@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 +1,21 @@
|
|
|
1
|
+
import type pino from "pino";
|
|
2
|
+
import type { SpeechStreamResult, TextToSpeechProvider } from "../../../speech-provider.js";
|
|
3
|
+
export type SherpaTtsPreset = "kokoro-en-v0_19" | "kitten-nano-en-v0_1-fp16";
|
|
4
|
+
export type SherpaTtsConfig = {
|
|
5
|
+
preset: SherpaTtsPreset;
|
|
6
|
+
modelDir: string;
|
|
7
|
+
speakerId?: number;
|
|
8
|
+
speed?: number;
|
|
9
|
+
lengthScale?: number;
|
|
10
|
+
numThreads?: number;
|
|
11
|
+
};
|
|
12
|
+
export declare class SherpaOnnxTTS implements TextToSpeechProvider {
|
|
13
|
+
private readonly tts;
|
|
14
|
+
private readonly speakerId;
|
|
15
|
+
private readonly speed;
|
|
16
|
+
private readonly logger;
|
|
17
|
+
constructor(config: SherpaTtsConfig, logger: pino.Logger);
|
|
18
|
+
synthesizeSpeech(text: string): Promise<SpeechStreamResult>;
|
|
19
|
+
free(): void;
|
|
20
|
+
}
|
|
21
|
+
//# sourceMappingURL=sherpa-tts.d.ts.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"sherpa-tts.d.ts","sourceRoot":"","sources":["../../../../../../../src/server/speech/providers/local/sherpa/sherpa-tts.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,IAAI,MAAM,MAAM,CAAC;AAI7B,OAAO,KAAK,EAAE,kBAAkB,EAAE,oBAAoB,EAAE,MAAM,6BAA6B,CAAC;AAI5F,MAAM,MAAM,eAAe,GAAG,iBAAiB,GAAG,0BAA0B,CAAC;AAE7E,MAAM,MAAM,eAAe,GAAG;IAC5B,MAAM,EAAE,eAAe,CAAC;IACxB,QAAQ,EAAE,MAAM,CAAC;IACjB,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,WAAW,CAAC,EAAE,MAAM,CAAC;IACrB,UAAU,CAAC,EAAE,MAAM,CAAC;CACrB,CAAC;AAQF,qBAAa,aAAc,YAAW,oBAAoB;IACxD,OAAO,CAAC,QAAQ,CAAC,GAAG,CAAM;IAC1B,OAAO,CAAC,QAAQ,CAAC,SAAS,CAAS;IACnC,OAAO,CAAC,QAAQ,CAAC,KAAK,CAAS;IAC/B,OAAO,CAAC,QAAQ,CAAC,MAAM,CAAc;gBAEzB,MAAM,EAAE,eAAe,EAAE,MAAM,EAAE,IAAI,CAAC,MAAM;IAwDlD,gBAAgB,CAAC,IAAI,EAAE,MAAM,GAAG,OAAO,CAAC,kBAAkB,CAAC;IAkCjE,IAAI,IAAI,IAAI;CAOb"}
|
|
@@ -0,0 +1,95 @@
|
|
|
1
|
+
import { Readable } from "node:stream";
|
|
2
|
+
import { existsSync } from "node:fs";
|
|
3
|
+
import { chunkBuffer, float32ToPcm16le } from "../../../audio.js";
|
|
4
|
+
import { loadSherpaOnnxNode } from "./sherpa-onnx-node-loader.js";
|
|
5
|
+
function assertFileExists(filePath, label) {
|
|
6
|
+
if (!existsSync(filePath)) {
|
|
7
|
+
throw new Error(`Missing ${label}: ${filePath}`);
|
|
8
|
+
}
|
|
9
|
+
}
|
|
10
|
+
export class SherpaOnnxTTS {
|
|
11
|
+
constructor(config, logger) {
|
|
12
|
+
if (config.preset !== "kokoro-en-v0_19" && config.preset !== "kitten-nano-en-v0_1-fp16") {
|
|
13
|
+
throw new Error(`Unsupported Sherpa TTS preset: ${config.preset}`);
|
|
14
|
+
}
|
|
15
|
+
this.logger = logger.child({ module: "speech", provider: "local", component: "tts" });
|
|
16
|
+
this.speakerId = config.speakerId ?? 0;
|
|
17
|
+
this.speed = config.speed ?? 1.0;
|
|
18
|
+
const sherpa = loadSherpaOnnxNode();
|
|
19
|
+
if (typeof sherpa.OfflineTts !== "function") {
|
|
20
|
+
throw new Error("sherpa-onnx-node OfflineTts is unavailable");
|
|
21
|
+
}
|
|
22
|
+
const modelFile = config.preset === "kokoro-en-v0_19" ? "model.onnx" : "model.fp16.onnx";
|
|
