@getpaseo/server 0.1.2
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/.env.example +20 -0
- package/README.md +107 -0
- package/agent-prompt.md +339 -0
- package/dist/scripts/daemon-runner.js +32 -0
- package/dist/scripts/daemon-runner.js.map +1 -0
- package/dist/scripts/dev-runner.js +19 -0
- package/dist/scripts/dev-runner.js.map +1 -0
- package/dist/scripts/mcp-stdio-socket-bridge-cli.mjs +62 -0
- package/dist/scripts/supervisor.js +95 -0
- package/dist/scripts/supervisor.js.map +1 -0
- package/dist/server/client/daemon-client.d.ts +383 -0
- package/dist/server/client/daemon-client.d.ts.map +1 -0
- package/dist/server/client/daemon-client.js +2443 -0
- package/dist/server/client/daemon-client.js.map +1 -0
- package/dist/server/server/agent/activity-curator.d.ts +8 -0
- package/dist/server/server/agent/activity-curator.d.ts.map +1 -0
- package/dist/server/server/agent/activity-curator.js +228 -0
- package/dist/server/server/agent/activity-curator.js.map +1 -0
- package/dist/server/server/agent/agent-management-mcp.d.ts +34 -0
- package/dist/server/server/agent/agent-management-mcp.d.ts.map +1 -0
- package/dist/server/server/agent/agent-management-mcp.js +619 -0
- package/dist/server/server/agent/agent-management-mcp.js.map +1 -0
- package/dist/server/server/agent/agent-manager.d.ts +182 -0
- package/dist/server/server/agent/agent-manager.d.ts.map +1 -0
- package/dist/server/server/agent/agent-manager.js +1066 -0
- package/dist/server/server/agent/agent-manager.js.map +1 -0
- package/dist/server/server/agent/agent-metadata-generator.d.ts +29 -0
- package/dist/server/server/agent/agent-metadata-generator.d.ts.map +1 -0
- package/dist/server/server/agent/agent-metadata-generator.js +157 -0
- package/dist/server/server/agent/agent-metadata-generator.js.map +1 -0
- package/dist/server/server/agent/agent-projections.d.ts +12 -0
- package/dist/server/server/agent/agent-projections.d.ts.map +1 -0
- package/dist/server/server/agent/agent-projections.js +238 -0
- package/dist/server/server/agent/agent-projections.js.map +1 -0
- package/dist/server/server/agent/agent-response-loop.d.ts +32 -0
- package/dist/server/server/agent/agent-response-loop.d.ts.map +1 -0
- package/dist/server/server/agent/agent-response-loop.js +224 -0
- package/dist/server/server/agent/agent-response-loop.js.map +1 -0
- package/dist/server/server/agent/agent-sdk-types.d.ts +360 -0
- package/dist/server/server/agent/agent-sdk-types.d.ts.map +1 -0
- package/dist/server/server/agent/agent-sdk-types.js +2 -0
- package/dist/server/server/agent/agent-sdk-types.js.map +1 -0
- package/dist/server/server/agent/agent-storage.d.ts +187 -0
- package/dist/server/server/agent/agent-storage.d.ts.map +1 -0
- package/dist/server/server/agent/agent-storage.js +328 -0
- package/dist/server/server/agent/agent-storage.js.map +1 -0
- package/dist/server/server/agent/audio-utils.d.ts +3 -0
- package/dist/server/server/agent/audio-utils.d.ts.map +1 -0
- package/dist/server/server/agent/audio-utils.js +19 -0
- package/dist/server/server/agent/audio-utils.js.map +1 -0
- package/dist/server/server/agent/dictation-debug.d.ts +13 -0
- package/dist/server/server/agent/dictation-debug.d.ts.map +1 -0
- package/dist/server/server/agent/dictation-debug.js +50 -0
- package/dist/server/server/agent/dictation-debug.js.map +1 -0
- package/dist/server/server/agent/llm-openai.d.ts +7 -0
- package/dist/server/server/agent/llm-openai.d.ts.map +1 -0
- package/dist/server/server/agent/llm-openai.js +8 -0
- package/dist/server/server/agent/llm-openai.js.map +1 -0
- package/dist/server/server/agent/mcp-server.d.ts +26 -0
- package/dist/server/server/agent/mcp-server.d.ts.map +1 -0
- package/dist/server/server/agent/mcp-server.js +762 -0
- package/dist/server/server/agent/mcp-server.js.map +1 -0
- package/dist/server/server/agent/model-resolver.d.ts +11 -0
- package/dist/server/server/agent/model-resolver.d.ts.map +1 -0
- package/dist/server/server/agent/model-resolver.js +21 -0
- package/dist/server/server/agent/model-resolver.js.map +1 -0
- package/dist/server/server/agent/orchestrator-instructions.d.ts +7 -0
- package/dist/server/server/agent/orchestrator-instructions.d.ts.map +1 -0
- package/dist/server/server/agent/orchestrator-instructions.js +51 -0
- package/dist/server/server/agent/orchestrator-instructions.js.map +1 -0
- package/dist/server/server/agent/orchestrator.d.ts +12 -0
- package/dist/server/server/agent/orchestrator.d.ts.map +1 -0
- package/dist/server/server/agent/orchestrator.js +12 -0
- package/dist/server/server/agent/orchestrator.js.map +1 -0
- package/dist/server/server/agent/pcm16-resampler.d.ts +14 -0
- package/dist/server/server/agent/pcm16-resampler.d.ts.map +1 -0
- package/dist/server/server/agent/pcm16-resampler.js +63 -0
- package/dist/server/server/agent/pcm16-resampler.js.map +1 -0
- package/dist/server/server/agent/provider-launch-config.d.ts +139 -0
- package/dist/server/server/agent/provider-launch-config.d.ts.map +1 -0
- package/dist/server/server/agent/provider-launch-config.js +83 -0
- package/dist/server/server/agent/provider-launch-config.js.map +1 -0
- package/dist/server/server/agent/provider-manifest.d.ts +20 -0
- package/dist/server/server/agent/provider-manifest.d.ts.map +1 -0
- package/dist/server/server/agent/provider-manifest.js +97 -0
- package/dist/server/server/agent/provider-manifest.js.map +1 -0
- package/dist/server/server/agent/provider-registry.d.ts +18 -0
- package/dist/server/server/agent/provider-registry.d.ts.map +1 -0
- package/dist/server/server/agent/provider-registry.js +45 -0
- package/dist/server/server/agent/provider-registry.js.map +1 -0
- package/dist/server/server/agent/providers/claude/tool-call-detail-parser.d.ts +3 -0
- package/dist/server/server/agent/providers/claude/tool-call-detail-parser.d.ts.map +1 -0
- package/dist/server/server/agent/providers/claude/tool-call-detail-parser.js +42 -0
- package/dist/server/server/agent/providers/claude/tool-call-detail-parser.js.map +1 -0
- package/dist/server/server/agent/providers/claude/tool-call-mapper.d.ts +16 -0
- package/dist/server/server/agent/providers/claude/tool-call-mapper.d.ts.map +1 -0
- package/dist/server/server/agent/providers/claude/tool-call-mapper.js +73 -0
- package/dist/server/server/agent/providers/claude/tool-call-mapper.js.map +1 -0
- package/dist/server/server/agent/providers/claude-agent.d.ts +35 -0
- package/dist/server/server/agent/providers/claude-agent.d.ts.map +1 -0
- package/dist/server/server/agent/providers/claude-agent.js +2056 -0
- package/dist/server/server/agent/providers/claude-agent.js.map +1 -0
- package/dist/server/server/agent/providers/codex/tool-call-detail-parser.d.ts +13 -0
- package/dist/server/server/agent/providers/codex/tool-call-detail-parser.d.ts.map +1 -0
- package/dist/server/server/agent/providers/codex/tool-call-detail-parser.js +67 -0
- package/dist/server/server/agent/providers/codex/tool-call-detail-parser.js.map +1 -0
- package/dist/server/server/agent/providers/codex/tool-call-mapper.d.ts +15 -0
- package/dist/server/server/agent/providers/codex/tool-call-mapper.d.ts.map +1 -0
- package/dist/server/server/agent/providers/codex/tool-call-mapper.js +640 -0
- package/dist/server/server/agent/providers/codex/tool-call-mapper.js.map +1 -0
- package/dist/server/server/agent/providers/codex-app-server-agent.d.ts +34 -0
- package/dist/server/server/agent/providers/codex-app-server-agent.d.ts.map +1 -0
- package/dist/server/server/agent/providers/codex-app-server-agent.js +2476 -0
- package/dist/server/server/agent/providers/codex-app-server-agent.js.map +1 -0
- package/dist/server/server/agent/providers/codex-rollout-timeline.d.ts +9 -0
- package/dist/server/server/agent/providers/codex-rollout-timeline.d.ts.map +1 -0
- package/dist/server/server/agent/providers/codex-rollout-timeline.js +486 -0
- package/dist/server/server/agent/providers/codex-rollout-timeline.js.map +1 -0
- package/dist/server/server/agent/providers/opencode/tool-call-detail-parser.d.ts +3 -0
- package/dist/server/server/agent/providers/opencode/tool-call-detail-parser.d.ts.map +1 -0
- package/dist/server/server/agent/providers/opencode/tool-call-detail-parser.js +33 -0
- package/dist/server/server/agent/providers/opencode/tool-call-detail-parser.js.map +1 -0
- package/dist/server/server/agent/providers/opencode/tool-call-mapper.d.ts +13 -0
- package/dist/server/server/agent/providers/opencode/tool-call-mapper.d.ts.map +1 -0
- package/dist/server/server/agent/providers/opencode/tool-call-mapper.js +75 -0
- package/dist/server/server/agent/providers/opencode/tool-call-mapper.js.map +1 -0
- package/dist/server/server/agent/providers/opencode-agent.d.ts +37 -0
- package/dist/server/server/agent/providers/opencode-agent.d.ts.map +1 -0
- package/dist/server/server/agent/providers/opencode-agent.js +822 -0
- package/dist/server/server/agent/providers/opencode-agent.js.map +1 -0
- package/dist/server/server/agent/providers/tool-call-detail-primitives.d.ts +1363 -0
- package/dist/server/server/agent/providers/tool-call-detail-primitives.d.ts.map +1 -0
- package/dist/server/server/agent/providers/tool-call-detail-primitives.js +534 -0
- package/dist/server/server/agent/providers/tool-call-detail-primitives.js.map +1 -0
- package/dist/server/server/agent/providers/tool-call-mapper-utils.d.ts +18 -0
- package/dist/server/server/agent/providers/tool-call-mapper-utils.d.ts.map +1 -0
- package/dist/server/server/agent/providers/tool-call-mapper-utils.js +119 -0
- package/dist/server/server/agent/providers/tool-call-mapper-utils.js.map +1 -0
- package/dist/server/server/agent/recordings-debug.d.ts +3 -0
- package/dist/server/server/agent/recordings-debug.d.ts.map +1 -0
- package/dist/server/server/agent/recordings-debug.js +19 -0
- package/dist/server/server/agent/recordings-debug.js.map +1 -0
- package/dist/server/server/agent/stt-debug.d.ts +10 -0
- package/dist/server/server/agent/stt-debug.d.ts.map +1 -0
