothing_but_a_processor_backend_says_so(self):
-        self.assertTrue(ggml.cpu_only_build(self.read(ggml.WHISPER, WHISPER_CPU)))
-        self.assertFalse(ggml.cpu_only_build(self.read(ggml.WHISPER, WHISPER_CUDA)))
+    def test_only_loaded_backends_are_reported(self):
+        self.assertTrue(ggml.cpu_only_loaded(self.read(ggml.WHISPER, WHISPER_CPU)))
+        self.assertFalse(ggml.cpu_only_loaded(self.read(ggml.WHISPER, WHISPER_CUDA)))
 
     def test_a_card_that_was_found_is_named(self):
         accel = self.read(ggml.WHISPER, WHISPER_CUDA)
@@ -697,6 +909,15 @@ class WhatItRunsOn(Local):
         self.assertEqual(accel.backend, "CPU")
         self.assertEqual(ggml.accel_kind(accel), "cpu")
 
+    def test_failed_initialisation_overrides_weights_on_the_card(self):
+        log = WHISPER_GPU_FAILED.replace("CPU total size", "Vulkan0 total size")
+        self.assertEqual(self.read(ggml.WHISPER, log).backend, "CPU")
+
+    def test_a_later_successful_attempt_replaces_the_failed_one(self):
+        log = WHISPER_GPU_FAILED + (
+            "whisper_backend_init_gpu: using Vulkan0 backend\n")
+        self.assertEqual(self.read(ggml.WHISPER, log).backend, "Vulkan")
+
     def test_the_card_named_is_the_one_that_ran(self):
         accel = self.read(ggml.WHISPER, WHISPER_MIXED)
         self.assertEqual(accel.backend, "CUDA")
@@ -723,12 +944,46 @@ class WhatItRunsOn(Local):
         self.assertEqual(accel.layers, "29/29")
         self.assertEqual(ggml.accel_detail(accel), "Vulkan, 29/29 layers")
 
+    def test_llama_names_the_allocated_device_not_the_first_loaded_backend(self):
+        log = (
+            "load_backend: loaded CUDA backend from /x.so\n"
+            "  Device 0: NVIDIA RTX 4070, compute capability 8.9, VMM: yes\n"
+            "ggml_vulkan: 0 = Intel UHD Graphics | uma: 1\n"
+            "ggml_vulkan: 1 = AMD Radeon RX 6600 | uma: 0\n"
+            "load_tensors:     CUDA_Host model buffer size = 32.00 MiB\n"
+            "load_tensors:       Vulkan0 model buffer size = 0.00 MiB\n"
+            + LLAMA_GPU.replace("Vulkan0 model", "Vulkan1 model"))
+        accel = self.read(ggml.LLAMA, log)
+        self.assertEqual(accel.backend, "Vulkan")
+        self.assertEqual(accel.device, "AMD Radeon RX 6600")
+
+    def test_llama_without_buffer_evidence_does_not_guess_the_backend(self):
+        log = LLAMA_GPU.replace(
+            "load_tensors:      Vulkan0 model buffer size = 2048.00 MiB\n", "")
+        accel = self.read(ggml.LLAMA, log)
+        self.assertEqual((accel.backend, accel.device), ("GPU", ""))
+        self.assertEqual(accel.layers, "29/29")
+
+    def test_llama_split_across_cards_does_not_name_only_one(self):
+        log = LLAMA_GPU + (
+            "load_tensors:      Vulkan1 model buffer size = 1024.00 MiB\n")
+        accel = self.read(ggml.LLAMA, log)
+        self.assertEqual((accel.backend, accel.device), ("GPU", ""))
+
+    def test_llama_static_metal_build_can_be_identified_by_its_buffer(self):
+        log = (
+            "ggml_metal_init: picking default device: Apple M2\n"
+            "load_tensors: offloaded 29/29 layers to GPU\n"
+            "load_tensors:       Metal model buffer size = 2048.00 MiB\n")
+        accel = self.read(ggml.LLAMA, log)
+        self.assertEqual((accel.backend, accel.device), ("Metal", "Apple M2"))
+
     def test_a_llama_that_offloaded_nothing_is_on_the_processor(self):
         accel = self.read(ggml.LLAMA, LLAMA_CPU)
         self.assertEqual(accel.backend, "CPU")
         self.assertEqual(ggml.accel_kind(accel), "cpu")
         # The build could have used the card; this run did not.
-        self.assertFalse(ggml.cpu_only_build(accel))
+        self.assertFalse(ggml.cpu_only_loaded(accel))
 
     def test_the_processor_is_not_named_twice(self):
         # whisper prints CPU as the backend and as the device, and saying it
@@ -776,7 +1031,7 @@ STAND_IN = textwrap.dedent("""
               flush=True)
         print("whisper_backend_init_gpu: device 0: Test Card (type: 1)",
               flush=True)
-        # The line that says one of them worked, which is the one read back.
+        # The attempt is followed by no failure in this stand-in.
         print("whisper_backend_init_gpu: using " + opt("--backend") + "0 backend",
               flush=True)
 
@@ -809,7 +1064,9 @@ 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)
@@ -832,6 +1089,8 @@ 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()
@@ -1107,6 +1366,122 @@ 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."""
 