23
|
+
const modelPath = `${config.modelDir}/${modelFile}`;
|
|
24
|
+
const voicesPath = `${config.modelDir}/voices.bin`;
|
|
25
|
+
const tokensPath = `${config.modelDir}/tokens.txt`;
|
|
26
|
+
const dataDir = `${config.modelDir}/espeak-ng-data`;
|
|
27
|
+
assertFileExists(modelPath, "TTS model");
|
|
28
|
+
assertFileExists(voicesPath, "TTS voices");
|
|
29
|
+
assertFileExists(tokensPath, "TTS tokens");
|
|
30
|
+
assertFileExists(dataDir, "TTS espeak-ng dataDir");
|
|
31
|
+
const modelConfig = config.preset === "kokoro-en-v0_19"
|
|
32
|
+
? {
|
|
33
|
+
kokoro: {
|
|
34
|
+
model: modelPath,
|
|
35
|
+
voices: voicesPath,
|
|
36
|
+
tokens: tokensPath,
|
|
37
|
+
dataDir,
|
|
38
|
+
lengthScale: config.lengthScale ?? 1.0,
|
|
39
|
+
},
|
|
40
|
+
}
|
|
41
|
+
: {
|
|
42
|
+
kitten: {
|
|
43
|
+
model: modelPath,
|
|
44
|
+
voices: voicesPath,
|
|
45
|
+
tokens: tokensPath,
|
|
46
|
+
dataDir,
|
|
47
|
+
lengthScale: config.lengthScale ?? 1.0,
|
|
48
|
+
},
|
|
49
|
+
};
|
|
50
|
+
const offlineTtsConfig = {
|
|
51
|
+
model: modelConfig,
|
|
52
|
+
numThreads: config.numThreads ?? 2,
|
|
53
|
+
provider: "cpu",
|
|
54
|
+
maxNumSentences: 1,
|
|
55
|
+
};
|
|
56
|
+
this.tts = new sherpa.OfflineTts(offlineTtsConfig);
|
|
57
|
+
this.logger.info({ preset: config.preset, modelDir: config.modelDir }, "Sherpa offline TTS initialized");
|
|
58
|
+
}
|
|
59
|
+
async synthesizeSpeech(text) {
|
|
60
|
+
const trimmed = text.trim();
|
|
61
|
+
if (!trimmed) {
|
|
62
|
+
throw new Error("Cannot synthesize empty text");
|
|
63
|
+
}
|
|
64
|
+
const audio = this.tts.generate({ text: trimmed, sid: this.speakerId, speed: this.speed });
|
|
65
|
+
const samples = audio && audio.samples instanceof Float32Array
|
|
66
|
+
? audio.samples
|
|
67
|
+
: audio && Array.isArray(audio.samples)
|
|
68
|
+
? Float32Array.from(audio.samples)
|
|
69
|
+
: null;
|
|
70
|
+
const sampleRate = audio && typeof audio.sampleRate === "number" && Number.isFinite(audio.sampleRate) && audio.sampleRate > 0
|
|
71
|
+
? audio.sampleRate
|
|
72
|
+
: typeof this.tts.sampleRate === "number"
|
|
73
|
+
? this.tts.sampleRate
|
|
74
|
+
: 24000;
|
|
75
|
+
if (!samples) {
|
|
76
|
+
throw new Error("Unexpected sherpa TTS output: missing Float32 samples");
|
|
77
|
+
}
|
|
78
|
+
const pcm16 = float32ToPcm16le(samples);
|
|
79
|
+
const chunkBytes = Math.max(2, Math.round(sampleRate * 0.05) * 2); // ~50ms
|
|
80
|
+
const chunks = chunkBuffer(pcm16, chunkBytes);
|
|
81
|
+
return {
|
|
82
|
+
stream: Readable.from(chunks),
|
|
83
|
+
format: `pcm;rate=${sampleRate}`,
|
|
84
|
+
};
|
|
85
|
+
}
|
|
86
|
+
free() {
|
|
87
|
+
try {
|
|
88
|
+
this.tts?.free?.();
|
|
89
|
+
}
|
|
90
|
+
catch {
|
|
91
|
+
// ignore
|
|
92
|
+
}
|
|
93
|
+
}
|
|
94
|
+
}
|
|
95
|
+