- package/dist/server/server/agent/stt-debug.js +33 -0
- package/dist/server/server/agent/stt-debug.js.map +1 -0
- package/dist/server/server/agent/stt-manager.d.ts +32 -0
- package/dist/server/server/agent/stt-manager.d.ts.map +1 -0
- package/dist/server/server/agent/stt-manager.js +231 -0
- package/dist/server/server/agent/stt-manager.js.map +1 -0
- package/dist/server/server/agent/system-prompt.d.ts +3 -0
- package/dist/server/server/agent/system-prompt.d.ts.map +1 -0
- package/dist/server/server/agent/system-prompt.js +19 -0
- package/dist/server/server/agent/system-prompt.js.map +1 -0
- package/dist/server/server/agent/tool-name-normalization.d.ts +7 -0
- package/dist/server/server/agent/tool-name-normalization.d.ts.map +1 -0
- package/dist/server/server/agent/tool-name-normalization.js +45 -0
- package/dist/server/server/agent/tool-name-normalization.js.map +1 -0
- package/dist/server/server/agent/tts-debug.d.ts +8 -0
- package/dist/server/server/agent/tts-debug.d.ts.map +1 -0
- package/dist/server/server/agent/tts-debug.js +24 -0
- package/dist/server/server/agent/tts-debug.js.map +1 -0
- package/dist/server/server/agent/tts-manager.d.ts +33 -0
- package/dist/server/server/agent/tts-manager.d.ts.map +1 -0
- package/dist/server/server/agent/tts-manager.js +261 -0
- package/dist/server/server/agent/tts-manager.js.map +1 -0
- package/dist/server/server/agent/wait-for-agent-tracker.d.ts +15 -0
- package/dist/server/server/agent/wait-for-agent-tracker.d.ts.map +1 -0
- package/dist/server/server/agent/wait-for-agent-tracker.js +53 -0
- package/dist/server/server/agent/wait-for-agent-tracker.js.map +1 -0
- package/dist/server/server/allowed-hosts.d.ts +13 -0
- package/dist/server/server/allowed-hosts.d.ts.map +1 -0
- package/dist/server/server/allowed-hosts.js +94 -0
- package/dist/server/server/allowed-hosts.js.map +1 -0
- package/dist/server/server/bootstrap.d.ts +49 -0
- package/dist/server/server/bootstrap.d.ts.map +1 -0
- package/dist/server/server/bootstrap.js +483 -0
- package/dist/server/server/bootstrap.js.map +1 -0
- package/dist/server/server/config.d.ts +13 -0
- package/dist/server/server/config.d.ts.map +1 -0
- package/dist/server/server/config.js +84 -0
- package/dist/server/server/config.js.map +1 -0
- package/dist/server/server/connection-offer.d.ts +19 -0
- package/dist/server/server/connection-offer.d.ts.map +1 -0
- package/dist/server/server/connection-offer.js +60 -0
- package/dist/server/server/connection-offer.js.map +1 -0
- package/dist/server/server/daemon-keypair.d.ts +8 -0
- package/dist/server/server/daemon-keypair.d.ts.map +1 -0
- package/dist/server/server/daemon-keypair.js +40 -0
- package/dist/server/server/daemon-keypair.js.map +1 -0
- package/dist/server/server/dictation/dictation-stream-manager.d.ts +76 -0
- package/dist/server/server/dictation/dictation-stream-manager.d.ts.map +1 -0
- package/dist/server/server/dictation/dictation-stream-manager.js +481 -0
- package/dist/server/server/dictation/dictation-stream-manager.js.map +1 -0
- package/dist/server/server/exports.d.ts +11 -0
- package/dist/server/server/exports.d.ts.map +1 -0
- package/dist/server/server/exports.js +11 -0
- package/dist/server/server/exports.js.map +1 -0
- package/dist/server/server/file-download/token-store.d.ts +25 -0
- package/dist/server/server/file-download/token-store.d.ts.map +1 -0
- package/dist/server/server/file-download/token-store.js +40 -0
- package/dist/server/server/file-download/token-store.js.map +1 -0
- package/dist/server/server/file-explorer/service.d.ts +41 -0
- package/dist/server/server/file-explorer/service.d.ts.map +1 -0
- package/dist/server/server/file-explorer/service.js +163 -0
- package/dist/server/server/file-explorer/service.js.map +1 -0
- package/dist/server/server/index.d.ts +2 -0
- package/dist/server/server/index.d.ts.map +1 -0
- package/dist/server/server/index.js +90 -0
- package/dist/server/server/index.js.map +1 -0
- package/dist/server/server/json-utils.d.ts +11 -0
- package/dist/server/server/json-utils.d.ts.map +1 -0
- package/dist/server/server/json-utils.js +45 -0
- package/dist/server/server/json-utils.js.map +1 -0
- package/dist/server/server/logger.d.ts +12 -0
- package/dist/server/server/logger.d.ts.map +1 -0
- package/dist/server/server/logger.js +29 -0
- package/dist/server/server/logger.js.map +1 -0
- package/dist/server/server/messages.d.ts +9 -0
- package/dist/server/server/messages.d.ts.map +1 -0
- package/dist/server/server/messages.js +29 -0
- package/dist/server/server/messages.js.map +1 -0
- package/dist/server/server/pairing-offer.d.ts +16 -0
- package/dist/server/server/pairing-offer.d.ts.map +1 -0
- package/dist/server/server/pairing-offer.js +45 -0
- package/dist/server/server/pairing-offer.js.map +1 -0
- package/dist/server/server/pairing-qr.d.ts +7 -0
- package/dist/server/server/pairing-qr.d.ts.map +1 -0
- package/dist/server/server/pairing-qr.js +45 -0
- package/dist/server/server/pairing-qr.js.map +1 -0
- package/dist/server/server/paseo-home.d.ts +2 -0
- package/dist/server/server/paseo-home.d.ts.map +1 -0
- package/dist/server/server/paseo-home.js +19 -0
- package/dist/server/server/paseo-home.js.map +1 -0
- package/dist/server/server/path-utils.d.ts +3 -0
- package/dist/server/server/path-utils.d.ts.map +1 -0
- package/dist/server/server/path-utils.js +20 -0
- package/dist/server/server/path-utils.js.map +1 -0
- package/dist/server/server/persisted-config.d.ts +500 -0
- package/dist/server/server/persisted-config.d.ts.map +1 -0
- package/dist/server/server/persisted-config.js +212 -0
- package/dist/server/server/persisted-config.js.map +1 -0
- package/dist/server/server/persistence-hooks.d.ts +24 -0
- package/dist/server/server/persistence-hooks.d.ts.map +1 -0
- package/dist/server/server/persistence-hooks.js +60 -0
- package/dist/server/server/persistence-hooks.js.map +1 -0
- package/dist/server/server/pid-lock.d.ts +19 -0
- package/dist/server/server/pid-lock.d.ts.map +1 -0
- package/dist/server/server/pid-lock.js +115 -0
- package/dist/server/server/pid-lock.js.map +1 -0
- package/dist/server/server/push/push-service.d.ts +21 -0
- package/dist/server/server/push/push-service.d.ts.map +1 -0
- package/dist/server/server/push/push-service.js +68 -0
- package/dist/server/server/push/push-service.js.map +1 -0
- package/dist/server/server/push/token-store.d.ts +18 -0
- package/dist/server/server/push/token-store.d.ts.map +1 -0
- package/dist/server/server/push/token-store.js +70 -0
- package/dist/server/server/push/token-store.js.map +1 -0
- package/dist/server/server/relay-transport.d.ts +22 -0
- package/dist/server/server/relay-transport.d.ts.map +1 -0
- package/dist/server/server/relay-transport.js +374 -0
- package/dist/server/server/relay-transport.js.map +1 -0
- package/dist/server/server/server-id.d.ts +17 -0
- package/dist/server/server/server-id.d.ts.map +1 -0
- package/dist/server/server/server-id.js +63 -0
- package/dist/server/server/server-id.js.map +1 -0
- package/dist/server/server/session.d.ts +360 -0
- package/dist/server/server/session.d.ts.map +1 -0
- package/dist/server/server/session.js +4615 -0
- package/dist/server/server/session.js.map +1 -0
- package/dist/server/server/speech/audio.d.ts +10 -0
- package/dist/server/server/speech/audio.d.ts.map +1 -0
- package/dist/server/server/speech/audio.js +101 -0
- package/dist/server/server/speech/audio.js.map +1 -0
- package/dist/server/server/speech/providers/local/config.d.ts +26 -0
- package/dist/server/server/speech/providers/local/config.d.ts.map +1 -0
- package/dist/server/server/speech/providers/local/config.js +93 -0
- package/dist/server/server/speech/providers/local/config.js.map +1 -0
- package/dist/server/server/speech/providers/local/models.d.ts +12 -0
- package/dist/server/server/speech/providers/local/models.d.ts.map +1 -0
- package/dist/server/server/speech/providers/local/models.js +18 -0
- package/dist/server/server/speech/providers/local/models.js.map +1 -0
- package/dist/server/server/speech/providers/local/pocket/pocket-tts-onnx.d.ts +24 -0
- package/dist/server/server/speech/providers/local/pocket/pocket-tts-onnx.d.ts.map +1 -0
- package/dist/server/server/speech/providers/local/pocket/pocket-tts-onnx.js +422 -0
- package/dist/server/server/speech/providers/local/pocket/pocket-tts-onnx.js.map +1 -0
- package/dist/server/server/speech/providers/local/runtime.d.ts +30 -0
- package/dist/server/server/speech/providers/local/runtime.d.ts.map +1 -0
- package/dist/server/server/speech/providers/local/runtime.js +254 -0
- package/dist/server/server/speech/providers/local/runtime.js.map +1 -0
- package/dist/server/server/speech/providers/local/sherpa/model-catalog.d.ts +117 -0
- package/dist/server/server/speech/providers/local/sherpa/model-catalog.d.ts.map +1 -0
- package/dist/server/server/speech/providers/local/sherpa/model-catalog.js +166 -0
- package/dist/server/server/speech/providers/local/sherpa/model-catalog.js.map +1 -0
- package/dist/server/server/speech/providers/local/sherpa/model-downloader.d.ts +17 -0
- package/dist/server/server/speech/providers/local/sherpa/model-downloader.d.ts.map +1 -0
- package/dist/server/server/speech/providers/local/sherpa/model-downloader.js +151 -0
- package/dist/server/server/speech/providers/local/sherpa/model-downloader.js.map +1 -0
- package/dist/server/server/speech/providers/local/sherpa/sherpa-offline-recognizer.d.ts +28 -0
- package/dist/server/server/speech/providers/local/sherpa/sherpa-offline-recognizer.d.ts.map +1 -0
- package/dist/server/server/speech/providers/local/sherpa/sherpa-offline-recognizer.js +68 -0
- package/dist/server/server/speech/providers/local/sherpa/sherpa-offline-recognizer.js.map +1 -0
- package/dist/server/server/speech/providers/local/sherpa/sherpa-online-recognizer.d.ts +37 -0
- package/dist/server/server/speech/providers/local/sherpa/sherpa-online-recognizer.d.ts.map +1 -0
- package/dist/server/server/speech/providers/local/sherpa/sherpa-online-recognizer.js +79 -0