@@ -1305,3 +1680,205 @@ 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 QHideEvent, QShowEvent, 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,
@@ -72,12 +77,14 @@ CHANGED = {
     "local_model": "ggml-small.bin",
     "local_gpu": False,
     "local_preload": False,
-    "local_threads": 6,
+    "local_threads": 1,
     "local_llm_model": "gemma-3-4b-it-Q4_K_M.gguf",
     "local_llm_repo": "ggml-org/gemma-4-E2B-it-GGUF",
     "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."""
@@ -286,6 +326,56 @@ class Settings(DikteTest):
                 self.assertEqual(shown, [provider])
                 self.assertFalse(box.isHidden())
 
+    def test_editable_boxes_live_in_forms_that_grow_the_field_column(self):
+        window = self.window(cfg.Config())
+
+        def contains(layout, target):
+            for index in range(layout.count()):
+                item = layout.itemAt(index)
+                widget = item.widget()
+                if widget is target or (widget is not None and
+                                        widget.isAncestorOf(target)):
+                    return True
+                child = item.layout()
+                if child is not None and contains(child, target):
+                    return True
+            return False
+
+        forms = window.findChildren(settings_ui.QFormLayout)
+        boxes = [
+            window.paste_shortcut,
+            window.transcribe_model,
+            window.file_model,
+            window.cleanup_model,
+            window.cleanup_gemini_model,
+            window.cleanup_opencode_model,
+            window.cleanup_agy_model,
+            window.cleanup_claude_model,
+            window.cleanup_codex_model,
+            window.assistant_model,
+            window.assistant_agy_model,
+            window.assistant_opencode_model,
+            window.assistant_codex_model,
+            window.assistant_openrouter_model,
+            window.meeting_model,
+            *(box for box, _status, _missing in
+              window._shortcut_rows.values()),
+        ]
+        for box in boxes:
+            form = next((candidate for candidate in forms
+                         if contains(candidate, box)), None)
+            with self.subTest(box=box.objectName() or box.currentText()):
+                self.assertIsNotNone(form)
+                self.assertEqual(
+                    form.fieldGrowthPolicy(),
+                    settings_ui.QFormLayout.FieldGrowthPolicy.AllNonFixedFieldsGrow,
+                )
+
+        self.assertEqual(
+            window.local_llm.layout().fieldGrowthPolicy(),
+            settings_ui.QFormLayout.FieldGrowthPolicy.AllNonFixedFieldsGrow,
+        )
+
     def test_codex_answering_refills_both_of_its_boxes(self):
         """The list Codex gave replaces the built-in one, in both places, and
         neither loses what was already picked."""
@@ -504,6 +594,20 @@ 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 +1155,121 @@ 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")
@@ -1203,26 +1422,25 @@ class LocalModels(DikteTest):
     def test_a_card_asked_for_and_missing_is_not_left_to_be_guessed_at(self):
         whisper, _ = self.shown(backend="CPU", device="CPU", available=["CPU"])
         self.assertIn("processor", whisper)
-        self.assertIn("no graphics backend", whisper)
+        self.assertIn("only the CPU backend was loaded", whisper)
 
-    def test_the_downloaded_build_is_told_what_to_install(self):
-        # whisper.cpp publishes nothing that reaches a card on this system, and
-        # a copy on the PATH is used ahead of Dikte's own, so that is the fix.
+    def test_a_download_is_not_assumed_to_lack_gpu_support(self):
         whisper, _ = self.shown(backend="CPU", device="CPU", available=["CPU"],
                                 downloaded=True)
-        self.assertIn("Dikte downloaded", whisper)
-        self.assertIn("whisper-server", whisper)
+        self.assertIn("only the CPU backend was loaded", whisper)
+        self.assertIn("driver errors", whisper)
+        self.assertNotIn("installing one", whisper)
 
     def test_a_system_build_is_not_told_to_install_itself(self):
         whisper, _ = self.shown(backend="CPU", device="CPU", available=["CPU"],
                                 downloaded=False)
-        self.assertIn("no graphics backend", whisper)
+        self.assertIn("only the CPU backend was loaded", whisper)
         self.assertNotIn("Dikte downloaded", whisper)
 
     def test_a_build_that_could_have_used_one_says_the_other_thing(self):
         whisper, _ = self.shown(backend="CPU", device="CPU",
                                 available=["CUDA", "CPU"])
-        self.assertIn("none was found", whisper)
+        self.assertIn("could not be used", whisper)
         self.assertNotIn("no graphics backend", whisper)
 