//# sourceMappingURL=sherpa-tts.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"sherpa-tts.js","sourceRoot":"","sources":["../../../../../../../src/server/speech/providers/local/sherpa/sherpa-tts.ts"],"names":[],"mappings":"AACA,OAAO,EAAE,QAAQ,EAAE,MAAM,aAAa,CAAC;AACvC,OAAO,EAAE,UAAU,EAAE,MAAM,SAAS,CAAC;AAGrC,OAAO,EAAE,WAAW,EAAE,gBAAgB,EAAE,MAAM,mBAAmB,CAAC;AAClE,OAAO,EAAE,kBAAkB,EAAE,MAAM,8BAA8B,CAAC;AAalE,SAAS,gBAAgB,CAAC,QAAgB,EAAE,KAAa;IACvD,IAAI,CAAC,UAAU,CAAC,QAAQ,CAAC,EAAE,CAAC;QAC1B,MAAM,IAAI,KAAK,CAAC,WAAW,KAAK,KAAK,QAAQ,EAAE,CAAC,CAAC;IACnD,CAAC;AACH,CAAC;AAED,MAAM,OAAO,aAAa;IAMxB,YAAY,MAAuB,EAAE,MAAmB;QACtD,IAAI,MAAM,CAAC,MAAM,KAAK,iBAAiB,IAAI,MAAM,CAAC,MAAM,KAAK,0BAA0B,EAAE,CAAC;YACxF,MAAM,IAAI,KAAK,CAAC,kCAAkC,MAAM,CAAC,MAAM,EAAE,CAAC,CAAC;QACrE,CAAC;QACD,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC,KAAK,CAAC,EAAE,MAAM,EAAE,QAAQ,EAAE,QAAQ,EAAE,OAAO,EAAE,SAAS,EAAE,KAAK,EAAE,CAAC,CAAC;QACtF,IAAI,CAAC,SAAS,GAAG,MAAM,CAAC,SAAS,IAAI,CAAC,CAAC;QACvC,IAAI,CAAC,KAAK,GAAG,MAAM,CAAC,KAAK,IAAI,GAAG,CAAC;QAEjC,MAAM,MAAM,GAAG,kBAAkB,EAAE,CAAC;QACpC,IAAI,OAAO,MAAM,CAAC,UAAU,KAAK,UAAU,EAAE,CAAC;YAC5C,MAAM,IAAI,KAAK,CAAC,4CAA4C,CAAC,CAAC;QAChE,CAAC;QAED,MAAM,SAAS,GAAG,MAAM,CAAC,MAAM,KAAK,iBAAiB,CAAC,CAAC,CAAC,YAAY,CAAC,CAAC,CAAC,iBAAiB,CAAC;QACzF,MAAM,SAAS,GAAG,GAAG,MAAM,CAAC,QAAQ,IAAI,SAAS,EAAE,CAAC;QACpD,MAAM,UAAU,GAAG,GAAG,MAAM,CAAC,QAAQ,aAAa,CAAC;QACnD,MAAM,UAAU,GAAG,GAAG,MAAM,CAAC,QAAQ,aAAa,CAAC;QACnD,MAAM,OAAO,GAAG,GAAG,MAAM,CAAC,QAAQ,iBAAiB,CAAC;QAEpD,gBAAgB,CAAC,SAAS,EAAE,WAAW,CAAC,CAAC;QACzC,gBAAgB,CAAC,UAAU,EAAE,YAAY,CAAC,CAAC;QAC3C,gBAAgB,CAAC,UAAU,EAAE,YAAY,CAAC,CAAC;QAC3C,gBAAgB,CAAC,OAAO,EAAE,uBAAuB,CAAC,CAAC;QAEnD,MAAM,WAAW,GACf,MAAM,CAAC,MAAM,KAAK,iBAAiB;YACjC,CAAC,CAAC;gBACE,MAAM,EAAE;oBACN,KAAK,EAAE,SAAS;oBAChB,MAAM,EAAE,UAAU;oBAClB,MAAM,EAAE,UAAU;oBAClB,OAAO;oBACP,WAAW,EAAE,MAAM,CAAC,WAAW,IAAI,GAAG;iBACvC;aACF;YACH,CAAC,CAAC;gBACE,MAAM,EAAE;oBACN,KAAK,EAAE,SAAS;oBAChB,MAAM,EAAE,UAAU;oBAClB,MAAM,EAAE,UAAU;oBAClB,OAAO;oBACP,WAAW,EAAE,MAAM,CAAC,WAAW,IAAI,GAAG;iBACvC;aACF,CAAC;QAER,MAAM,gBAAgB,GAAG;YACvB,KAAK,EAAE,WAAW;YAClB,UAAU,EAAE,MAAM,CAAC,UAAU,IAAI,CAAC;YAClC,QAAQ,EAAE,KAAK;YACf,eAAe,EAAE,CAAC;SACnB,CAAC;QAEF,IAAI,CAAC,GAAG,GAAG,IAAI,MAAM,CAAC,UAAU,CAAC,gBAAgB,CAAC,CAAC;QACnD,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,EAAE,MAAM,EAAE,MAAM,CAAC,MAAM,EAAE,QAAQ,EAAE,MAAM,CAAC,QAAQ,EAAE,EAAE,gCAAgC,CAAC,CAAC;IAC3G,CAAC;IAED,KAAK,CAAC,gBAAgB,CAAC,IAAY;QACjC,MAAM,OAAO,GAAG,IAAI,CAAC,IAAI,EAAE,CAAC;QAC5B,IAAI,CAAC,OAAO,EAAE,CAAC;YACb,MAAM,IAAI,KAAK,CAAC,8BAA8B,CAAC,CAAC;QAClD,CAAC;QAED,MAAM,KAAK,GAAG,IAAI,CAAC,GAAG,CAAC,QAAQ,CAAC,EAAE,IAAI,EAAE,OAAO,EAAE,GAAG,EAAE,IAAI,CAAC,SAAS,EAAE,KAAK,EAAE,IAAI,CAAC,KAAK,EAAE,CAAC,CAAC;QAC3F,MAAM,OAAO,GACX,KAAK,IAAI,KAAK,CAAC,OAAO,YAAY,YAAY;YAC5C,CAAC,CAAC,KAAK,CAAC,OAAO;YACf,CAAC,CAAC,KAAK,IAAI,KAAK,CAAC,OAAO,CAAC,KAAK,CAAC,OAAO,CAAC;gBACrC,CAAC,CAAC,YAAY,CAAC,IAAI,CAAC,KAAK,CAAC,OAAmB,CAAC;gBAC9C,CAAC,CAAC,IAAI,CAAC;QACb,MAAM,UAAU,GACd,KAAK,IAAI,OAAO,KAAK,CAAC,UAAU,KAAK,QAAQ,IAAI,MAAM,CAAC,QAAQ,CAAC,KAAK,CAAC,UAAU,CAAC,IAAI,KAAK,CAAC,UAAU,GAAG,CAAC;YACxG,CAAC,CAAC,KAAK,CAAC,UAAU;YAClB,CAAC,CAAC,OAAO,IAAI,CAAC,GAAG,CAAC,UAAU,KAAK,QAAQ;gBACvC,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,UAAU;gBACrB,CAAC,CAAC,KAAK,CAAC;QAEd,IAAI,CAAC,OAAO,EAAE,CAAC;YACb,MAAM,IAAI,KAAK,CAAC,uDAAuD,CAAC,CAAC;QAC3E,CAAC;QAED,MAAM,KAAK,GAAG,gBAAgB,CAAC,OAAO,CAAC,CAAC;QACxC,MAAM,UAAU,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,CAAC,KAAK,CAAC,UAAU,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,QAAQ;QAC3E,MAAM,MAAM,GAAG,WAAW,CAAC,KAAK,EAAE,UAAU,CAAC,CAAC;QAE9C,OAAO;YACL,MAAM,EAAE,QAAQ,CAAC,IAAI,CAAC,MAAM,CAAC;YAC7B,MAAM,EAAE,YAAY,UAAU,EAAE;SACjC,CAAC;IACJ,CAAC;IAED,IAAI;QACF,IAAI,CAAC;YACH,IAAI,CAAC,GAAG,EAAE,IAAI,EAAE,EAAE,CAAC;QACrB,CAAC;QAAC,MAAM,CAAC;YACP,SAAS;QACX,CAAC;IACH,CAAC;CACF"}