- package/dist/server/server/speech/providers/local/sherpa/sherpa-online-recognizer.js.map +1 -0
- package/dist/server/server/speech/providers/local/sherpa/sherpa-onnx-loader.d.ts +7 -0
- package/dist/server/server/speech/providers/local/sherpa/sherpa-onnx-loader.d.ts.map +1 -0
- package/dist/server/server/speech/providers/local/sherpa/sherpa-onnx-loader.js +11 -0
- package/dist/server/server/speech/providers/local/sherpa/sherpa-onnx-loader.js.map +1 -0
- package/dist/server/server/speech/providers/local/sherpa/sherpa-onnx-node-loader.d.ts +7 -0
- package/dist/server/server/speech/providers/local/sherpa/sherpa-onnx-node-loader.d.ts.map +1 -0
- package/dist/server/server/speech/providers/local/sherpa/sherpa-onnx-node-loader.js +44 -0
- package/dist/server/server/speech/providers/local/sherpa/sherpa-onnx-node-loader.js.map +1 -0
- package/dist/server/server/speech/providers/local/sherpa/sherpa-parakeet-realtime-session.d.ts +28 -0
- package/dist/server/server/speech/providers/local/sherpa/sherpa-parakeet-realtime-session.d.ts.map +1 -0
- package/dist/server/server/speech/providers/local/sherpa/sherpa-parakeet-realtime-session.js +131 -0
- package/dist/server/server/speech/providers/local/sherpa/sherpa-parakeet-realtime-session.js.map +1 -0
- package/dist/server/server/speech/providers/local/sherpa/sherpa-parakeet-stt.d.ts +21 -0
- package/dist/server/server/speech/providers/local/sherpa/sherpa-parakeet-stt.d.ts.map +1 -0
- package/dist/server/server/speech/providers/local/sherpa/sherpa-parakeet-stt.js +132 -0
- package/dist/server/server/speech/providers/local/sherpa/sherpa-parakeet-stt.js.map +1 -0
- package/dist/server/server/speech/providers/local/sherpa/sherpa-realtime-session.d.ts +23 -0
- package/dist/server/server/speech/providers/local/sherpa/sherpa-realtime-session.d.ts.map +1 -0
- package/dist/server/server/speech/providers/local/sherpa/sherpa-realtime-session.js +112 -0
- package/dist/server/server/speech/providers/local/sherpa/sherpa-realtime-session.js.map +1 -0
- package/dist/server/server/speech/providers/local/sherpa/sherpa-stt.d.ts +23 -0
- package/dist/server/server/speech/providers/local/sherpa/sherpa-stt.d.ts.map +1 -0
- package/dist/server/server/speech/providers/local/sherpa/sherpa-stt.js +140 -0
- package/dist/server/server/speech/providers/local/sherpa/sherpa-stt.js.map +1 -0
- package/dist/server/server/speech/providers/local/sherpa/sherpa-tts.d.ts +21 -0
- package/dist/server/server/speech/providers/local/sherpa/sherpa-tts.d.ts.map +1 -0
- package/dist/server/server/speech/providers/local/sherpa/sherpa-tts.js +95 -0
- package/dist/server/server/speech/providers/local/sherpa/sherpa-tts.js.map +1 -0
- package/dist/server/server/speech/providers/openai/config.d.ts +22 -0
- package/dist/server/server/speech/providers/openai/config.d.ts.map +1 -0
- package/dist/server/server/speech/providers/openai/config.js +94 -0
- package/dist/server/server/speech/providers/openai/config.js.map +1 -0
- package/dist/server/server/speech/providers/openai/realtime-transcription-session.d.ts +42 -0
- package/dist/server/server/speech/providers/openai/realtime-transcription-session.d.ts.map +1 -0
- package/dist/server/server/speech/providers/openai/realtime-transcription-session.js +165 -0
- package/dist/server/server/speech/providers/openai/realtime-transcription-session.js.map +1 -0
- package/dist/server/server/speech/providers/openai/runtime.d.ts +27 -0
- package/dist/server/server/speech/providers/openai/runtime.d.ts.map +1 -0
- package/dist/server/server/speech/providers/openai/runtime.js +103 -0
- package/dist/server/server/speech/providers/openai/runtime.js.map +1 -0
- package/dist/server/server/speech/providers/openai/stt.d.ts +22 -0
- package/dist/server/server/speech/providers/openai/stt.d.ts.map +1 -0
- package/dist/server/server/speech/providers/openai/stt.js +208 -0
- package/dist/server/server/speech/providers/openai/stt.js.map +1 -0
- package/dist/server/server/speech/providers/openai/tts.d.ts +18 -0
- package/dist/server/server/speech/providers/openai/tts.d.ts.map +1 -0
- package/dist/server/server/speech/providers/openai/tts.js +46 -0
- package/dist/server/server/speech/providers/openai/tts.js.map +1 -0
- package/dist/server/server/speech/speech-config-resolver.d.ts +11 -0
- package/dist/server/server/speech/speech-config-resolver.d.ts.map +1 -0
- package/dist/server/server/speech/speech-config-resolver.js +64 -0
- package/dist/server/server/speech/speech-config-resolver.js.map +1 -0
- package/dist/server/server/speech/speech-provider.d.ts +59 -0
- package/dist/server/server/speech/speech-provider.d.ts.map +1 -0
- package/dist/server/server/speech/speech-provider.js +2 -0
- package/dist/server/server/speech/speech-provider.js.map +1 -0
- package/dist/server/server/speech/speech-runtime.d.ts +20 -0
- package/dist/server/server/speech/speech-runtime.d.ts.map +1 -0
- package/dist/server/server/speech/speech-runtime.js +119 -0
- package/dist/server/server/speech/speech-runtime.js.map +1 -0
- package/dist/server/server/speech/speech-types.d.ts +20 -0
- package/dist/server/server/speech/speech-types.d.ts.map +1 -0
- package/dist/server/server/speech/speech-types.js +7 -0
- package/dist/server/server/speech/speech-types.js.map +1 -0
- package/dist/server/server/terminal-mcp/index.d.ts +4 -0
- package/dist/server/server/terminal-mcp/index.d.ts.map +1 -0
- package/dist/server/server/terminal-mcp/index.js +3 -0
- package/dist/server/server/terminal-mcp/index.js.map +1 -0
- package/dist/server/server/terminal-mcp/server.d.ts +10 -0
- package/dist/server/server/terminal-mcp/server.d.ts.map +1 -0
- package/dist/server/server/terminal-mcp/server.js +217 -0
- package/dist/server/server/terminal-mcp/server.js.map +1 -0
- package/dist/server/server/terminal-mcp/terminal-manager.d.ts +123 -0
- package/dist/server/server/terminal-mcp/terminal-manager.d.ts.map +1 -0
- package/dist/server/server/terminal-mcp/terminal-manager.js +351 -0
- package/dist/server/server/terminal-mcp/terminal-manager.js.map +1 -0
- package/dist/server/server/terminal-mcp/tmux.d.ts +207 -0
- package/dist/server/server/terminal-mcp/tmux.d.ts.map +1 -0
- package/dist/server/server/terminal-mcp/tmux.js +924 -0
- package/dist/server/server/terminal-mcp/tmux.js.map +1 -0
- package/dist/server/server/types.d.ts +5 -0
- package/dist/server/server/types.d.ts.map +1 -0
- package/dist/server/server/types.js +3 -0
- package/dist/server/server/types.js.map +1 -0
- package/dist/server/server/utils/diff-highlighter.d.ts +52 -0
- package/dist/server/server/utils/diff-highlighter.d.ts.map +1 -0
- package/dist/server/server/utils/diff-highlighter.js +244 -0
- package/dist/server/server/utils/diff-highlighter.js.map +1 -0
- package/dist/server/server/utils/syntax-highlighter.d.ts +10 -0
- package/dist/server/server/utils/syntax-highlighter.d.ts.map +1 -0
- package/dist/server/server/utils/syntax-highlighter.js +141 -0
- package/dist/server/server/utils/syntax-highlighter.js.map +1 -0
- package/dist/server/server/voice-config.d.ts +14 -0
- package/dist/server/server/voice-config.d.ts.map +1 -0
- package/dist/server/server/voice-config.js +51 -0
- package/dist/server/server/voice-config.js.map +1 -0
- package/dist/server/server/voice-mcp-bridge-command.d.ts +17 -0
- package/dist/server/server/voice-mcp-bridge-command.d.ts.map +1 -0
- package/dist/server/server/voice-mcp-bridge-command.js +31 -0
- package/dist/server/server/voice-mcp-bridge-command.js.map +1 -0
- package/dist/server/server/voice-mcp-bridge.d.ts +18 -0
- package/dist/server/server/voice-mcp-bridge.d.ts.map +1 -0
- package/dist/server/server/voice-mcp-bridge.js +109 -0
- package/dist/server/server/voice-mcp-bridge.js.map +1 -0
- package/dist/server/server/voice-permission-policy.d.ts +4 -0
- package/dist/server/server/voice-permission-policy.d.ts.map +1 -0
- package/dist/server/server/voice-permission-policy.js +13 -0
- package/dist/server/server/voice-permission-policy.js.map +1 -0
- package/dist/server/server/voice-types.d.ts +17 -0
- package/dist/server/server/voice-types.d.ts.map +1 -0
- package/dist/server/server/voice-types.js +2 -0
- package/dist/server/server/voice-types.js.map +1 -0
- package/dist/server/server/websocket-server.d.ts +80 -0
- package/dist/server/server/websocket-server.d.ts.map +1 -0
- package/dist/server/server/websocket-server.js +447 -0
- package/dist/server/server/websocket-server.js.map +1 -0
- package/dist/server/shared/agent-lifecycle.d.ts +3 -0
- package/dist/server/shared/agent-lifecycle.d.ts.map +1 -0
- package/dist/server/shared/agent-lifecycle.js +8 -0
- package/dist/server/shared/agent-lifecycle.js.map +1 -0
- package/dist/server/shared/connection-offer.d.ts +62 -0
- package/dist/server/shared/connection-offer.d.ts.map +1 -0
- package/dist/server/shared/connection-offer.js +17 -0
- package/dist/server/shared/connection-offer.js.map +1 -0
- package/dist/server/shared/daemon-endpoints.d.ts +19 -0
- package/dist/server/shared/daemon-endpoints.d.ts.map +1 -0
- package/dist/server/shared/daemon-endpoints.js +98 -0
- package/dist/server/shared/daemon-endpoints.js.map +1 -0
- package/dist/server/shared/messages.d.ts +36729 -0
- package/dist/server/shared/messages.d.ts.map +1 -0
- package/dist/server/shared/messages.js +1666 -0
- package/dist/server/shared/messages.js.map +1 -0
- package/dist/server/shared/path-utils.d.ts +2 -0
- package/dist/server/shared/path-utils.d.ts.map +1 -0
- package/dist/server/shared/path-utils.js +16 -0
- package/dist/server/shared/path-utils.js.map +1 -0
- package/dist/server/shared/tool-call-display.d.ts +11 -0
- package/dist/server/shared/tool-call-display.d.ts.map +1 -0
- package/dist/server/shared/tool-call-display.js +82 -0
- package/dist/server/shared/tool-call-display.js.map +1 -0
- package/dist/server/terminal/terminal-manager.d.ts +14 -0
- package/dist/server/terminal/terminal-manager.d.ts.map +1 -0
- package/dist/server/terminal/terminal-manager.js +67 -0