     def test_a_processor_nobody_argued_about_is_stated_plainly(self):
@@ -1270,6 +1488,28 @@ 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".
@@ -1282,6 +1522,319 @@ 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))
@@ -1300,6 +1853,26 @@ 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,
@@ -1325,3 +1898,15 @@ class LocalModels(DikteTest):
                     # isHidden rather than isVisible: the window itself is never
                     # shown in a test, so nothing in it is ever visible.
                     self.assertEqual(other.isHidden(), name != chosen)
+
+    def test_local_threads_range_is_bounded_by_cpu_count(self):
+        with mock.patch("os.cpu_count", return_value=8):
+            window = self.window(cfg.Config())
+            self.assertEqual(window.local_threads.minimum(), 0)
+            self.assertEqual(window.local_threads.maximum(), 8)
+
+    def test_local_threads_range_has_safe_minimum_when_cpu_count_is_none(self):
+        with mock.patch("os.cpu_count", return_value=None):
+            window = self.window(cfg.Config())
+            self.assertEqual(window.local_threads.minimum(), 0)
+            self.assertEqual(window.local_threads.maximum(), 1)

tests/test_worker.py

@@ -35,7 +35,7 @@ class Chain(DikteTest):
                   cleaned="Book it for Thursday.",
                   cleanup_error=None, answer=("Booked.", ""), rms=None,
                   clipboard=b"what was there before", paste_error=None,
-                  focus=None):
+                  detected="en", focus=None):
         pipeline = worker.Pipeline(self.conf)
         done, failures, stages, cancels = [], [], [], []
         pipeline.finished.connect(lambda *args: done.append(args))
@@ -45,14 +45,19 @@ class Chain(DikteTest):
 
         cleanup = (mock.Mock(side_effect=cleanup_error) if cleanup_error
                    else mock.Mock(return_value=cleaned))
+        # Auto mode takes the detection path; a fixed language the plain one.
+        # Both are mocked so the chain runs either way without a server.
+        behavior = {"side_effect": transcribe_error} if transcribe_error \
+            else {"return_value": transcript}
+        detect_behavior = {"side_effect": transcribe_error} if transcribe_error \
+            else {"return_value": (transcript, detected)}
         calls = {}
         # The chain reports its own failures on stderr, which a test run has no
         # use for.
         with contextlib.redirect_stderr(io.StringIO()), \
-                mock.patch.object(
-                    api, "transcribe",
-                    **({"side_effect": transcribe_error} if transcribe_error
-                       else {"return_value": transcript})) as tr, \
+                mock.patch.object(api, "transcribe", **behavior) as tr, \
+                mock.patch.object(api, "transcribe_detected",
+                                  **detect_behavior) as tdet, \
                 mock.patch.object(api, "cleanup", cleanup), \
                 mock.patch.object(assistant, "ask", return_value=answer) as ask_call, \
                 mock.patch.object(paste, "copy") as copy, \
@@ -62,7 +67,8 @@ class Chain(DikteTest):
                                   return_value=clipboard) as read_clipboard, \
                 mock.patch.object(worker.time, "sleep", lambda seconds: None):
             press.side_effect = paste_error
-            calls = {"transcribe": tr, "cleanup": cleanup, "ask": ask_call,
+            calls = {"transcribe": tr, "transcribe_detected": tdet,
+                     "cleanup": cleanup, "ask": ask_call,
                      "copy": copy, "copy_bytes": copy_bytes, "press": press,
                      "read_clipboard": read_clipboard}
             pipeline._work(self.wav, duration,
@@ -77,7 +83,8 @@ class Chain(DikteTest):
         run = self.run_chain()
         self.assertEqual(run["failures"], [])
         self.assertEqual(run["done"][0],
-                         ("uh, book it for Thursday", "Book it for Thursday.", ""))
+                         ("uh, book it for Thursday", "Book it for Thursday.",
+                          "", "en"))
         run["copy"].assert_called_once_with("Book it for Thursday.")
         run["press"].assert_called_once_with(self.conf["paste_shortcut"],
                                              focus=None)
@@ -129,7 +136,7 @@ class Chain(DikteTest):
         self.conf["restore_clipboard"] = True
         run = self.run_chain(paste_error=paste.PasteError("not trusted"))
         self.assertEqual(run["failures"], [])
-        raw, text, warning = run["done"][0]
+        raw, text, warning, _lang = run["done"][0]
         self.assertIn("not trusted", warning)
         run["copy_bytes"].assert_not_called()
 
@@ -182,17 +189,41 @@ class Chain(DikteTest):
         self.assertEqual(run["transcribe"].call_args.kwargs["language"], "tr")
         self.assertEqual(run["transcribe"].call_args.kwargs["prompt"], "Paraşüt")
 