|
|
@@ -0,0 +1,22 @@
|
|
|
1
|
+
import type { PersistedConfig } from "../../../persisted-config.js";
|
|
2
|
+
import type { RequestedSpeechProviders } from "../../speech-types.js";
|
|
3
|
+
import type { STTConfig } from "./stt.js";
|
|
4
|
+
import type { TTSConfig } from "./tts.js";
|
|
5
|
+
export declare const DEFAULT_OPENAI_REALTIME_TRANSCRIPTION_MODEL = "gpt-4o-transcribe";
|
|
6
|
+
export declare const DEFAULT_OPENAI_TTS_MODEL = "tts-1";
|
|
7
|
+
export type OpenAiSpeechProviderConfig = {
|
|
8
|
+
apiKey?: string;
|
|
9
|
+
stt?: Partial<STTConfig> & {
|
|
10
|
+
apiKey?: string;
|
|
11
|
+
};
|
|
12
|
+
tts?: Partial<TTSConfig> & {
|
|
13
|
+
apiKey?: string;
|
|
14
|
+
};
|
|
15
|
+
realtimeTranscriptionModel?: string;
|
|
16
|
+
};
|
|
17
|
+
export declare function resolveOpenAiSpeechConfig(params: {
|
|
18
|
+
env: NodeJS.ProcessEnv;
|
|
19
|
+
persisted: PersistedConfig;
|
|
20
|
+
providers: RequestedSpeechProviders;
|
|
21
|
+
}): OpenAiSpeechProviderConfig | undefined;
|
|
22
|
+
//# sourceMappingURL=config.d.ts.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"config.d.ts","sourceRoot":"","sources":["../../../../../../src/server/speech/providers/openai/config.ts"],"names":[],"mappings":"AAEA,OAAO,KAAK,EAAE,eAAe,EAAE,MAAM,8BAA8B,CAAC;AACpE,OAAO,KAAK,EAAE,wBAAwB,EAAE,MAAM,uBAAuB,CAAC;AACtE,OAAO,KAAK,EAAE,SAAS,EAAE,MAAM,UAAU,CAAC;AAC1C,OAAO,KAAK,EAAE,SAAS,EAAE,MAAM,UAAU,CAAC;AAE1C,eAAO,MAAM,2CAA2C,sBAAsB,CAAC;AAC/E,eAAO,MAAM,wBAAwB,UAAU,CAAC;AAEhD,MAAM,MAAM,0BAA0B,GAAG;IACvC,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,GAAG,CAAC,EAAE,OAAO,CAAC,SAAS,CAAC,GAAG;QAAE,MAAM,CAAC,EAAE,MAAM,CAAA;KAAE,CAAC;IAC/C,GAAG,CAAC,EAAE,OAAO,CAAC,SAAS,CAAC,GAAG;QAAE,MAAM,CAAC,EAAE,MAAM,CAAA;KAAE,CAAC;IAC/C,0BAA0B,CAAC,EAAE,MAAM,CAAC;CACrC,CAAC;AAiDF,wBAAgB,yBAAyB,CAAC,MAAM,EAAE;IAChD,GAAG,EAAE,MAAM,CAAC,UAAU,CAAC;IACvB,SAAS,EAAE,eAAe,CAAC;IAC3B,SAAS,EAAE,wBAAwB,CAAC;CACrC,GAAG,0BAA0B,GAAG,SAAS,CAyDzC"}
|
|
@@ -0,0 +1,94 @@
|
|
|
1
|
+
import { z } from "zod";
|
|
2
|
+
export const DEFAULT_OPENAI_REALTIME_TRANSCRIPTION_MODEL = "gpt-4o-transcribe";
|
|
3
|
+
export const DEFAULT_OPENAI_TTS_MODEL = "tts-1";
|
|
4
|
+
const OpenAiTtsVoiceSchema = z.enum([
|
|
5
|
+
"alloy",
|
|
6
|
+
"echo",
|
|
7
|
+
"fable",
|
|
8
|
+
"onyx",
|
|
9
|
+
"nova",
|
|
10
|
+
"shimmer",
|
|
11
|
+
]);
|
|
12
|
+
const OpenAiTtsModelSchema = z.enum(["tts-1", "tts-1-hd"]);
|
|
13
|
+
const NumberLikeSchema = z.union([
|
|
14
|
+
z.number(),
|
|
15
|
+
z.string().trim().min(1),
|
|
16
|
+
]);
|
|
17
|
+
const OptionalFiniteNumberSchema = NumberLikeSchema
|
|
18
|
+
.pipe(z.coerce.number().finite())