- package/dist/server/terminal/terminal-manager.js.map +1 -0
- package/dist/server/terminal/terminal.d.ts +67 -0
- package/dist/server/terminal/terminal.d.ts.map +1 -0
- package/dist/server/terminal/terminal.js +190 -0
- package/dist/server/terminal/terminal.js.map +1 -0
- package/dist/server/utils/checkout-git.d.ts +138 -0
- package/dist/server/utils/checkout-git.d.ts.map +1 -0
- package/dist/server/utils/checkout-git.js +1079 -0
- package/dist/server/utils/checkout-git.js.map +1 -0
- package/dist/server/utils/path.d.ts +5 -0
- package/dist/server/utils/path.d.ts.map +1 -0
- package/dist/server/utils/path.js +15 -0
- package/dist/server/utils/path.js.map +1 -0
- package/dist/server/utils/project-icon.d.ts +39 -0
- package/dist/server/utils/project-icon.d.ts.map +1 -0
- package/dist/server/utils/project-icon.js +391 -0
- package/dist/server/utils/project-icon.js.map +1 -0
- package/dist/server/utils/worktree-metadata.d.ts +21 -0
- package/dist/server/utils/worktree-metadata.d.ts.map +1 -0
- package/dist/server/utils/worktree-metadata.js +74 -0
- package/dist/server/utils/worktree-metadata.js.map +1 -0
- package/dist/server/utils/worktree.d.ts +95 -0
- package/dist/server/utils/worktree.d.ts.map +1 -0
- package/dist/server/utils/worktree.js +568 -0
- package/dist/server/utils/worktree.js.map +1 -0
- package/package.json +108 -0
|
@@ -0,0 +1,131 @@
|
|
|
1
|
+
import { EventEmitter } from "node:events";
|
|
2
|
+
import { v4 as uuidv4 } from "uuid";
|
|
3
|
+
import { pcm16lePeakAbs, pcm16leToFloat32 } from "../../../audio.js";
|
|
4
|
+
export class SherpaParakeetRealtimeTranscriptionSession extends EventEmitter {
|
|
5
|
+
constructor(params) {
|
|
6
|
+
super();
|
|
7
|
+
this.connected = false;
|
|
8
|
+
this.currentSegmentId = null;
|
|
9
|
+
this.previousSegmentId = null;
|
|
10
|
+
this.lastPartialText = "";
|
|
11
|
+
this.pcm16 = Buffer.alloc(0);
|
|
12
|
+
this.lastDecodeAt = 0;
|
|
13
|
+
this.decoding = false;
|
|
14
|
+
this.pendingDecode = false;
|
|
15
|
+
this.engine = params.engine;
|
|
16
|
+
this.requiredSampleRate = this.engine.sampleRate;
|
|
17
|
+
this.minDecodeIntervalMs = params.minDecodeIntervalMs ?? 350;
|
|
18
|
+
}
|
|
19
|
+
async connect() {
|
|
20
|
+
if (this.connected) {
|
|
21
|
+
return;
|
|
22
|
+
}
|
|
23
|
+
this.currentSegmentId = uuidv4();
|
|
24
|
+
this.connected = true;
|
|
25
|
+
}
|
|
26
|
+
appendPcm16(chunk) {
|
|
27
|
+
if (!this.connected || !this.currentSegmentId) {
|
|
28
|
+
this.emit("error", new Error("Parakeet realtime session not connected"));
|
|
29
|
+
return;
|
|
30
|
+
}
|
|
31
|
+
try {
|
|
32
|
+
this.pcm16 = this.pcm16.length === 0 ? chunk : Buffer.concat([this.pcm16, chunk]);
|
|
33
|
+
void this.maybeDecode(false);
|
|
34
|
+
}
|
|
35
|
+
catch (err) {
|
|
36
|
+
this.emit("error", err instanceof Error ? err : new Error(String(err)));
|
|
37
|
+
}
|
|
38
|
+
}
|
|
39
|
+
commit() {
|
|
40
|
+
if (!this.connected || !this.currentSegmentId) {
|
|
41
|
+
this.emit("error", new Error("Parakeet realtime session not connected"));
|
|
42
|
+
return;
|
|
43
|
+
}
|
|
44
|
+
void (async () => {
|
|
45
|
+
try {
|
|
46
|
+
await this.maybeDecode(true);
|
|
47
|
+
const finalText = this.lastPartialText;
|
|
48
|
+
const segmentId = this.currentSegmentId;
|
|
49
|
+
const previousSegmentId = this.previousSegmentId;
|
|
50
|
+
this.emit("committed", { segmentId, previousSegmentId });
|
|
51
|
+
this.emit("transcript", { segmentId, transcript: finalText, isFinal: true });
|
|
52
|
+
this.previousSegmentId = segmentId;
|
|
53
|
+
this.currentSegmentId = uuidv4();
|
|
54
|
+
this.lastPartialText = "";
|
|
55
|
+
this.pcm16 = Buffer.alloc(0);
|
|
56
|
+
}
|
|
57
|
+
catch (err) {
|
|
58
|
+
this.emit("error", err instanceof Error ? err : new Error(String(err)));
|
|
59
|
+
}
|
|
60
|
+
})();
|
|
61
|
+
}
|
|
62
|
+
clear() {
|
|
63
|
+
if (!this.connected) {
|
|
64
|
+
return;
|
|
65
|
+
}
|
|
66
|
+
this.pcm16 = Buffer.alloc(0);
|
|
67
|
+
this.currentSegmentId = uuidv4();
|
|
68
|
+
this.lastPartialText = "";
|
|
69
|
+
}
|
|
70
|
+
close() {
|
|
71
|
+
this.connected = false;
|
|
72
|
+
this.currentSegmentId = null;
|
|
73
|
+
this.pcm16 = Buffer.alloc(0);
|
|
74
|
+
}
|
|
75
|
+
async maybeDecode(force) {
|
|
76
|
+
if (!this.connected || !this.currentSegmentId) {
|
|
77
|
+
return;
|
|
78
|
+
}
|
|
79
|
+
const now = Date.now();
|
|
80
|
+
if (!force && now - this.lastDecodeAt < this.minDecodeIntervalMs) {
|
|
81
|
+
return;
|
|
82
|
+
}
|
|
83
|
+
if (this.decoding) {
|
|
84
|
+
this.pendingDecode = true;
|
|
85
|
+
return;
|
|
86
|
+
}
|
|
87
|
+
this.decoding = true;
|
|
88
|
+
try {
|
|
89
|
+
const text = await this.decodeNow();
|
|
90
|
+
this.lastDecodeAt = Date.now();
|
|
91
|
+
if (text !== this.lastPartialText) {
|
|
92
|
+
this.lastPartialText = text;
|
|
93
|
+
this.emit("transcript", { segmentId: this.currentSegmentId, transcript: text, isFinal: false });
|
|
94
|
+
}
|
|
95
|
+
}
|
|
96
|
+
finally {
|
|
97
|
+
this.decoding = false;
|
|
98
|
+
if (this.pendingDecode) {
|
|
99
|
+
this.pendingDecode = false;
|
|
100
|
+
await this.maybeDecode(true);
|
|
101
|
+
}
|
|
102
|
+
}
|
|
103
|
+
}
|
|
104
|
+
async decodeNow() {
|
|
105
|
+
if (this.pcm16.length === 0) {
|
|
106
|
+
return "";
|
|
107
|
+
}
|
|
108
|
+
const peak = pcm16lePeakAbs(this.pcm16);
|
|
109
|
+
const peakFloat = peak / 32768.0;
|
|
110
|
+
const targetPeak = 0.6;
|
|
111
|
+
const maxGain = 50;
|
|
112
|
+
const gain = peakFloat > 0 && peakFloat < targetPeak ? Math.min(maxGain, targetPeak / peakFloat) : 1;
|
|
113
|
+
const stream = this.engine.createStream();
|
|
114
|
+
try {
|
|
115
|
+
const floatSamples = pcm16leToFloat32(this.pcm16, gain);
|
|
116
|
+
this.engine.acceptWaveform(stream, this.engine.sampleRate, floatSamples);
|
|
117
|
+
this.engine.recognizer.decode(stream);
|
|
118
|
+
const result = this.engine.recognizer.getResult(stream);
|
|
119
|
+
return String(result?.text ?? result ?? "").trim();
|
|
120
|
+
}
|
|
121
|
+
finally {
|
|
122
|
+
try {
|
|
123
|
+
stream.free?.();
|
|
124
|
+
}
|
|
125
|
+
catch {
|
|
126
|
+
// ignore
|
|
127
|
+
}
|
|
128
|
+
}
|
|
129
|
+
}
|
|
130
|
+
}
|
|
131
|
+
//# sourceMappingURL=sherpa-parakeet-realtime-session.js.map
|
package/dist/server/server/speech/providers/local/sherpa/sherpa-parakeet-realtime-session.js.map
ADDED
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"sherpa-parakeet-realtime-session.js","sourceRoot":"","sources":["../../../../../../../src/server/speech/providers/local/sherpa/sherpa-parakeet-realtime-session.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAC3C,OAAO,EAAE,EAAE,IAAI,MAAM,EAAE,MAAM,MAAM,CAAC;AAGpC,OAAO,EAAE,cAAc,EAAE,gBAAgB,EAAE,MAAM,mBAAmB,CAAC;AAGrE,MAAM,OAAO,0CACX,SAAQ,YAAY;IAiBpB,YAAY,MAA+E;QACzF,KAAK,EAAE,CAAC;QAdF,cAAS,GAAG,KAAK,CAAC;QAGlB,qBAAgB,GAAkB,IAAI,CAAC;QACvC,sBAAiB,GAAkB,IAAI,CAAC;QACxC,oBAAe,GAAG,EAAE,CAAC;QAErB,UAAK,GAAW,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;QAChC,iBAAY,GAAG,CAAC,CAAC;QACjB,aAAQ,GAAG,KAAK,CAAC;QACjB,kBAAa,GAAG,KAAK,CAAC;QAK5B,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC,MAAM,CAAC;QAC5B,IAAI,CAAC,kBAAkB,GAAG,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC;QACjD,IAAI,CAAC,mBAAmB,GAAG,MAAM,CAAC,mBAAmB,IAAI,GAAG,CAAC;IAC/D,CAAC;IAED,KAAK,CAAC,OAAO;QACX,IAAI,IAAI,CAAC,SAAS,EAAE,CAAC;YACnB,OAAO;QACT,CAAC;QACD,IAAI,CAAC,gBAAgB,GAAG,MAAM,EAAE,CAAC;QACjC,IAAI,CAAC,SAAS,GAAG,IAAI,CAAC;IACxB,CAAC;IAED,WAAW,CAAC,KAAa;QACvB,IAAI,CAAC,IAAI,CAAC,SAAS,IAAI,CAAC,IAAI,CAAC,gBAAgB,EAAE,CAAC;YAC9C,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,IAAI,KAAK,CAAC,yCAAyC,CAAC,CAAC,CAAC;YACzE,OAAO;QACT,CAAC;QAED,IAAI,CAAC;YACH,IAAI,CAAC,KAAK,GAAG,IAAI,CAAC,KAAK,CAAC,MAAM,KAAK,CAAC,CAAC,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,MAAM,CAAC,MAAM,CAAC,CAAC,IAAI,CAAC,KAAK,EAAE,KAAK,CAAC,CAAC,CAAC;YAClF,KAAK,IAAI,CAAC,WAAW,CAAC,KAAK,CAAC,CAAC;QAC/B,CAAC;QAAC,OAAO,GAAG,EAAE,CAAC;YACb,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,GAAG,YAAY,KAAK,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,IAAI,KAAK,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC;QAC1E,CAAC;IACH,CAAC;IAED,MAAM;QACJ,IAAI,CAAC,IAAI,CAAC,SAAS,IAAI,CAAC,IAAI,CAAC,gBAAgB,EAAE,CAAC;YAC9C,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,IAAI,KAAK,CAAC,yCAAyC,CAAC,CAAC,CAAC;YACzE,OAAO;QACT,CAAC;QAED,KAAK,CAAC,KAAK,IAAI,EAAE;YACf,IAAI,CAAC;gBACH,MAAM,IAAI,CAAC,WAAW,CAAC,IAAI,CAAC,CAAC;gBAC7B,MAAM,SAAS,GAAG,IAAI,CAAC,eAAe,CAAC;gBACvC,MAAM,SAAS,GAAG,IAAI,CAAC,gBAAiB,CAAC;gBACzC,MAAM,iBAAiB,GAAG,IAAI,CAAC,iBAAiB,CAAC;gBAEjD,IAAI,CAAC,IAAI,CAAC,WAAW,EAAE,EAAE,SAAS,EAAE,iBAAiB,EAAE,CAAC,CAAC;gBACzD,IAAI,CAAC,IAAI,CAAC,YAAY,EAAE,EAAE,SAAS,EAAE,UAAU,EAAE,SAAS,EAAE,OAAO,EAAE,IAAI,EAAE,CAAC,CAAC;gBAE7E,IAAI,CAAC,iBAAiB,GAAG,SAAS,CAAC;gBACnC,IAAI,CAAC,gBAAgB,GAAG,MAAM,EAAE,CAAC;gBACjC,IAAI,CAAC,eAAe,GAAG,EAAE,CAAC;gBAC1B,IAAI,CAAC,KAAK,GAAG,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;YAC/B,CAAC;YAAC,OAAO,GAAG,EAAE,CAAC;gBACb,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,GAAG,YAAY,KAAK,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,IAAI,KAAK,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC;YAC1E,CAAC;QACH,CAAC,CAAC,EAAE,CAAC;IACP,CAAC;IAED,KAAK;QACH,IAAI,CAAC,IAAI,CAAC,SAAS,EAAE,CAAC;YACpB,OAAO;QACT,CAAC;QACD,IAAI,CAAC,KAAK,GAAG,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;QAC7B,IAAI,CAAC,gBAAgB,GAAG,MAAM,EAAE,CAAC;QACjC,IAAI,CAAC,eAAe,GAAG,EAAE,CAAC;IAC5B,CAAC;IAED,KAAK;QACH,IAAI,CAAC,SAAS,GAAG,KAAK,CAAC;QACvB,IAAI,CAAC,gBAAgB,GAAG,IAAI,CAAC;QAC7B,IAAI,CAAC,KAAK,GAAG,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;IAC/B,CAAC;IAEO,KAAK,CAAC,WAAW,CAAC,KAAc;QACtC,IAAI,CAAC,IAAI,CAAC,SAAS,IAAI,CAAC,IAAI,CAAC,gBAAgB,EAAE,CAAC;YAC9C,OAAO;QACT,CAAC;QAED,MAAM,GAAG,GAAG,IAAI,CAAC,GAAG,EAAE,CAAC;QACvB,IAAI,CAAC,KAAK,IAAI,GAAG,GAAG,IAAI,CAAC,YAAY,GAAG,IAAI,CAAC,mBAAmB,EAAE,CAAC;YACjE,OAAO;QACT,CAAC;QAED,IAAI,IAAI,CAAC,QAAQ,EAAE,CAAC;YAClB,IAAI