feat(transcribe): add speaker diarization support

.github/scripts/sync_extras.py

@@ -37,6 +37,7 @@
     "rag": ("RAG proxy (ChromaDB, embeddings)", ["chromadb", "pydantic_ai"]),
     "memory": ("Long-term memory proxy", ["chromadb", "yaml", "pydantic_ai"]),
     "vad": ("Voice Activity Detection (Silero VAD via ONNX)", ["onnxruntime"]),
+    "diarization": ("Speaker diarization (pyannote.audio)", ["pyannote.audio"]),
     "whisper": ("Local Whisper ASR (faster-whisper)", ["faster_whisper"]),
     "whisper-mlx": ("MLX Whisper for Apple Silicon", ["mlx_whisper"]),
     "tts": ("Local Piper TTS", ["piper"]),

.github/workflows/pytest.yml

@@ -33,7 +33,7 @@ jobs:
         run: uv run --all-extras pytest -vvv
       - name: Run pytest (non-macOS - exclude mlx-whisper)
         if: matrix.os != 'macos-latest'
-        run: uv run --extra audio --extra llm --extra rag --extra memory --extra vad --extra faster-whisper --extra piper --extra kokoro --extra server --extra speed --extra test pytest -vvv
+        run: uv run --extra audio --extra diarization --extra llm --extra rag --extra memory --extra vad --extra faster-whisper --extra piper --extra kokoro --extra server --extra speed --extra test pytest -vvv
       - name: Upload coverage reports to Codecov
         if: matrix.os == 'ubuntu-latest' && matrix.python-version == '3.13'
         uses: codecov/codecov-action@v5

README.md

@@ -421,6 +421,7 @@ agent-cli install-extras rag memory vad
  Available extras:
 
   • audio - Audio recording/playback
+  • diarization - Speaker diarization (pyannote.audio)
   • faster-whisper - Whisper ASR via CTranslate2
   • kokoro - Kokoro neural TTS (GPU)
   • llm - LLM framework (pydantic-ai)
@@ -444,9 +445,9 @@ agent-cli install-extras rag memory vad
 
 
 ╭─ Arguments ────────────────────────────────────────────────────────────────────────────╮
-│   extras      [EXTRAS]...  Extras to install: audio, faster-whisper, kokoro, llm,      │
-│                            memory, mlx-whisper, piper, rag, server, speed, vad,        │
-│                            whisper-transformers, wyoming                               │
+│   extras      [EXTRAS]...  Extras to install: audio, diarization, faster-whisper,      │
+│                            kokoro, llm, memory, mlx-whisper, piper, rag, server,       │
+│                            speed, vad, whisper-transformers, wyoming                   │
 ╰────────────────────────────────────────────────────────────────────────────────────────╯
 ╭─ Options ──────────────────────────────────────────────────────────────────────────────╮
 │ --list  -l        Show available extras with descriptions (what each one enables)      │
@@ -730,7 +731,7 @@ the `[defaults]` section of your configuration file.
 │ --llm                   --no-llm          Clean up transcript with LLM: fix errors,    │
 │                                           add punctuation, remove filler words. Uses   │
 │                                           --extra-instructions if set (via CLI or      │
-│                                           config file).                                │
+│                                           config file). Not compatible with --diarize. │
 │                                           [default: no-llm]                            │
 ╰────────────────────────────────────────────────────────────────────────────────────────╯
 ╭─ Audio Recovery ───────────────────────────────────────────────────────────────────────╮
@@ -852,6 +853,44 @@ the `[defaults]` section of your configuration file.
 │                                                                  provide context for   │
 │                                                                  LLM cleanup.          │
 ╰────────────────────────────────────────────────────────────────────────────────────────╯
+╭─ Diarization ──────────────────────────────────────────────────────────────────────────╮
+│ --diarize           --no-diarize                       Enable speaker diarization      │
+│                                                        (requires pyannote-audio).      │
+│                                                        Install with: pip install       │
+│                                                        agent-cli[diarization]          │
+│                                                        [default: no-diarize]           │
+│ --diarize-format                        [inline|json]  Output format for diarization   │
+│                                                        ('inline' for [Speaker N]:      │
+│                                                        text, 'json' for structured     │
+│                                                        output).                        │
+│                                                        [default: inline]               │
+│ --hf-token                              TEXT           HuggingFace token for pyannote  │
+│                                                        models. Required for            │
+│                                                        diarization. Token must have    │
+│                                                        'Read access to contents of all │
+│                                                        public gated repos you can      │
+│                                                        access' permission. Accept      │
+│                                                        licenses at:                    │
+│                                                        https://hf.co/pyannote/speaker… │
+│                                                        https://hf.co/pyannote/segment… │
+│                                                        https://hf.co/pyannote/wespeak… │
+│                                                        [env var: HF_TOKEN]             │
+│ --min-speakers                          INTEGER        Minimum number of speakers      │
+│                                                        (optional hint for              │
+│                                                        diarization).                   │
+│ --max-speakers                          INTEGER        Maximum number of speakers      │
+│                                                        (optional hint for              │
+│                                                        diarization).                   │
+│ --align-words       --no-align-words                   Use wav2vec2 forced alignment   │
+│                                                        for word-level speaker          │
+│                                                        assignment (more accurate but   │
+│                                                        slower).                        │
+│                                                        [default: no-align-words]       │
+│ --align-language                        TEXT           Language code for word          │
+│                                                        alignment model (e.g., 'en',    │
+│                                                        'fr', 'de', 'es', 'it').        │
+│                                                        [default: en]                   │
+╰────────────────────────────────────────────────────────────────────────────────────────╯
 
 ```
 
@@ -1050,7 +1089,7 @@ uv tool install "agent-cli[vad]" -p 3.13
 ╭─ LLM Configuration ────────────────────────────────────────────────────────────────────╮
 │ --llm    --no-llm      Clean up transcript with LLM: fix errors, add punctuation,      │
 │                        remove filler words. Uses --extra-instructions if set (via CLI  │
-│                        or config file).                                                │
+│                        or config file). Not compatible with --diarize.                 │
 │                        [default: no-llm]                                               │
 ╰────────────────────────────────────────────────────────────────────────────────────────╯
 ╭─ Process Management ───────────────────────────────────────────────────────────────────╮

agent_cli/_extras.json

@@ -5,6 +5,12 @@
       "sounddevice"
     ]
   ],
+  "diarization": [
+    "Speaker diarization (pyannote.audio)",
+    [
+      "pyannote.audio"
+    ]
+  ],
   "faster-whisper": [
     "Whisper ASR via CTranslate2",
     [