|
|
19
|
+
.optional();
|
|
20
|
+
const OptionalTrimmedStringSchema = z
|
|
21
|
+
.string()
|
|
22
|
+
.trim()
|
|
23
|
+
.optional()
|
|
24
|
+
.transform((value) => (value && value.length > 0 ? value : undefined));
|
|
25
|
+
const OpenAiSpeechResolutionSchema = z.object({
|
|
26
|
+
apiKey: OptionalTrimmedStringSchema,
|
|
27
|
+
sttConfidenceThreshold: OptionalFiniteNumberSchema,
|
|
28
|
+
sttModel: OptionalTrimmedStringSchema,
|
|
29
|
+
ttsVoice: z
|
|
30
|
+
.string()
|
|
31
|
+
.trim()
|
|
32
|
+
.toLowerCase()
|
|
33
|
+
.pipe(OpenAiTtsVoiceSchema)
|
|
34
|
+
.default("alloy"),
|
|
35
|
+
ttsModel: z
|
|
36
|
+
.string()
|
|
37
|
+
.trim()
|
|
38
|
+
.toLowerCase()
|
|
39
|
+
.pipe(OpenAiTtsModelSchema)
|
|
40
|
+
.default(DEFAULT_OPENAI_TTS_MODEL),
|
|
41
|
+
realtimeTranscriptionModel: OptionalTrimmedStringSchema.default(DEFAULT_OPENAI_REALTIME_TRANSCRIPTION_MODEL),
|
|
42
|
+
});
|
|
43
|
+
export function resolveOpenAiSpeechConfig(params) {
|
|
44
|
+
const parsed = OpenAiSpeechResolutionSchema.parse({
|
|
45
|
+
apiKey: params.env.OPENAI_API_KEY ?? params.persisted.providers?.openai?.apiKey,
|
|
46
|
+
sttConfidenceThreshold: params.env.STT_CONFIDENCE_THRESHOLD ??
|
|
47
|
+
params.persisted.features?.dictation?.stt?.confidenceThreshold,
|
|
48
|
+
sttModel: params.env.STT_MODEL ??
|
|
49
|
+
(params.providers.voiceStt.provider === "openai"
|
|
50
|
+
? params.persisted.features?.voiceMode?.stt?.model
|
|
51
|
+
: undefined) ??
|
|
52
|
+
(params.providers.dictationStt.provider === "openai"
|
|
53
|
+
? params.persisted.features?.dictation?.stt?.model
|
|
54
|
+
: undefined),
|
|
55
|
+
ttsVoice: params.env.TTS_VOICE ??
|
|
56
|
+
(params.providers.voiceTts.provider === "openai"
|
|
57
|
+
? params.persisted.features?.voiceMode?.tts?.voice
|
|
58
|
+
: undefined) ??
|
|
59
|
+
"alloy",
|
|
60
|
+
ttsModel: params.env.TTS_MODEL ??
|
|
61
|
+
(params.providers.voiceTts.provider === "openai"
|
|
62
|
+
? params.persisted.features?.voiceMode?.tts?.model
|
|
63
|
+
: undefined) ??
|
|
64
|
+
DEFAULT_OPENAI_TTS_MODEL,
|
|
65
|
+
realtimeTranscriptionModel: params.env.OPENAI_REALTIME_TRANSCRIPTION_MODEL ??
|
|
66
|
+
(params.providers.dictationStt.provider === "openai"
|
|
67
|
+
? params.persisted.features?.dictation?.stt?.model
|
|
68
|
+
: undefined) ??
|
|
69
|
+
DEFAULT_OPENAI_REALTIME_TRANSCRIPTION_MODEL,
|
|
70
|
+
});
|
|
71
|
+
if (!parsed.apiKey) {
|
|
72
|
+
return undefined;
|
|
73
|
+
}
|
|
74
|
+
return {
|
|
75
|
+
apiKey: parsed.apiKey,
|
|
76
|
+
stt: {
|
|
77
|
+
apiKey: parsed.apiKey,
|
|
78
|
+
...(parsed.sttConfidenceThreshold !== undefined
|
|
79
|
+
? { confidenceThreshold: parsed.sttConfidenceThreshold }
|
|
80
|
+
: {}),
|
|
81
|
+
...(parsed.sttModel
|
|
82
|
+
? { model: parsed.sttModel }
|
|
83
|
+
: {}),
|
|
84
|
+
},
|
|
85
|
+
tts: {
|
|
86
|
+
apiKey: parsed.apiKey,
|
|
87
|
+
voice: parsed.ttsVoice,
|
|
88
|
+
model: parsed.ttsModel,
|
|
89
|
+
responseFormat: "pcm",
|
|
90
|
+
},
|
|
91
|
+
realtimeTranscriptionModel: parsed.realtimeTranscriptionModel,
|
|
92
|
+
};
|
|
93
|
+
}
|
|
94