,CAAC,aAAa,GAAG,IAAI,CAAC;YAC1B,OAAO;QACT,CAAC;QAED,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC;QACrB,IAAI,CAAC;YACH,MAAM,IAAI,GAAG,MAAM,IAAI,CAAC,SAAS,EAAE,CAAC;YACpC,IAAI,CAAC,YAAY,GAAG,IAAI,CAAC,GAAG,EAAE,CAAC;YAC/B,IAAI,IAAI,KAAK,IAAI,CAAC,eAAe,EAAE,CAAC;gBAClC,IAAI,CAAC,eAAe,GAAG,IAAI,CAAC;gBAC5B,IAAI,CAAC,IAAI,CAAC,YAAY,EAAE,EAAE,SAAS,EAAE,IAAI,CAAC,gBAAgB,EAAE,UAAU,EAAE,IAAI,EAAE,OAAO,EAAE,KAAK,EAAE,CAAC,CAAC;YAClG,CAAC;QACH,CAAC;gBAAS,CAAC;YACT,IAAI,CAAC,QAAQ,GAAG,KAAK,CAAC;YACtB,IAAI,IAAI,CAAC,aAAa,EAAE,CAAC;gBACvB,IAAI,CAAC,aAAa,GAAG,KAAK,CAAC;gBAC3B,MAAM,IAAI,CAAC,WAAW,CAAC,IAAI,CAAC,CAAC;YAC/B,CAAC;QACH,CAAC;IACH,CAAC;IAEO,KAAK,CAAC,SAAS;QACrB,IAAI,IAAI,CAAC,KAAK,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;YAC5B,OAAO,EAAE,CAAC;QACZ,CAAC;QAED,MAAM,IAAI,GAAG,cAAc,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;QACxC,MAAM,SAAS,GAAG,IAAI,GAAG,OAAO,CAAC;QACjC,MAAM,UAAU,GAAG,GAAG,CAAC;QACvB,MAAM,OAAO,GAAG,EAAE,CAAC;QACnB,MAAM,IAAI,GAAG,SAAS,GAAG,CAAC,IAAI,SAAS,GAAG,UAAU,CAAC,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,OAAO,EAAE,UAAU,GAAG,SAAS,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;QAErG,MAAM,MAAM,GAAG,IAAI,CAAC,MAAM,CAAC,YAAY,EAAE,CAAC;QAC1C,IAAI,CAAC;YACH,MAAM,YAAY,GAAG,gBAAgB,CAAC,IAAI,CAAC,KAAK,EAAE,IAAI,CAAC,CAAC;YACxD,IAAI,CAAC,MAAM,CAAC,cAAc,CAAC,MAAM,EAAE,IAAI,CAAC,MAAM,CAAC,UAAU,EAAE,YAAY,CAAC,CAAC;YACzE,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC,MAAM,CAAC,MAAM,CAAC,CAAC;YACtC,MAAM,MAAM,GAAG,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC,SAAS,CAAC,MAAM,CAAC,CAAC;YACxD,OAAO,MAAM,CAAC,MAAM,EAAE,IAAI,IAAI,MAAM,IAAI,EAAE,CAAC,CAAC,IAAI,EAAE,CAAC;QACrD,CAAC;gBAAS,CAAC;YACT,IAAI,CAAC;gBACH,MAAM,CAAC,IAAI,EAAE,EAAE,CAAC;YAClB,CAAC;YAAC,MAAM,CAAC;gBACP,SAAS;YACX,CAAC;QACH,CAAC;IACH,CAAC;CACF"}
|
|
@@ -0,0 +1,21 @@
|
|
|
1
|
+
import type pino from "pino";
|
|
2
|
+
import type { SpeechToTextProvider, StreamingTranscriptionSession, TranscriptionResult } from "../../../speech-provider.js";
|
|
3
|
+
import { SherpaOfflineRecognizerEngine } from "./sherpa-offline-recognizer.js";
|
|
4
|
+
export type SherpaParakeetSttConfig = {
|
|
5
|
+
engine: SherpaOfflineRecognizerEngine;
|
|
6
|
+
silencePeakThreshold?: number;
|
|
7
|
+
};
|
|
8
|
+
export declare class SherpaOnnxParakeetSTT implements SpeechToTextProvider {
|
|
9
|
+
private readonly engine;
|
|
10
|
+
private readonly silencePeakThreshold;
|
|
11
|
+
private readonly logger;
|
|
12
|
+
readonly id: "local";
|
|
13
|
+
constructor(config: SherpaParakeetSttConfig, logger: pino.Logger);
|
|
14
|
+
createSession(params: {
|
|
15
|
+
logger: pino.Logger;
|
|
16
|
+
language?: string;
|
|
17
|
+
prompt?: string;
|
|
18
|
+
}): StreamingTranscriptionSession;
|
|
19
|
+
transcribeAudio(audioBuffer: Buffer, format: string): Promise<TranscriptionResult>;
|
|
20
|
+
}
|
|
21
|
+
//# sourceMappingURL=sherpa-parakeet-stt.d.ts.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"sherpa-parakeet-stt.d.ts","sourceRoot":"","sources":["../../../../../../../src/server/speech/providers/local/sherpa/sherpa-parakeet-stt.ts"],"names":[],"mappings":"AAEA,OAAO,KAAK,IAAI,MAAM,MAAM,CAAC;AAE7B,OAAO,KAAK,EACV,oBAAoB,EACpB,6BAA6B,EAC7B,mBAAmB,EACpB,MAAM,6BAA6B,CAAC;AAGrC,OAAO,EAAE,6BAA6B,EAAE,MAAM,gCAAgC,CAAC;AAE/E,MAAM,MAAM,uBAAuB,GAAG;IACpC,MAAM,EAAE,6BAA6B,CAAC;IACtC,oBAAoB,CAAC,EAAE,MAAM,CAAC;CAC/B,CAAC;AAEF,qBAAa,qBAAsB,YAAW,oBAAoB;IAChE,OAAO,CAAC,QAAQ,CAAC,MAAM,CAAgC;IACvD,OAAO,CAAC,QAAQ,CAAC,oBAAoB,CAAS;IAC9C,OAAO,CAAC,QAAQ,CAAC,MAAM,CAAc;IACrC,SAAgB,EAAE,EAAG,OAAO,CAAU;gBAE1B,MAAM,EAAE,uBAAuB,EAAE,MAAM,EAAE,IAAI,CAAC,MAAM;IAMzD,aAAa,CAAC,MAAM,EAAE;QAC3B,MAAM,EAAE,IAAI,CAAC,MAAM,CAAC;QACpB,QAAQ,CAAC,EAAE,MAAM,CAAC;QAClB,MAAM,CAAC,EAAE,MAAM,CAAC;KACjB,GAAG,6BAA6B;IAoEpB,eAAe,CAAC,WAAW,EAAE,MAAM,EAAE,MAAM,EAAE,MAAM,GAAG,OAAO,CAAC,mBAAmB,CAAC;CAwDhG"}
|
|
@@ -0,0 +1,132 @@
|
|
|
1
|
+
import { EventEmitter } from "node:events";
|
|
2
|
+
import { v4 as uuidv4 } from "uuid";
|
|
3
|
+
import { Pcm16MonoResampler } from "../../../../agent/pcm16-resampler.js";
|
|
4
|
+
import { parsePcm16MonoWav, parsePcmRateFromFormat, pcm16lePeakAbs, pcm16leToFloat32 } from "../../../audio.js";
|
|
5
|
+
export class SherpaOnnxParakeetSTT {
|
|
6
|
+
constructor(config, logger) {
|
|
7
|
+
this.id = "local";
|
|
8
|
+
this.engine = config.engine;
|
|
9
|
+
this.silencePeakThreshold = config.silencePeakThreshold ?? 300;
|
|
10
|
+
this.logger = logger.child({ module: "speech", provider: "local", component: "parakeet-stt" });
|
|
11
|
+
}
|
|
12
|
+
createSession(params) {
|
|
13
|
+
const emitter = new EventEmitter();
|
|
14
|
+
const logger = params.logger.child({ provider: "local", component: "parakeet-stt-session" });
|
|
15
|
+
const requiredSampleRate = this.engine.sampleRate;
|
|
16
|
+
let connected = false;
|
|
17
|
+
let segmentId = uuidv4();
|
|
18
|
+
let previousSegmentId = null;
|
|
19
|
+
let pcm16 = Buffer.alloc(0);
|
|
20
|
+
return {
|
|
21
|
+
requiredSampleRate,
|
|
22
|
+
async connect() {
|
|
23
|
+
connected = true;
|
|
24
|
+
},
|
|
25
|
+
appendPcm16(chunk) {
|
|
26
|
+
if (!connected) {
|
|
27
|
+
emitter.emit("error", new Error("STT session not connected"));
|
|
28
|
+
return;
|
|
29
|
+
}
|
|
30
|
+
pcm16 = pcm16.length === 0 ? chunk : Buffer.concat([pcm16, chunk]);
|
|
31
|
+
},
|
|
32
|
+
commit: () => {
|
|
33
|
+
if (!connected) {
|
|
34
|
+
emitter.emit("error", new Error("STT session not connected"));
|
|
35
|
+
return;
|
|
36
|
+
}
|
|
37
|
+
const committedId = segmentId;
|
|
38
|
+
const prev = previousSegmentId;
|
|
39
|
+
emitter.emit("committed", { segmentId: committedId, previousSegmentId: prev });
|
|
40
|
+
void (async () => {
|
|
41
|
+
try {
|
|
42
|
+
const rt = await this.transcribeAudio(pcm16, `audio/pcm;rate=${requiredSampleRate}`);
|
|
43
|
+
emitter.emit("transcript", {
|
|
44
|
+
segmentId: committedId,
|
|
45
|
+
transcript: rt.text,
|
|
46
|
+
isFinal: true,
|
|
47
|
+
language: rt.language,
|
|
48
|
+
logprobs: rt.logprobs,
|
|
49
|
+
avgLogprob: rt.avgLogprob,
|
|
50
|
+
isLowConfidence: rt.isLowConfidence,
|
|
51
|
+
});
|
|
52
|
+
}
|
|
53
|
+
catch (err) {
|
|
54
|
+
emitter.emit("error", err);
|
|
55
|
+
}
|
|
56
|
+
finally {
|
|
57
|
+
previousSegmentId = committedId;
|
|
58
|
+
segmentId = uuidv4();
|
|
59
|
+
pcm16 = Buffer.alloc(0);
|
|
60
|
+
logger.debug({ bytes: pcm16.length }, "Parakeet session reset");
|
|
61
|
+
}
|
|
62
|
+
})();
|
|
63
|
+
},
|
|
64
|
+
clear() {
|
|
65
|
+
pcm16 = Buffer.alloc(0);
|
|
66
|
+
segmentId = uuidv4();
|
|
67
|
+
},
|
|
68
|
+
close() {
|
|
69
|
+
connected = false;
|
|
70
|
+
pcm16 = Buffer.alloc(0);
|
|
71
|
+
},
|
|
72
|
+
on(event, handler) {
|
|
73
|
+
emitter.on(event, handler);
|
|
74
|
+
return undefined;
|
|
75
|
+
},
|
|
76
|
+
};
|
|
77
|
+
}
|
|
78
|
+
async transcribeAudio(audioBuffer, format) {
|
|
79
|
+
const start = Date.now();
|
|
80
|
+
let inputRate;
|
|
81
|
+
let pcm16;
|
|
82
|
+
if (format.toLowerCase().includes("audio/wav")) {
|
|
83
|
+
const parsed = parsePcm16MonoWav(audioBuffer);
|
|
84
|
+
inputRate = parsed.sampleRate;
|
|
85
|
+
pcm16 = parsed.pcm16;
|
|
86
|
+
}
|
|
87
|
+
else if (format.toLowerCase().includes("audio/pcm")) {
|
|
88
|
+
inputRate = parsePcmRateFromFormat(format, this.engine.sampleRate) ?? this.engine.sampleRate;
|
|
89
|
+
pcm16 = audioBuffer;
|
|
90
|
+
}
|
|
91
|
+
else {
|
|
92
|
+
throw new Error(`Unsupported audio format for sherpa Parakeet STT: ${format}`);
|
|
93
|
+
}
|
|
94
|
+
const peak = pcm16lePeakAbs(pcm16);
|
|
95
|
+
if (peak < this.silencePeakThreshold) {
|
|
96
|
+
return { text: "", duration: Date.now() - start, isLowConfidence: true };
|
|
97
|
+
}
|
|
98
|
+
let pcmForModel = pcm16;
|
|
99
|
+
if (inputRate !== this.engine.sampleRate) {
|
|
100
|
+
const resampler = new Pcm16MonoResampler({ inputRate, outputRate: this.engine.sampleRate });
|
|
101
|
+
pcmForModel = resampler.processChunk(pcm16);
|
|
102
|
+
inputRate = this.engine.sampleRate;
|
|
103
|
+
}
|
|
104
|
+
const peakForModel = pcm16lePeakAbs(pcmForModel);
|
|
105
|
+
const peakFloat = peakForModel / 32768.0;
|
|
106
|
+
const targetPeak = 0.6;
|
|
107
|
+
const maxGain = 50;
|
|
108
|
+
const gain = peakFloat > 0 && peakFloat < targetPeak
|
|
109
|
+
? Math.min(maxGain, targetPeak / peakFloat)
|
|
110
|
+
: 1;
|
|
111
|
+
const stream = this.engine.createStream();
|
|
112
|
+
try {
|
|
113
|
+
const floatSamples = pcm16leToFloat32(pcmForModel, gain);
|
|
114
|
+
this.engine.acceptWaveform(stream, inputRate, floatSamples);
|
|
115
|
+
this.engine.recognizer.decode(stream);
|
|
116
|
+
const result = this.engine.recognizer.getResult(stream);
|
|
117
|
+
const text = String(result?.text ?? result ?? "").trim();
|
|
118
|
+
const duration = Date.now() - start;
|
|
119
|
+
this.logger.debug({ duration, textLength: text.length }, "Parakeet transcription complete");
|
|
120
|
+
return { text, duration, ...(text.length === 0 ? { isLowConfidence: true } : {}) };
|
|
121
|
+
}
|
|
122
|
+
finally {
|
|
123
|
+
try {
|
|
124
|
+
stream.free?.();
|
|
125
|
+
}
|
|
126
|
+
catch {
|
|
127
|
+
// ignore
|
|
128
|
+
}
|
|
129
|
+
}
|
|
130
|
+
}
|
|
131
|
+
}
|
|
132
|
+