+    def test_auto_mode_asks_for_the_detected_language_and_records_it(self):
+        run = self.run_chain(detected="tr")
+        told = run["transcribe_detected"].call_args.kwargs
+        self.assertEqual(told["language"], "auto")
+        self.assertEqual(cfg.read_history()[0]["speech_language"], "tr")
+        self.assertEqual(run["done"][0][3], "tr")
+        run["transcribe"].assert_not_called()
+
+    def test_the_detected_language_is_told_to_the_cleanup_prompt(self):
+        # The mock stands in for api.cleanup, which the cleanup module calls
+        # with (text, key, model, system_prompt, …); the prompt is the fourth.
+        self.conf["transcribe_prompt"] = "Paraşüt"
+        run = self.run_chain(detected="tr")
+        prompt = run["cleanup"].call_args.args[3]
+        # Turkish was detected, so the Turkish glossary rule is appended.
+        self.assertIn("KONUŞMACININ KULLANDIĞI İSİM VE TERİMLER", prompt)
+
+    def test_a_fixed_language_needs_no_detection(self):
+        self.conf["language"] = "en"
+        run = self.run_chain()
+        run["transcribe"].assert_called_once()
+        run["transcribe_detected"].assert_not_called()
+        self.assertEqual(cfg.read_history()[0]["speech_language"], "en")
+
     # ---- silence and stock phrases ----------------------------------------
 
     def test_room_tone_costs_no_api_call(self):
         run = self.run_chain(rms=[0.00001] * 60)
-        run["transcribe"].assert_not_called()
+        run["transcribe_detected"].assert_not_called()
         self.assertIn("No speech", run["failures"][0])
 
     def test_the_silence_check_can_be_switched_off(self):
         self.conf["skip_silent"] = False
         run = self.run_chain(rms=[0.00001] * 60)
-        run["transcribe"].assert_called_once()
+        run["transcribe_detected"].assert_called_once()
 
     def test_a_stock_phrase_from_a_short_clip_is_thrown_away(self):
         run = self.run_chain(duration=2.0, transcript="Altyazı M.K.")
@@ -208,7 +239,7 @@ class Chain(DikteTest):
 
     def test_a_failed_cleanup_still_pastes_the_transcript(self):
         run = self.run_chain(cleanup_error=api.ApiError("rate limited"))
-        _raw, text, warning = run["done"][0]
+        _raw, text, warning, _lang = run["done"][0]
         self.assertEqual(text, "uh, book it for Thursday")
         self.assertIn("rate limited", warning)
         run["copy"].assert_called_once_with("uh, book it for Thursday")
@@ -220,6 +251,9 @@ class Chain(DikteTest):
         self.assertEqual(cfg.read_history()[0]["cleanup_error"], "bad key")
 
     def test_a_failed_transcription_ends_the_run(self):
+        # This path mocks api.transcribe, so it wants
+        # the plain (fixed-language) transcription.
+        self.conf["language"] = "tr"
         pipeline = worker.Pipeline(self.conf)
         failures = []
         pipeline.failed.connect(failures.append)
@@ -231,6 +265,9 @@ class Chain(DikteTest):
         copy.assert_not_called()
 
     def test_a_clipboard_that_will_not_take_it(self):
+        # This path mocks api.transcribe, so it wants
+        # the plain (fixed-language) transcription.
+        self.conf["language"] = "tr"
         pipeline = worker.Pipeline(self.conf)
         failures = []
         pipeline.failed.connect(failures.append)
@@ -243,6 +280,9 @@ class Chain(DikteTest):
         self.assertIn("wl-copy", failures[0])
 
     def test_an_unexpected_error_is_reported_rather_than_swallowed(self):
+        # This path mocks api.transcribe, so it wants
+        # the plain (fixed-language) transcription.
+        self.conf["language"] = "tr"
         pipeline = worker.Pipeline(self.conf)
         failures = []
         pipeline.failed.connect(failures.append)
@@ -280,6 +320,9 @@ class Chain(DikteTest):
         run["press"].assert_not_called()
 
     def test_a_command_that_was_cancelled(self):
+        # This path mocks api.transcribe, so it wants
+        # the plain (fixed-language) transcription.
+        self.conf["language"] = "tr"
         pipeline = worker.Pipeline(self.conf)
         cancels = []
         pipeline.cancelled.connect(lambda: cancels.append(True))
@@ -289,6 +332,9 @@ class Chain(DikteTest):
         self.assertEqual(cancels, [True])
 
     def test_an_agent_that_is_not_installed(self):
+        # This path mocks api.transcribe, so it wants
+        # the plain (fixed-language) transcription.
+        self.conf["language"] = "tr"
         pipeline = worker.Pipeline(self.conf)
         failures = []
         pipeline.failed.connect(failures.append)