|
+
//# sourceMappingURL=config.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"config.js","sourceRoot":"","sources":["../../../../../../src/server/speech/providers/openai/config.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,CAAC,EAAE,MAAM,KAAK,CAAC;AAOxB,MAAM,CAAC,MAAM,2CAA2C,GAAG,mBAAmB,CAAC;AAC/E,MAAM,CAAC,MAAM,wBAAwB,GAAG,OAAO,CAAC;AAShD,MAAM,oBAAoB,GAAG,CAAC,CAAC,IAAI,CAAC;IAClC,OAAO;IACP,MAAM;IACN,OAAO;IACP,MAAM;IACN,MAAM;IACN,SAAS;CACV,CAAC,CAAC;AAEH,MAAM,oBAAoB,GAAG,CAAC,CAAC,IAAI,CAAC,CAAC,OAAO,EAAE,UAAU,CAAC,CAAC,CAAC;AAE3D,MAAM,gBAAgB,GAAG,CAAC,CAAC,KAAK,CAAC;IAC/B,CAAC,CAAC,MAAM,EAAE;IACV,CAAC,CAAC,MAAM,EAAE,CAAC,IAAI,EAAE,CAAC,GAAG,CAAC,CAAC,CAAC;CACzB,CAAC,CAAC;AAEH,MAAM,0BAA0B,GAAG,gBAAgB;KAChD,IAAI,CAAC,CAAC,CAAC,MAAM,CAAC,MAAM,EAAE,CAAC,MAAM,EAAE,CAAC;KAChC,QAAQ,EAAE,CAAC;AAEd,MAAM,2BAA2B,GAAG,CAAC;KAClC,MAAM,EAAE;KACR,IAAI,EAAE;KACN,QAAQ,EAAE;KACV,SAAS,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,CAAC,KAAK,IAAI,KAAK,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,CAAC;AAEzE,MAAM,4BAA4B,GAAG,CAAC,CAAC,MAAM,CAAC;IAC5C,MAAM,EAAE,2BAA2B;IACnC,sBAAsB,EAAE,0BAA0B;IAClD,QAAQ,EAAE,2BAA2B;IACrC,QAAQ,EAAE,CAAC;SACR,MAAM,EAAE;SACR,IAAI,EAAE;SACN,WAAW,EAAE;SACb,IAAI,CAAC,oBAAoB,CAAC;SAC1B,OAAO,CAAC,OAAO,CAAC;IACnB,QAAQ,EAAE,CAAC;SACR,MAAM,EAAE;SACR,IAAI,EAAE;SACN,WAAW,EAAE;SACb,IAAI,CAAC,oBAAoB,CAAC;SAC1B,OAAO,CAAC,wBAAwB,CAAC;IACpC,0BAA0B,EAAE,2BAA2B,CAAC,OAAO,CAC7D,2CAA2C,CAC5C;CACF,CAAC,CAAC;AAEH,MAAM,UAAU,yBAAyB,CAAC,MAIzC;IACC,MAAM,MAAM,GAAG,4BAA4B,CAAC,KAAK,CAAC;QAChD,MAAM,EAAE,MAAM,CAAC,GAAG,CAAC,cAAc,IAAI,MAAM,CAAC,SAAS,CAAC,SAAS,EAAE,MAAM,EAAE,MAAM;QAC/E,sBAAsB,EACpB,MAAM,CAAC,GAAG,CAAC,wBAAwB;YACnC,MAAM,CAAC,SAAS,CAAC,QAAQ,EAAE,SAAS,EAAE,GAAG,EAAE,mBAAmB;QAChE,QAAQ,EACN,MAAM,CAAC,GAAG,CAAC,SAAS;YACpB,CAAC,MAAM,CAAC,SAAS,CAAC,QAAQ,CAAC,QAAQ,KAAK,QAAQ;gBAC9C,CAAC,CAAC,MAAM,CAAC,SAAS,CAAC,QAAQ,EAAE,SAAS,EAAE,GAAG,EAAE,KAAK;gBAClD,CAAC,CAAC,SAAS,CAAC;YACd,CAAC,MAAM,CAAC,SAAS,CAAC,YAAY,CAAC,QAAQ,KAAK,QAAQ;gBAClD,CAAC,CAAC,MAAM,CAAC,SAAS,CAAC,QAAQ,EAAE,SAAS,EAAE,GAAG,EAAE,KAAK;gBAClD,CAAC,CAAC,SAAS,CAAC;QAChB,QAAQ,EACN,MAAM,CAAC,GAAG,CAAC,SAAS;YACpB,CAAC,MAAM,CAAC,SAAS,CAAC,QAAQ,CAAC,QAAQ,KAAK,QAAQ;gBAC9C,CAAC,CAAC,MAAM,CAAC,SAAS,CAAC,QAAQ,EAAE,SAAS,EAAE,GAAG,EAAE,KAAK;gBAClD,CAAC,CAAC,SAAS,CAAC;YACd,OAAO;QACT,QAAQ,EACN,MAAM,CAAC,GAAG,CAAC,SAAS;YACpB,CAAC,MAAM,CAAC,SAAS,CAAC,QAAQ,CAAC,QAAQ,KAAK,QAAQ;gBAC9C,CAAC,CAAC,MAAM,CAAC,SAAS,CAAC,QAAQ,EAAE,SAAS,EAAE,GAAG,EAAE,KAAK;gBAClD,CAAC,CAAC,SAAS,CAAC;YACd,wBAAwB;QAC1B,0BAA0B,EACxB,MAAM,CAAC,GAAG,CAAC,mCAAmC;YAC9C,CAAC,MAAM,CAAC,SAAS,CAAC,YAAY,CAAC,QAAQ,KAAK,QAAQ;gBAClD,CAAC,CAAC,MAAM,CAAC,SAAS,CAAC,QAAQ,EAAE,SAAS,EAAE,GAAG,EAAE,KAAK;gBAClD,CAAC,CAAC,SAAS,CAAC;YACd,2CAA2C;KAC9C,CAAC,CAAC;IAEH,IAAI,CAAC,MAAM,CAAC,MAAM,EAAE,CAAC;QACnB,OAAO,SAAS,CAAC;IACnB,CAAC;IAED,OAAO;QACL,MAAM,EAAE,MAAM,CAAC,MAAM;QACrB,GAAG,EAAE;YACH,MAAM,EAAE,MAAM,CAAC,MAAM;YACrB,GAAG,CAAC,MAAM,CAAC,sBAAsB,KAAK,SAAS;gBAC7C,CAAC,CAAC,EAAE,mBAAmB,EAAE,MAAM,CAAC,sBAAsB,EAAE;gBACxD,CAAC,CAAC,EAAE,CAAC;YACP,GAAG,CAAC,MAAM,CAAC,QAAQ;gBACjB,CAAC,CAAC,EAAE,KAAK,EAAE,MAAM,CAAC,QAAQ,EAAE;gBAC5B,CAAC,CAAC,EAAE,CAAC;SACR;QACD,GAAG,EAAE;YACH,MAAM,EAAE,MAAM,CAAC,MAAM;YACrB,KAAK,EAAE,MAAM,CAAC,QAAQ;YACtB,KAAK,EAAE,MAAM,CAAC,QAAQ;YACtB,cAAc,EAAE,KAAK;SACtB;QACD,0BAA0B,EAAE,MAAM,CAAC,0BAA0B;KAC9D,CAAC;AACJ,CAAC"}