//# sourceMappingURL=sherpa-parakeet-stt.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"sherpa-parakeet-stt.js","sourceRoot":"","sources":["../../../../../../../src/server/speech/providers/local/sherpa/sherpa-parakeet-stt.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAC3C,OAAO,EAAE,EAAE,IAAI,MAAM,EAAE,MAAM,MAAM,CAAC;AAQpC,OAAO,EAAE,kBAAkB,EAAE,MAAM,sCAAsC,CAAC;AAC1E,OAAO,EAAE,iBAAiB,EAAE,sBAAsB,EAAE,cAAc,EAAE,gBAAgB,EAAE,MAAM,mBAAmB,CAAC;AAQhH,MAAM,OAAO,qBAAqB;IAMhC,YAAY,MAA+B,EAAE,MAAmB;QAFhD,OAAE,GAAG,OAAgB,CAAC;QAGpC,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC,MAAM,CAAC;QAC5B,IAAI,CAAC,oBAAoB,GAAG,MAAM,CAAC,oBAAoB,IAAI,GAAG,CAAC;QAC/D,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC,KAAK,CAAC,EAAE,MAAM,EAAE,QAAQ,EAAE,QAAQ,EAAE,OAAO,EAAE,SAAS,EAAE,cAAc,EAAE,CAAC,CAAC;IACjG,CAAC;IAEM,aAAa,CAAC,MAIpB;QACC,MAAM,OAAO,GAAG,IAAI,YAAY,EAAE,CAAC;QACnC,MAAM,MAAM,GAAG,MAAM,CAAC,MAAM,CAAC,KAAK,CAAC,EAAE,QAAQ,EAAE,OAAO,EAAE,SAAS,EAAE,sBAAsB,EAAE,CAAC,CAAC;QAC7F,MAAM,kBAAkB,GAAG,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC;QAClD,IAAI,SAAS,GAAG,KAAK,CAAC;QACtB,IAAI,SAAS,GAAG,MAAM,EAAE,CAAC;QACzB,IAAI,iBAAiB,GAAkB,IAAI,CAAC;QAC5C,IAAI,KAAK,GAAW,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;QAEpC,OAAO;YACL,kBAAkB;YAClB,KAAK,CAAC,OAAO;gBACX,SAAS,GAAG,IAAI,CAAC;YACnB,CAAC;YACD,WAAW,CAAC,KAAa;gBACvB,IAAI,CAAC,SAAS,EAAE,CAAC;oBACd,OAAe,CAAC,IAAI,CAAC,OAAO,EAAE,IAAI,KAAK,CAAC,2BAA2B,CAAC,CAAC,CAAC;oBACvE,OAAO;gBACT,CAAC;gBACD,KAAK,GAAG,KAAK,CAAC,MAAM,KAAK,CAAC,CAAC,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,MAAM,CAAC,MAAM,CAAC,CAAC,KAAK,EAAE,KAAK,CAAC,CAAC,CAAC;YACrE,CAAC;YACD,MAAM,EAAE,GAAG,EAAE;gBACX,IAAI,CAAC,SAAS,EAAE,CAAC;oBACd,OAAe,CAAC,IAAI,CAAC,OAAO,EAAE,IAAI,KAAK,CAAC,2BAA2B,CAAC,CAAC,CAAC;oBACvE,OAAO;gBACT,CAAC;gBAED,MAAM,WAAW,GAAG,SAAS,CAAC;gBAC9B,MAAM,IAAI,GAAG,iBAAiB,CAAC;gBAC9B,OAAe,CAAC,IAAI,CAAC,WAAW,EAAE,EAAE,SAAS,EAAE,WAAW,EAAE,iBAAiB,EAAE,IAAI,EAAE,CAAC,CAAC;gBAExF,KAAK,CAAC,KAAK,IAAI,EAAE;oBACf,IAAI,CAAC;wBACH,MAAM,EAAE,GAAG,MAAM,IAAI,CAAC,eAAe,CAAC,KAAK,EAAE,kBAAkB,kBAAkB,EAAE,CAAC,CAAC;wBACpF,OAAe,CAAC,IAAI,CAAC,YAAY,EAAE;4BAClC,SAAS,EAAE,WAAW;4BACtB,UAAU,EAAE,EAAE,CAAC,IAAI;4BACnB,OAAO,EAAE,IAAI;4BACb,QAAQ,EAAE,EAAE,CAAC,QAAQ;4BACrB,QAAQ,EAAE,EAAE,CAAC,QAAQ;4BACrB,UAAU,EAAE,EAAE,CAAC,UAAU;4BACzB,eAAe,EAAE,EAAE,CAAC,eAAe;yBACpC,CAAC,CAAC;oBACL,CAAC;oBAAC,OAAO,GAAG,EAAE,CAAC;wBACZ,OAAe,CAAC,IAAI,CAAC,OAAO,EAAE,GAAG,CAAC,CAAC;oBACtC,CAAC;4BAAS,CAAC;wBACT,iBAAiB,GAAG,WAAW,CAAC;wBAChC,SAAS,GAAG,MAAM,EAAE,CAAC;wBACrB,KAAK,GAAG,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;wBACxB,MAAM,CAAC,KAAK,CAAC,EAAE,KAAK,EAAE,KAAK,CAAC,MAAM,EAAE,EAAE,wBAAwB,CAAC,CAAC;oBAClE,CAAC;gBACH,CAAC,CAAC,EAAE,CAAC;YACP,CAAC;YACD,KAAK;gBACH,KAAK,GAAG,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;gBACxB,SAAS,GAAG,MAAM,EAAE,CAAC;YACvB,CAAC;YACD,KAAK;gBACH,SAAS,GAAG,KAAK,CAAC;gBAClB,KAAK,GAAG,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;YAC1B,CAAC;YACD,EAAE,CAAC,KAAU,EAAE,OAAY;gBACzB,OAAO,CAAC,EAAE,CAAC,KAAK,EAAE,OAAO,CAAC,CAAC;gBAC3B,OAAO,SAAS,CAAC;YACnB,CAAC;SACF,CAAC;IACJ,CAAC;IAEM,KAAK,CAAC,eAAe,CAAC,WAAmB,EAAE,MAAc;QAC9D,MAAM,KAAK,GAAG,IAAI,CAAC,GAAG,EAAE,CAAC;QAEzB,IAAI,SAAiB,CAAC;QACtB,IAAI,KAAa,CAAC;QAElB,IAAI,MAAM,CAAC,WAAW,EAAE,CAAC,QAAQ,CAAC,WAAW,CAAC,EAAE,CAAC;YAC/C,MAAM,MAAM,GAAG,iBAAiB,CAAC,WAAW,CAAC,CAAC;YAC9C,SAAS,GAAG,MAAM,CAAC,UAAU,CAAC;YAC9B,KAAK,GAAG,MAAM,CAAC,KAAK,CAAC;QACvB,CAAC;aAAM,IAAI,MAAM,CAAC,WAAW,EAAE,CAAC,QAAQ,CAAC,WAAW,CAAC,EAAE,CAAC;YACtD,SAAS,GAAG,sBAAsB,CAAC,MAAM,EAAE,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC,IAAI,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC;YAC7F,KAAK,GAAG,WAAW,CAAC;QACtB,CAAC;aAAM,CAAC;YACN,MAAM,IAAI,KAAK,CAAC,qDAAqD,MAAM,EAAE,CAAC,CAAC;QACjF,CAAC;QAED,MAAM,IAAI,GAAG,cAAc,CAAC,KAAK,CAAC,CAAC;QACnC,IAAI,IAAI,GAAG,IAAI,CAAC,oBAAoB,EAAE,CAAC;YACrC,OAAO,EAAE,IAAI,EAAE,EAAE,EAAE,QAAQ,EAAE,IAAI,CAAC,GAAG,EAAE,GAAG,KAAK,EAAE,eAAe,EAAE,IAAI,EAAE,CAAC;QAC3E,CAAC;QAED,IAAI,WAAW,GAAG,KAAK,CAAC;QACxB,IAAI,SAAS,KAAK,IAAI,CAAC,MAAM,CAAC,UAAU,EAAE,CAAC;YACzC,MAAM,SAAS,GAAG,IAAI,kBAAkB,CAAC,EAAE,SAAS,EAAE,UAAU,EAAE,IAAI,CAAC,MAAM,CAAC,UAAU,EAAE,CAAC,CAAC;YAC5F,WAAW,GAAG,SAAS,CAAC,YAAY,CAAC,KAAK,CAAC,CAAC;YAC5C,SAAS,GAAG,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC;QACrC,CAAC;QAED,MAAM,YAAY,GAAG,cAAc,CAAC,WAAW,CAAC,CAAC;QACjD,MAAM,SAAS,GAAG,YAAY,GAAG,OAAO,CAAC;QACzC,MAAM,UAAU,GAAG,GAAG,CAAC;QACvB,MAAM,OAAO,GAAG,EAAE,CAAC;QACnB,MAAM,IAAI,GACR,SAAS,GAAG,CAAC,IAAI,SAAS,GAAG,UAAU;YACrC,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,OAAO,EAAE,UAAU,GAAG,SAAS,CAAC;YAC3C,CAAC,CAAC,CAAC,CAAC;QAER,MAAM,MAAM,GAAG,IAAI,CAAC,MAAM,CAAC,YAAY,EAAE,CAAC;QAC1C,IAAI,CAAC;YACH,MAAM,YAAY,GAAG,gBAAgB,CAAC,WAAW,EAAE,IAAI,CAAC,CAAC;YACzD,IAAI,CAAC,MAAM,CAAC,cAAc,CAAC,MAAM,EAAE,SAAS,EAAE,YAAY,CAAC,CAAC;YAC5D,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC,MAAM,CAAC,MAAM,CAAC,CAAC;YACtC,MAAM,MAAM,GAAG,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC,SAAS,CAAC,MAAM,CAAC,CAAC;YACxD,MAAM,IAAI,GAAG,MAAM,CAAC,MAAM,EAAE,IAAI,IAAI,MAAM,IAAI,EAAE,CAAC,CAAC,IAAI,EAAE,CAAC;YACzD,MAAM,QAAQ,GAAG,IAAI,CAAC,GAAG,EAAE,GAAG,KAAK,CAAC;YACpC,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,EAAE,QAAQ,EAAE,UAAU,EAAE,IAAI,CAAC,MAAM,EAAE,EAAE,iCAAiC,CAAC,CAAC;YAC5F,OAAO,EAAE,IAAI,EAAE,QAAQ,EAAE,GAAG,CAAC,IAAI,CAAC,MAAM,KAAK,CAAC,CAAC,CAAC,CAAC,EAAE,eAAe,EAAE,IAAI,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,CAAC;QACrF,CAAC;gBAAS,CAAC;YACT,IAAI,CAAC;gBACH,MAAM,CAAC,IAAI,EAAE,EAAE,CAAC;YAClB,CAAC;YAAC,MAAM,CAAC;gBACP,SAAS;YACX,CAAC;QACH,CAAC;IACH,CAAC;CACF"}
|
|
@@ -0,0 +1,23 @@
|
|
|
1
|
+
import { EventEmitter } from "node:events";
|
|
2
|
+
import type { StreamingTranscriptionSession } from "../../../speech-provider.js";
|
|
3
|
+
import { SherpaOnlineRecognizerEngine } from "./sherpa-online-recognizer.js";
|
|
4
|
+
export declare class SherpaRealtimeTranscriptionSession extends EventEmitter implements StreamingTranscriptionSession {
|
|
5
|
+
private readonly engine;
|
|
6
|
+
private stream;
|
|
7
|
+
private connected;
|
|
8
|
+
readonly requiredSampleRate: number;
|
|
9
|
+
private currentSegmentId;
|
|
10
|
+
private previousSegmentId;
|
|
11
|
+
private lastPartialText;
|
|
12
|
+
private readonly tailPaddingMs;
|
|
13
|
+
constructor(params: {
|
|
14
|
+
engine: SherpaOnlineRecognizerEngine;
|
|
15
|
+
tailPaddingMs?: number;
|
|
16
|
+
});
|
|
17
|
+
connect(): Promise<void>;
|
|
18
|
+
appendPcm16(pcm16le: Buffer): void;
|
|
19
|
+
commit(): void;
|
|
20
|
+
clear(): void;
|
|
21
|
+
close(): void;
|
|
22
|
+
}
|
|
23
|
+
//# sourceMappingURL=sherpa-realtime-session.d.ts.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"sherpa-realtime-session.d.ts","sourceRoot":"","sources":["../../../../../../../src/server/speech/providers/local/sherpa/sherpa-realtime-session.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAG3C,OAAO,KAAK,EAAE,6BAA6B,EAAE,MAAM,6BAA6B,CAAC;AAEjF,OAAO,EAAE,4BAA4B,EAAE,MAAM,+BAA+B,CAAC;AAE7E,qBAAa,kCACX,SAAQ,YACR,YAAW,6BAA6B;IAExC,OAAO,CAAC,QAAQ,CAAC,MAAM,CAA+B;IACtD,OAAO,CAAC,MAAM,CAAoB;IAClC,OAAO,CAAC,SAAS,CAAS;IAE1B,SAAgB,kBAAkB,EAAE,MAAM,CAAC;IAC3C,OAAO,CAAC,gBAAgB,CAAuB;IAC/C,OAAO,CAAC,iBAAiB,CAAuB;IAChD,OAAO,CAAC,eAAe,CAAM;IAC7B,OAAO,CAAC,QAAQ,CAAC,aAAa,CAAS;gBAE3B,MAAM,EAAE;QAAE,MAAM,EAAE,4BAA4B,CAAC;QAAC,aAAa,CAAC,EAAE,MAAM,CAAA;KAAE;IAO9E,OAAO,IAAI,OAAO,CAAC,IAAI,CAAC;IAS9B,WAAW,CAAC,OAAO,EAAE,MAAM,GAAG,IAAI;IAoClC,MAAM,IAAI,IAAI;IAgCd,KAAK,IAAI,IAAI;IAab,KAAK,IAAI,IAAI;CAad"}
|
|
@@ -0,0 +1,112 @@
|
|
|
1
|
+
import { EventEmitter } from "node:events";
|
|
2
|
+
import { v4 as uuidv4 } from "uuid";
|
|
3
|
+
import { pcm16lePeakAbs, pcm16leToFloat32 } from "../../../audio.js";
|
|
4
|
+
export class SherpaRealtimeTranscriptionSession extends EventEmitter {
|
|
5
|
+
constructor(params) {
|
|
6
|
+
super();
|
|
7
|
+
this.stream = null;
|
|
8
|
+
this.connected = false;
|
|
9
|
+
this.currentSegmentId = null;
|
|
10
|
+
this.previousSegmentId = null;
|
|
11
|
+
this.lastPartialText = "";
|
|
12
|
+
this.engine = params.engine;
|
|
13
|
+
this.requiredSampleRate = this.engine.sampleRate;
|
|
14
|
+
this.tailPaddingMs = params.tailPaddingMs ?? 500;
|
|
15
|
+
}
|
|
16
|
+
async connect() {
|
|
17
|
+
if (this.connected) {
|
|
18
|
+
return;
|
|
19
|
+
}
|
|
20
|
+
this.stream = this.engine.createStream();
|
|
21
|
+
this.currentSegmentId = uuidv4();
|
|
22
|
+
this.connected = true;
|
|
23
|
+
}
|
|
24
|
+
appendPcm16(pcm16le) {
|
|
25
|
+
if (!this.connected || !this.stream || !this.currentSegmentId) {
|
|
26
|
+
this.emit("error", new Error("Sherpa realtime session not connected"));
|
|
27
|
+
return;
|
|
28
|
+
}
|
|
29
|
+
try {
|
|
30
|
+
const peak = pcm16lePeakAbs(pcm16le);
|
|
31
|
+
const peakFloat = peak / 32768.0;
|
|
32
|
+
const targetPeak = 0.6;
|
|
33
|
+
const maxGain = 50;
|
|
34
|
+
const gain = peakFloat > 0 && peakFloat < targetPeak
|
|
35
|
+
? Math.min(maxGain, targetPeak / peakFloat)
|
|
36
|
+
: 1;
|
|
37
|
+
const floatSamples = pcm16leToFloat32(pcm16le, gain);
|
|
38
|
+
this.stream.acceptWaveform(this.engine.sampleRate, floatSamples);
|
|
39
|
+
while (this.engine.recognizer.isReady(this.stream)) {
|
|
40
|
+
this.engine.recognizer.decode(this.stream);
|
|
41
|
+
}
|
|
42
|
+
const text = String(this.engine.recognizer.getResult(this.stream)?.text ?? "").trim();
|
|
43
|
+
if (text !== this.lastPartialText) {
|
|
44
|
+
this.lastPartialText = text;
|
|
45
|
+
this.emit("transcript", {
|
|
46
|
+
segmentId: this.currentSegmentId,
|
|
47
|
+
transcript: text,
|
|
48
|
+
isFinal: false,
|
|
49
|
+
});
|
|
50
|
+
}
|
|
51
|
+
}
|
|
52
|
+
catch (err) {
|
|
53
|
+
this.emit("error", err instanceof Error ? err : new Error(String(err)));
|
|
54
|
+
}
|
|
55
|
+
}
|
|
56
|
+
commit() {
|
|
57
|
+
if (!this.connected || !this.stream || !this.currentSegmentId) {
|
|
58
|
+
this.emit("error", new Error("Sherpa realtime session not connected"));
|
|
59
|
+
return;
|
|
60
|
+
}
|
|
61
|
+
try {
|
|
62
|
+
const padSamples = Math.max(0, Math.round((this.engine.sampleRate * this.tailPaddingMs) / 1000));
|
|
63
|
+
if (padSamples > 0) {
|
|
64
|
+
this.stream.acceptWaveform(this.engine.sampleRate, new Float32Array(padSamples));
|
|
65
|
+
}
|
|
66
|
+
while (this.engine.recognizer.isReady(this.stream)) {
|
|
67
|
+
this.engine.recognizer.decode(this.stream);
|
|
68
|
+
}
|
|
69
|
+
const finalText = String(this.engine.recognizer.getResult(this.stream)?.text ?? "").trim();
|
|
70
|
+
const segmentId = this.currentSegmentId;
|
|
71
|
+
const previousSegmentId = this.previousSegmentId;
|
|
72
|
+
this.emit("committed", { segmentId, previousSegmentId });
|
|
73
|
+
this.emit("transcript", { segmentId, transcript: finalText, isFinal: true });
|
|
74
|
+
this.previousSegmentId = segmentId;