|
|
@@ -0,0 +1,42 @@
|
|
|
1
|
+
import type pino from "pino";
|
|
2
|
+
import { EventEmitter } from "node:events";
|
|
3
|
+
import type { StreamingTranscriptionSession } from "../../speech-provider.js";
|
|
4
|
+
type OpenAITurnDetection = null | {
|
|
5
|
+
type: "server_vad";
|
|
6
|
+
create_response?: boolean;
|
|
7
|
+
threshold?: number;
|
|
8
|
+
prefix_padding_ms?: number;
|
|
9
|
+
silence_duration_ms?: number;
|
|
10
|
+
} | {
|
|
11
|
+
type: "semantic_vad";
|
|
12
|
+
create_response?: boolean;
|
|
13
|
+
eagerness?: "low" | "medium" | "high";
|
|
14
|
+
};
|
|
15
|
+
export declare class OpenAIRealtimeTranscriptionSession extends EventEmitter implements StreamingTranscriptionSession {
|
|
16
|
+
readonly requiredSampleRate = 24000;
|
|
17
|
+
private readonly apiKey;
|
|
18
|
+
private readonly logger;
|
|
19
|
+
private readonly transcriptionModel;
|
|
20
|
+
private readonly language?;
|
|
21
|
+
private readonly prompt?;
|
|
22
|
+
private readonly turnDetection;
|
|
23
|
+
private ws;
|
|
24
|
+
private ready;
|
|
25
|
+
private closing;
|
|
26
|
+
private partialByItemId;
|
|
27
|
+
constructor(params: {
|
|
28
|
+
apiKey: string;
|
|
29
|
+
logger: pino.Logger;
|
|
30
|
+
transcriptionModel: string;
|
|
31
|
+
language?: string;
|
|
32
|
+
prompt?: string;
|
|
33
|
+
turnDetection?: OpenAITurnDetection;
|
|
34
|
+
});
|
|
35
|
+
connect(): Promise<void>;
|
|
36
|
+
appendPcm16(pcm16le: Buffer): void;
|
|
37
|
+
commit(): void;
|
|
38
|
+
clear(): void;
|
|
39
|
+
close(): void;
|
|
40
|
+
}
|
|
41
|
+
export {};
|
|
42
|
+
//# sourceMappingURL=realtime-transcription-session.d.ts.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"realtime-transcription-session.d.ts","sourceRoot":"","sources":["../../../../../../src/server/speech/providers/openai/realtime-transcription-session.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,IAAI,MAAM,MAAM,CAAC;AAE7B,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAC3C,OAAO,KAAK,EAAE,6BAA6B,EAAE,MAAM,0BAA0B,CAAC;AAE9E,KAAK,mBAAmB,GACpB,IAAI,GACJ;IACE,IAAI,EAAE,YAAY,CAAC;IACnB,eAAe,CAAC,EAAE,OAAO,CAAC;IAC1B,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,iBAAiB,CAAC,EAAE,MAAM,CAAC;IAC3B,mBAAmB,CAAC,EAAE,MAAM,CAAC;CAC9B,GACD;IACE,IAAI,EAAE,cAAc,CAAC;IACrB,eAAe,CAAC,EAAE,OAAO,CAAC;IAC1B,SAAS,CAAC,EAAE,KAAK,GAAG,QAAQ,GAAG,MAAM,CAAC;CACvC,CAAC;AA6CN,qBAAa,kCACX,SAAQ,YACR,YAAW,6BAA6B;IAExC,SAAgB,kBAAkB,SAAS;IAC3C,OAAO,CAAC,QAAQ,CAAC,MAAM,CAAS;IAChC,OAAO,CAAC,QAAQ,CAAC,MAAM,CAAc;IACrC,OAAO,CAAC,QAAQ,CAAC,kBAAkB,CAAS;IAC5C,OAAO,CAAC,QAAQ,CAAC,QAAQ,CAAC,CAAS;IACnC,OAAO,CAAC,QAAQ,CAAC,MAAM,CAAC,CAAS;IACjC,OAAO,CAAC,QAAQ,CAAC,aAAa,CAAsB;IAEpD,OAAO,CAAC,EAAE,CAA0B;IACpC,OAAO,CAAC,KAAK,CAA8B;IAC3C,OAAO,CAAC,OAAO,CAAS;IACxB,OAAO,CAAC,eAAe,CAA6B;gBAExC,MAAM,EAAE;QAClB,MAAM,EAAE,MAAM,CAAC;QACf,MAAM,EAAE,IAAI,CAAC,MAAM,CAAC;QACpB,kBAAkB,EAAE,MAAM,CAAC;QAC3B,QAAQ,CAAC,EAAE,MAAM,CAAC;QAClB,MAAM,CAAC,EAAE,MAAM,CAAC;QAChB,aAAa,CAAC,EAAE,mBAAmB,CAAC;KACrC;IAUY,OAAO,IAAI,OAAO,CAAC,IAAI,CAAC;IAgI9B,WAAW,CAAC,OAAO,EAAE,MAAM,GAAG,IAAI;IASlC,MAAM,IAAI,IAAI;IAQd,KAAK,IAAI,IAAI;IAQb,KAAK,IAAI,IAAI;CAWrB"}