|
|
75
|
+
this.currentSegmentId = uuidv4();
|
|
76
|
+
this.lastPartialText = "";
|
|
77
|
+
this.engine.recognizer.reset(this.stream);
|
|
78
|
+
}
|
|
79
|
+
catch (err) {
|
|
80
|
+
this.emit("error", err instanceof Error ? err : new Error(String(err)));
|
|
81
|
+
}
|
|
82
|
+
}
|
|
83
|
+
clear() {
|
|
84
|
+
if (!this.connected || !this.stream) {
|
|
85
|
+
return;
|
|
86
|
+
}
|
|
87
|
+
try {
|
|
88
|
+
this.engine.recognizer.reset(this.stream);
|
|
89
|
+
this.currentSegmentId = uuidv4();
|
|
90
|
+
this.lastPartialText = "";
|
|
91
|
+
}
|
|
92
|
+
catch (err) {
|
|
93
|
+
this.emit("error", err instanceof Error ? err : new Error(String(err)));
|
|
94
|
+
}
|
|
95
|
+
}
|
|
96
|
+
close() {
|
|
97
|
+
if (!this.stream) {
|
|
98
|
+
return;
|
|
99
|
+
}
|
|
100
|
+
try {
|
|
101
|
+
this.stream.free?.();
|
|
102
|
+
}
|
|
103
|
+
catch {
|
|
104
|
+
// ignore
|
|
105
|
+
}
|
|
106
|
+
finally {
|
|
107
|
+
this.stream = null;
|
|
108
|
+
this.connected = false;
|
|
109
|
+
}
|
|
110
|
+
}
|
|
111
|
+
}
|
|
112
|
+
//# sourceMappingURL=sherpa-realtime-session.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"sherpa-realtime-session.js","sourceRoot":"","sources":["../../../../../../../src/server/speech/providers/local/sherpa/sherpa-realtime-session.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAC3C,OAAO,EAAE,EAAE,IAAI,MAAM,EAAE,MAAM,MAAM,CAAC;AAGpC,OAAO,EAAE,cAAc,EAAE,gBAAgB,EAAE,MAAM,mBAAmB,CAAC;AAGrE,MAAM,OAAO,kCACX,SAAQ,YAAY;IAapB,YAAY,MAAwE;QAClF,KAAK,EAAE,CAAC;QAVF,WAAM,GAAe,IAAI,CAAC;QAC1B,cAAS,GAAG,KAAK,CAAC;QAGlB,qBAAgB,GAAkB,IAAI,CAAC;QACvC,sBAAiB,GAAkB,IAAI,CAAC;QACxC,oBAAe,GAAG,EAAE,CAAC;QAK3B,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC,MAAM,CAAC;QAC5B,IAAI,CAAC,kBAAkB,GAAG,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC;QACjD,IAAI,CAAC,aAAa,GAAG,MAAM,CAAC,aAAa,IAAI,GAAG,CAAC;IACnD,CAAC;IAED,KAAK,CAAC,OAAO;QACX,IAAI,IAAI,CAAC,SAAS,EAAE,CAAC;YACnB,OAAO;QACT,CAAC;QACD,IAAI,CAAC,MAAM,GAAG,IAAI,CAAC,MAAM,CAAC,YAAY,EAAE,CAAC;QACzC,IAAI,CAAC,gBAAgB,GAAG,MAAM,EAAE,CAAC;QACjC,IAAI,CAAC,SAAS,GAAG,IAAI,CAAC;IACxB,CAAC;IAED,WAAW,CAAC,OAAe;QACzB,IAAI,CAAC,IAAI,CAAC,SAAS,IAAI,CAAC,IAAI,CAAC,MAAM,IAAI,CAAC,IAAI,CAAC,gBAAgB,EAAE,CAAC;YAC9D,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,IAAI,KAAK,CAAC,uCAAuC,CAAC,CAAC,CAAC;YACvE,OAAO;QACT,CAAC;QAED,IAAI,CAAC;YACH,MAAM,IAAI,GAAG,cAAc,CAAC,OAAO,CAAC,CAAC;YACrC,MAAM,SAAS,GAAG,IAAI,GAAG,OAAO,CAAC;YACjC,MAAM,UAAU,GAAG,GAAG,CAAC;YACvB,MAAM,OAAO,GAAG,EAAE,CAAC;YACnB,MAAM,IAAI,GACR,SAAS,GAAG,CAAC,IAAI,SAAS,GAAG,UAAU;gBACrC,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,OAAO,EAAE,UAAU,GAAG,SAAS,CAAC;gBAC3C,CAAC,CAAC,CAAC,CAAC;YACR,MAAM,YAAY,GAAG,gBAAgB,CAAC,OAAO,EAAE,IAAI,CAAC,CAAC;YACrD,IAAI,CAAC,MAAM,CAAC,cAAc,CAAC,IAAI,CAAC,MAAM,CAAC,UAAU,EAAE,YAAY,CAAC,CAAC;YAEjE,OAAO,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC,OAAO,CAAC,IAAI,CAAC,MAAM,CAAC,EAAE,CAAC;gBACnD,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC,MAAM,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC;YAC7C,CAAC;YAED,MAAM,IAAI,GAAG,MAAM,CAAC,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC,SAAS,CAAC,IAAI,CAAC,MAAM,CAAC,EAAE,IAAI,IAAI,EAAE,CAAC,CAAC,IAAI,EAAE,CAAC;YACtF,IAAI,IAAI,KAAK,IAAI,CAAC,eAAe,EAAE,CAAC;gBAClC,IAAI,CAAC,eAAe,GAAG,IAAI,CAAC;gBAC5B,IAAI,CAAC,IAAI,CAAC,YAAY,EAAE;oBACtB,SAAS,EAAE,IAAI,CAAC,gBAAgB;oBAChC,UAAU,EAAE,IAAI;oBAChB,OAAO,EAAE,KAAK;iBACf,CAAC,CAAC;YACL,CAAC;QACH,CAAC;QAAC,OAAO,GAAG,EAAE,CAAC;YACb,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,GAAG,YAAY,KAAK,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,IAAI,KAAK,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC;QAC1E,CAAC;IACH,CAAC;IAED,MAAM;QACJ,IAAI,CAAC,IAAI,CAAC,SAAS,IAAI,CAAC,IAAI,CAAC,MAAM,IAAI,CAAC,IAAI,CAAC,gBAAgB,EAAE,CAAC;YAC9D,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,IAAI,KAAK,CAAC,uCAAuC,CAAC,CAAC,CAAC;YACvE,OAAO;QACT,CAAC;QAED,IAAI,CAAC;YACH,MAAM,UAAU,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,CAAC,KAAK,CAAC,CAAC,IAAI,CAAC,MAAM,CAAC,UAAU,GAAG,IAAI,CAAC,aAAa,CAAC,GAAG,IAAI,CAAC,CAAC,CAAC;YACjG,IAAI,UAAU,GAAG,CAAC,EAAE,CAAC;gBACnB,IAAI,CAAC,MAAM,CAAC,cAAc,CAAC,IAAI,CAAC,MAAM,CAAC,UAAU,EAAE,IAAI,YAAY,CAAC,UAAU,CAAC,CAAC,CAAC;YACnF,CAAC;YAED,OAAO,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC,OAAO,CAAC,IAAI,CAAC,MAAM,CAAC,EAAE,CAAC;gBACnD,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC,MAAM,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC;YAC7C,CAAC;YAED,MAAM,SAAS,GAAG,MAAM,CAAC,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC,SAAS,CAAC,IAAI,CAAC,MAAM,CAAC,EAAE,IAAI,IAAI,EAAE,CAAC,CAAC,IAAI,EAAE,CAAC;YAC3F,MAAM,SAAS,GAAG,IAAI,CAAC,gBAAgB,CAAC;YACxC,MAAM,iBAAiB,GAAG,IAAI,CAAC,iBAAiB,CAAC;YAEjD,IAAI,CAAC,IAAI,CAAC,WAAW,EAAE,EAAE,SAAS,EAAE,iBAAiB,EAAE,CAAC,CAAC;YACzD,IAAI,CAAC,IAAI,CAAC,YAAY,EAAE,EAAE,SAAS,EAAE,UAAU,EAAE,SAAS,EAAE,OAAO,EAAE,IAAI,EAAE,CAAC,CAAC;YAE7E,IAAI,CAAC,iBAAiB,GAAG,SAAS,CAAC;YACnC,IAAI,CAAC,gBAAgB,GAAG,MAAM,EAAE,CAAC;YACjC,IAAI,CAAC,eAAe,GAAG,EAAE,CAAC;YAC1B,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC,KAAK,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC;QAC5C,CAAC;QAAC,OAAO,GAAG,EAAE,CAAC;YACb,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,GAAG,YAAY,KAAK,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,IAAI,KAAK,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC;QAC1E,CAAC;IACH,CAAC;IAED,KAAK;QACH,IAAI,CAAC,IAAI,CAAC,SAAS,IAAI,CAAC,IAAI,CAAC,MAAM,EAAE,CAAC;YACpC,OAAO;QACT,CAAC;QACD,IAAI,CAAC;YACH,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC,KAAK,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC;YAC1C,IAAI,CAAC,gBAAgB,GAAG,MAAM,EAAE,CAAC;YACjC,IAAI,CAAC,eAAe,GAAG,EAAE,CAAC;QAC5B,CAAC;QAAC,OAAO,GAAG,EAAE,CAAC;YACb,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,GAAG,YAAY,KAAK,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,IAAI,KAAK,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC;QAC1E,CAAC;IACH,CAAC;IAED,KAAK;QACH,IAAI,CAAC,IAAI,CAAC,MAAM,EAAE,CAAC;YACjB,OAAO;QACT,CAAC;QACD,IAAI,CAAC;YACH,IAAI,CAAC,MAAM,CAAC,IAAI,EAAE,EAAE,CAAC;QACvB,CAAC;QAAC,MAAM,CAAC;YACP,SAAS;QACX,CAAC;gBAAS,CAAC;YACT,IAAI,CAAC,MAAM,GAAG,IAAI,CAAC;YACnB,IAAI,CAAC,SAAS,GAAG,KAAK,CAAC;QACzB,CAAC;IACH,CAAC;CACF"}
|
|
@@ -0,0 +1,23 @@
|
|
|
1
|
+
import type pino from "pino";
|
|
2
|
+
import type { SpeechToTextProvider, StreamingTranscriptionSession, TranscriptionResult } from "../../../speech-provider.js";
|
|
3
|
+
import { SherpaOnlineRecognizerEngine } from "./sherpa-online-recognizer.js";
|
|
4
|
+
export type SherpaSttConfig = {
|
|
5
|
+
engine: SherpaOnlineRecognizerEngine;
|
|
6
|
+
silencePeakThreshold?: number;
|
|
7
|
+
tailPaddingMs?: number;
|
|
8
|
+
};
|
|
9
|
+
export declare class SherpaOnnxSTT implements SpeechToTextProvider {
|
|
10
|
+
private readonly engine;
|
|
11
|
+
private readonly silencePeakThreshold;
|
|
12
|
+
private readonly tailPaddingMs;
|
|
13
|
+
private readonly logger;
|
|
14
|
+
readonly id: "local";
|
|
15
|
+
constructor(config: SherpaSttConfig, logger: pino.Logger);
|
|
16
|
+
createSession(params: {
|
|
17
|
+
logger: pino.Logger;
|
|
18
|
+
language?: string;
|
|
19
|
+
prompt?: string;
|
|
20
|
+
}): StreamingTranscriptionSession;
|
|
21
|
+
transcribeAudio(audioBuffer: Buffer, format: string): Promise<TranscriptionResult>;
|
|
22
|
+
}
|
|
23
|
+
//# sourceMappingURL=sherpa-stt.d.ts.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"sherpa-stt.d.ts","sourceRoot":"","sources":["../../../../../../../src/server/speech/providers/local/sherpa/sherpa-stt.ts"],"names":[],"mappings":"AAEA,OAAO,KAAK,IAAI,MAAM,MAAM,CAAC;AAE7B,OAAO,KAAK,EACV,oBAAoB,EACpB,6BAA6B,EAC7B,mBAAmB,EACpB,MAAM,6BAA6B,CAAC;AAGrC,OAAO,EAAE,4BAA4B,EAAE,MAAM,+BAA+B,CAAC;AAE7E,MAAM,MAAM,eAAe,GAAG;IAC5B,MAAM,EAAE,4BAA4B,CAAC;IACrC,oBAAoB,CAAC,EAAE,MAAM,CAAC;IAC9B,aAAa,CAAC,EAAE,MAAM,CAAC;CACxB,CAAC;AAEF,qBAAa,aAAc,YAAW,oBAAoB;IACxD,OAAO,CAAC,QAAQ,CAAC,MAAM,CAA+B;IACtD,OAAO,CAAC,QAAQ,CAAC,oBAAoB,CAAS;IAC9C,OAAO,CAAC,QAAQ,CAAC,aAAa,CAAS;IACvC,OAAO,CAAC,QAAQ,CAAC,MAAM,CAAc;IACrC,SAAgB,EAAE,EAAG,OAAO,CAAU;gBAE1B,MAAM,EAAE,eAAe,EAAE,MAAM,EAAE,IAAI,CAAC,MAAM;IAOjD,aAAa,CAAC,MAAM,EAAE;QAC3B,MAAM,EAAE,IAAI,CAAC,MAAM,CAAC;QACpB,QAAQ,CAAC,EAAE,MAAM,CAAC;QAClB,MAAM,CAAC,EAAE,MAAM,CAAC;KACjB,GAAG,6BAA6B;IAmEpB,eAAe,CAAC,WAAW,EAAE,MAAM,EAAE,MAAM,EAAE,MAAM,GAAG,OAAO,CAAC,mBAAmB,CAAC;CAoEhG"}
|
|
@@ -0,0 +1,140 @@
|
|
|
1
|
+
import { EventEmitter } from "node:events";
|
|
2
|
+
import { v4 as uuidv4 } from "uuid";
|
|
3
|
+
import { Pcm16MonoResampler } from "../../../../agent/pcm16-resampler.js";
|
|
4
|
+
import { parsePcm16MonoWav, parsePcmRateFromFormat, pcm16lePeakAbs, pcm16leToFloat32 } from "../../../audio.js";
|
|
5
|
+
export class SherpaOnnxSTT {
|
|
6
|
+
constructor(config, logger) {
|
|
7
|
+
this.id = "local";
|
|
8
|
+
this.engine = config.engine;
|
|
9
|
+
this.silencePeakThreshold = config.silencePeakThreshold ?? 300;
|
|
10
|
+
this.tailPaddingMs = config.tailPaddingMs ?? 500;
|
|
11
|
+
this.logger = logger.child({ module: "speech", provider: "local", component: "stt" });
|
|
12
|
+
}
|
|
13
|
+
createSession(params) {
|
|
14
|
+
const emitter = new EventEmitter();
|
|
15
|
+
void params;
|
|
16
|
+
const requiredSampleRate = this.engine.sampleRate;
|
|
17
|
+
let connected = false;
|
|
18
|
+
let segmentId = uuidv4();
|
|
19
|
+
let previousSegmentId = null;
|
|
20
|
+
let pcm16 = Buffer.alloc(0);
|
|
21
|
+
return {
|
|
22
|
+
requiredSampleRate,
|
|
23
|
+
async connect() {
|
|
24
|
+
connected = true;
|
|
25
|
+
},
|
|
26
|
+
appendPcm16(chunk) {
|
|
27
|
+
if (!connected) {
|
|
28
|
+
emitter.emit("error", new Error("STT session not connected"));
|
|
29
|
+
return;
|
|
30
|
+
}
|
|
31
|
+
pcm16 = pcm16.length === 0 ? chunk : Buffer.concat([pcm16, chunk]);
|
|
32
|
+
},
|
|
33
|
+
commit: () => {
|
|
34
|
+
if (!connected) {
|
|
35
|
+
emitter.emit("error", new Error("STT session not connected"));
|
|
36
|
+
return;
|
|
37
|
+
}
|
|
38
|
+
const committedId = segmentId;
|
|
39
|
+
const prev = previousSegmentId;
|
|
40
|
+
emitter.emit("committed", { segmentId: committedId, previousSegmentId: prev });
|
|
41
|
+
void (async () => {
|
|
42
|
+
try {
|
|
43
|
+
const rt = await this.transcribeAudio(pcm16, `audio/pcm;rate=${requiredSampleRate}`);
|
|
44
|
+
emitter.emit("transcript", {
|
|
45
|
+
segmentId: committedId,
|
|
46
|
+
transcript: rt.text,
|
|
47
|
+
isFinal: true,
|
|
48
|
+
language: rt.language,
|
|
49
|
+
logprobs: rt.logprobs,
|
|
50
|
+
avgLogprob: rt.avgLogprob,
|
|
51
|
+
isLowConfidence: rt.isLowConfidence,
|
|
52
|
+
});
|
|
53
|
+
}
|
|
54
|
+
catch (err) {
|
|
55
|
+
emitter.emit("error", err);
|
|
56
|
+
}
|
|
57
|
+
finally {
|
|
58
|
+
previousSegmentId = committedId;
|
|
59
|
+
segmentId = uuidv4();
|
|
60
|
+
pcm16 = Buffer.alloc(0);
|
|
61
|
+
}
|
|
62
|
+
})();
|
|
63
|
+
},
|
|
64
|
+
clear() {
|
|
65
|
+
pcm16 = Buffer.alloc(0);
|
|
66
|
+
segmentId = uuidv4();
|
|
67
|
+
},
|
|
68
|
+
close() {
|
|
69
|
+
connected = false;
|
|
70
|
+
pcm16 = Buffer.alloc(0);
|
|
71
|
+
},
|
|
72
|
+
on(event, handler) {
|
|
73
|
+
emitter.on(event, handler);
|
|
74
|
+
return undefined;
|
|
75
|
+
},
|
|
76
|
+
};
|
|
77
|
+
}
|
|
78
|
+
async transcribeAudio(audioBuffer, format) {
|
|
79
|
+
const start = Date.now();
|
|
80
|
+
let inputRate;
|
|
81
|
+
let pcm16;
|
|
82
|
+
if (format.toLowerCase().includes("audio/wav")) {
|
|
83
|
+
const parsed = parsePcm16MonoWav(audioBuffer);
|
|
84
|
+
inputRate = parsed.sampleRate;
|
|
85
|
+
pcm16 = parsed.pcm16;
|
|
86
|
+
}
|
|
87
|
+
else if (format.toLowerCase().includes("audio/pcm")) {
|
|
88
|
+
inputRate = parsePcmRateFromFormat(format, this.engine.sampleRate) ?? this.engine.sampleRate;
|
|
89
|
+
pcm16 = audioBuffer;
|
|
90
|
+
}
|
|
91
|
+
else {
|
|
92
|
+
throw new Error(`Unsupported audio format for sherpa STT: ${format}`);
|
|
93
|
+
}
|
|
94
|
+
const peak = pcm16lePeakAbs(pcm16);
|
|
95
|
+
if (peak < this.silencePeakThreshold) {
|
|
96
|
+
return { text: "", duration: Date.now() - start, isLowConfidence: true };
|
|
97
|
+
}
|
|
98
|
+
let pcmForModel = pcm16;
|
|
99
|
+
if (inputRate !== this.engine.sampleRate) {
|
|
100
|
+
const resampler = new Pcm16MonoResampler({ inputRate, outputRate: this.engine.sampleRate });
|
|
101
|
+
pcmForModel = resampler.processChunk(pcm16);
|
|
102
|
+
inputRate = this.engine.sampleRate;
|
|
103
|
+
}
|
|
104
|
+
const peakForModel = pcm16lePeakAbs(pcmForModel);
|
|
105
|
+
const peakFloat = peakForModel / 32768.0;
|
|
106
|
+
const targetPeak = 0.6;
|
|
107
|
+
const maxGain = 50;
|
|
108
|
+
const gain = peakFloat > 0 && peakFloat < targetPeak
|
|
109
|
+
? Math.min(maxGain, targetPeak / peakFloat)
|
|
110
|
+
: 1;
|
|
111
|
+
const stream = this.engine.createStream();
|
|
112
|
+
try {
|
|
113
|
+
const floatSamples = pcm16leToFloat32(pcmForModel, gain);
|
|
114
|
+
stream.acceptWaveform(inputRate, floatSamples);
|
|
115
|
+
while (this.engine.recognizer.isReady(stream)) {
|
|
116
|
+
this.engine.recognizer.decode(stream);
|
|
117
|
+
}
|
|
118
|
+
const padSamples = Math.max(0, Math.round((this.engine.sampleRate * this.tailPaddingMs) / 1000));
|
|
119
|
+
if (padSamples > 0) {
|
|
120
|
+
stream.acceptWaveform(this.engine.sampleRate, new Float32Array(padSamples));
|
|
121
|
+
}
|
|
122
|
+
while (this.engine.recognizer.isReady(stream)) {
|
|
123
|
+
this.engine.recognizer.decode(stream);
|
|
124
|
+
}
|
|
125
|
+
const text = String(this.engine.recognizer.getResult(stream)?.text ?? "").trim();
|
|
126
|
+
const duration = Date.now() - start;
|
|
127
|
+
this.logger.debug({ duration, textLength: text.length }, "Sherpa transcription complete");
|
|
128
|
+
return { text, duration, ...(text.length === 0 ? { isLowConfidence: true } : {}) };
|
|
129
|
+
}
|
|
130
|
+
finally {
|
|
131
|
+
try {
|
|
132
|
+
stream.free?.();
|
|
133
|
+
}
|
|
134
|
+
catch {
|
|
135
|
+
// ignore
|
|
136
|
+
}
|
|
137
|
+
}
|
|
138
|
+
}
|
|
139
|
+
}
|
|
140
|
+
//# sourceMappingURL=sherpa-stt.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"sherpa-stt.js","sourceRoot":"","sources":["../../../../../../../src/server/speech/providers/local/sherpa/sherpa-stt.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAC3C,OAAO,EAAE,EAAE,IAAI,MAAM,EAAE,MAAM,MAAM,CAAC;AAQpC,OAAO,EAAE,kBAAkB,EAAE,MAAM,sCAAsC,CAAC;AAC1E,OAAO,EAAE,iBAAiB,EAAE,sBAAsB,EAAE,cAAc,EAAE,gBAAgB,EAAE,MAAM,mBAAmB,CAAC;AAShH,MAAM,OAAO,aAAa;IAOxB,YAAY,MAAuB,EAAE,MAAmB;QAFxC,OAAE,GAAG,OAAgB,CAAC;QAGpC,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC,MAAM,CAAC;QAC5B,IAAI,CAAC,oBAAoB,GAAG,MAAM,CAAC,oBAAoB,IAAI,GAAG,CAAC;QAC/D,IAAI,CAAC,aAAa,GAAG,MAAM,CAAC,aAAa,IAAI,GAAG,CAAC;QACjD,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC,KAAK,CAAC,EAAE,MAAM,EAAE,QAAQ,EAAE,QAAQ,EAAE,OAAO,EAAE,SAAS,EAAE,KAAK,EAAE,CAAC,CAAC;IACxF,CAAC;IAEM,aAAa,CAAC,MAIpB;QACC,MAAM,OAAO,GAAG,IAAI,YAAY,EAAE,CAAC;QACnC,KAAK,MAAM,CAAC;QACZ,MAAM,kBAAkB,GAAG,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC;QAClD,IAAI,SAAS,GAAG,KAAK,CAAC;QACtB,IAAI,SAAS,GAAG,MAAM,EAAE,CAAC;QACzB,IAAI,iBAAiB,GAAkB,IAAI,CAAC;QAC5C,IAAI,KAAK,GAAW,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;QAEpC,OAAO;YACL,kBAAkB;YAClB,KAAK,CAAC,OAAO;gBACX,SAAS,GAAG,IAAI,CAAC;YACnB,CAAC;YACD,WAAW,CAAC,KAAa;gBACvB,IAAI,CAAC,SAAS,EAAE,CAAC;oBACd,OAAe,CAAC,IAAI,CAAC,OAAO,EAAE,IAAI,KAAK,CAAC,2BAA2B,CAAC,CAAC,CAAC;oBACvE,OAAO;gBACT,CAAC;gBACD,KAAK,GAAG,KAAK,CAAC,MAAM,KAAK,CAAC,CAAC,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,MAAM,CAAC,MAAM,CAAC,CAAC,KAAK,EAAE,KAAK,CAAC,CAAC,CAAC;YACrE,CAAC;YACD,MAAM,EAAE,GAAG,EAAE;gBACX,IAAI,CAAC,SAAS,EAAE,CAAC;oBACd,OAAe,CAAC,IAAI,CAAC,OAAO,EAAE,IAAI,KAAK,CAAC,2BAA2B,CAAC,CAAC,CAAC;oBACvE,OAAO;gBACT,CAAC;gBAED,MAAM,WAAW,GAAG,SAAS,CAAC;gBAC9B,MAAM,IAAI,GAAG,iBAAiB,CAAC;gBAC9B,OAAe,CAAC,IAAI,CAAC,WAAW,EAAE,EAAE,SAAS,EAAE,WAAW,EAAE,iBAAiB,EAAE,IAAI,EAAE,CAAC,CAAC;gBAExF,KAAK,CAAC,KAAK,IAAI,EAAE;oBACf,IAAI,CAAC;wBACH,MAAM,EAAE,GAAG,MAAM,IAAI,CAAC,eAAe,CAAC,KAAK,EAAE,kBAAkB,kBAAkB,EAAE,CAAC,CAAC;wBACpF,OAAe,CAAC,IAAI,CAAC,YAAY,EAAE;4BAClC,SAAS,EAAE,WAAW;4BACtB,UAAU,EAAE,EAAE,CAAC,IAAI;4BACnB,OAAO,EAAE,IAAI;4BACb,QAAQ,EAAE,EAAE,CAAC,QAAQ;4BACrB,QAAQ,EAAE,EAAE,CAAC,QAAQ;4BACrB,UAAU,EAAE,EAAE,CAAC,UAAU;4BACzB,eAAe,EAAE,EAAE,CAAC,eAAe;yBACpC,CAAC,CAAC;oBACL,CAAC;oBAAC,OAAO,GAAG,EAAE,CAAC;wBACZ,OAAe,CAAC,IAAI,CAAC,OAAO,EAAE,GAAG,CAAC,CAAC;oBACtC,CAAC;4BAAS,CAAC;wBACT,iBAAiB,GAAG,WAAW,CAAC;wBAChC,SAAS,GAAG,MAAM,EAAE,CAAC;wBACrB,KAAK,GAAG,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;oBAC1B,CAAC;gBACH,CAAC,CAAC,EAAE,CAAC;YACP,CAAC;YACD,KAAK;gBACH,KAAK,GAAG,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;gBACxB,SAAS,GAAG,MAAM,EAAE,CAAC;YACvB,CAAC;YACD,KAAK;gBACH,SAAS,GAAG,KAAK,CAAC;gBAClB,KAAK,GAAG,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;YAC1B,CAAC;YACD,EAAE,CAAC,KAAU,EAAE,OAAY;gBACzB,OAAO,CAAC,EAAE,CAAC,KAAK,EAAE,OAAO,CAAC,CAAC;gBAC3B,OAAO,SAAS,CAAC;YACnB,CAAC;SACF,CAAC;IACJ,CAAC;IAEM,KAAK,CAAC,eAAe,CAAC,WAAmB,EAAE,MAAc;QAC9D,MAAM,KAAK,GAAG,IAAI,CAAC,GAAG,EAAE,CAAC;QAEzB,IAAI,SAAiB,CAAC;QACtB,IAAI,KAAa,CAAC;QAElB,IAAI,MAAM,CAAC,WAAW,EAAE,CAAC,QAAQ,CAAC,WAAW,CAAC,EAAE,CAAC;YAC/C,MAAM,MAAM,GAAG,iBAAiB,CAAC,WAAW,CAAC,CAAC;YAC9C,SAAS,GAAG,MAAM,CAAC,UAAU,CAAC;YAC9B,KAAK,GAAG,MAAM,CAAC,KAAK,CAAC;QACvB,CAAC;aAAM,IAAI,MAAM,CAAC,WAAW,EAAE,CAAC,QAAQ,CAAC,WAAW,CAAC,EAAE,CAAC;YACtD,SAAS,GAAG,sBAAsB,CAAC,MAAM,EAAE,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC,IAAI,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC;YAC7F,KAAK,GAAG,WAAW,CAAC;QACtB,CAAC;aAAM,CAAC;YACN,MAAM,IAAI,KAAK,CAAC,4CAA4C,MAAM,EAAE,CAAC,CAAC;QACxE,CAAC;QAED,MAAM,IAAI,GAAG,cAAc,CAAC,KAAK,CAAC,CAAC;QACnC,IAAI,IAAI,GAAG,IAAI,CAAC,oBAAoB,EAAE,CAAC;YACrC,OAAO,EAAE,IAAI,EAAE,EAAE,EAAE,QAAQ,EAAE,IAAI,CAAC,GAAG,EAAE,GAAG,KAAK,EAAE,eAAe,EAAE,IAAI,EAAE,CAAC;QAC3E,CAAC;QAED,IAAI,WAAW,GAAG,KAAK,CAAC;QACxB,IAAI,SAAS,KAAK,IAAI,CAAC,MAAM,CAAC,UAAU,EAAE,CAAC;YACzC,MAAM,SAAS,GAAG,IAAI,kBAAkB,CAAC,EAAE,SAAS,EAAE,UAAU,EAAE,IAAI,CAAC,MAAM,CAAC,UAAU,EAAE,CAAC,CAAC;YAC5F,WAAW,GAAG,SAAS,CAAC,YAAY,CAAC,KAAK,CAAC,CAAC;YAC5C,SAAS,GAAG,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC;QACrC,CAAC;QAED,MAAM,YAAY,GAAG,cAAc,CAAC,WAAW,CAAC,CAAC;QACjD,MAAM,SAAS,GAAG,YAAY,GAAG,OAAO,CAAC;QACzC,MAAM,UAAU,GAAG,GAAG,CAAC;QACvB,MAAM,OAAO,GAAG,EAAE,CAAC;QACnB,MAAM,IAAI,GACR,SAAS,GAAG,CAAC,IAAI,SAAS,GAAG,UAAU;YACrC,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,OAAO,EAAE,UAAU,GAAG,SAAS,CAAC;YAC3C,CAAC,CAAC,CAAC,CAAC;QAER,MAAM,MAAM,GAAG,IAAI,CAAC,MAAM,CAAC,YAAY,EAAE,CAAC;QAC1C,IAAI,CAAC;YACH,MAAM,YAAY,GAAG,gBAAgB,CAAC,WAAW,EAAE,IAAI,CAAC,CAAC;YACzD,MAAM,CAAC,cAAc,CAAC,SAAS,EAAE,YAAY,CAAC,CAAC;YAE/C,OAAO,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC,OAAO,CAAC,MAAM,CAAC,EAAE,CAAC;gBAC9C,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC,MAAM,CAAC,MAAM,CAAC,CAAC;YACxC,CAAC;YAED,MAAM,UAAU,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,CAAC,KAAK,CAAC,CAAC,IAAI,CAAC,MAAM,CAAC,UAAU,GAAG,IAAI,CAAC,aAAa,CAAC,GAAG,IAAI,CAAC,CAAC,CAAC;YACjG,IAAI,UAAU,GAAG,CAAC,EAAE,CAAC;gBACnB,MAAM,CAAC,cAAc,CAAC,IAAI,CAAC,MAAM,CAAC,UAAU,EAAE,IAAI,YAAY,CAAC,UAAU,CAAC,CAAC,CAAC;YAC9E,CAAC;YAED,OAAO,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC,OAAO,CAAC,MAAM,CAAC,EAAE,CAAC;gBAC9C,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC,MAAM,CAAC,MAAM,CAAC,CAAC;YACxC,CAAC;YAED,MAAM,IAAI,GAAG,MAAM,CAAC,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC,SAAS,CAAC,MAAM,CAAC,EAAE,IAAI,IAAI,EAAE,CAAC,CAAC,IAAI,EAAE,CAAC;YACjF,MAAM,QAAQ,GAAG,IAAI,CAAC,GAAG,EAAE,GAAG,KAAK,CAAC;YACpC,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,EAAE,QAAQ,EAAE,UAAU,EAAE,IAAI,CAAC,MAAM,EAAE,EAAE,+BAA+B,CAAC,CAAC;YAC1F,OAAO,EAAE,IAAI,EAAE,QAAQ,EAAE,GAAG,CAAC,IAAI,CAAC,MAAM,KAAK,CAAC,CAAC,CAAC,CAAC,EAAE,eAAe,EAAE,IAAI,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,CAAC;QACrF,CAAC;gBAAS,CAAC;YACT,IAAI,CAAC;gBACH,MAAM,CAAC,IAAI,EAAE,EAAE,CAAC;YAClB,CAAC;YAAC,MAAM,CAAC;gBACP,SAAS;YACX,CAAC;QACH,CAAC;IACH,CAAC;CACF"}
|