|
|
@@ -0,0 +1,165 @@
|
|
|
1
|
+
import WebSocket from "ws";
|
|
2
|
+
import { EventEmitter } from "node:events";
|
|
3
|
+
export class OpenAIRealtimeTranscriptionSession extends EventEmitter {
|
|
4
|
+
constructor(params) {
|
|
5
|
+
super();
|
|
6
|
+
this.requiredSampleRate = 24000;
|
|
7
|
+
this.ws = null;
|
|
8
|
+
this.ready = null;
|
|
9
|
+
this.closing = false;
|
|
10
|
+
this.partialByItemId = new Map();
|
|
11
|
+
this.apiKey = params.apiKey;
|
|
12
|
+
this.logger = params.logger.child({ provider: "openai", component: "realtime-transcription" });
|
|
13
|
+
this.transcriptionModel = params.transcriptionModel;
|
|
14
|
+
this.language = params.language;
|
|
15
|
+
this.prompt = params.prompt;
|
|
16
|
+
this.turnDetection = params.turnDetection ?? null;
|
|
17
|
+
}
|
|
18
|
+
async connect() {
|
|
19
|
+
if (this.ready) {
|
|
20
|
+
return this.ready;
|
|
21
|
+
}
|
|
22
|
+
this.closing = false;
|
|
23
|
+
this.ready = new Promise((resolve, reject) => {
|
|
24
|
+
const url = "wss://api.openai.com/v1/realtime?intent=transcription";
|
|
25
|
+
const ws = new WebSocket(url, {
|
|
26
|
+
headers: {
|
|
27
|
+
Authorization: `Bearer ${this.apiKey}`,
|
|
28
|
+
},
|
|
29
|
+
});
|
|
30
|
+
this.ws = ws;
|
|
31
|
+
let resolved = false;
|
|
32
|
+
const fail = (error) => {
|
|
33
|
+
if (resolved) {
|
|
34
|
+
this.emit("error", error);
|
|
35
|
+
return;
|
|
36
|
+
}
|
|
37
|
+
resolved = true;
|
|
38
|
+
reject(error);
|
|
39
|
+
};
|
|
40
|
+
ws.on("open", () => {
|
|
41
|
+
this.logger.debug("OpenAI realtime transcription websocket connected");
|
|
42
|
+
const update = {
|
|
43
|
+
type: "session.update",
|
|
44
|
+
session: {
|
|
45
|
+
type: "transcription",
|
|
46
|
+
audio: {
|
|
47
|
+
input: {
|
|
48
|
+
format: { type: "audio/pcm", rate: 24000 },
|
|
49
|
+
transcription: {
|
|
50
|
+
model: this.transcriptionModel,
|
|
51
|
+
...(this.language ? { language: this.language } : {}),
|
|
52
|
+
...(this.prompt ? { prompt: this.prompt } : {}),
|
|
53
|
+
},
|
|
54
|
+
turn_detection: this.turnDetection,
|
|
55
|
+
},
|
|
56
|
+
},
|
|
57
|
+
},
|
|
58
|
+
};
|
|
59
|
+
ws.send(JSON.stringify(update));
|
|
60
|
+
});
|
|
61
|
+
ws.on("message", (data) => {
|
|
62
|
+
const text = typeof data === "string" ? data : data.toString("utf-8");
|
|
63
|
+
let parsed;
|
|
64
|
+
try {
|
|
65
|
+
parsed = JSON.parse(text);
|
|
66
|
+
}
|
|
67
|
+
catch {
|
|
68
|
+
return;
|
|
69
|
+
}
|
|
70
|
+
const event = parsed;
|
|
71
|
+
if (event.type === "session.created" ||
|
|
72
|
+
event.type === "session.updated") {
|
|
73
|
+
if (!resolved) {
|
|
74
|
+
resolved = true;
|
|
75
|
+
resolve();
|
|
76
|
+
}
|
|
77
|
+
return;
|
|
78
|
+
}
|
|
79
|
+
if (event.type === "input_audio_buffer.committed") {
|
|
80
|
+
this.emit("committed", {
|
|
81
|
+
segmentId: event.item_id,
|
|
82
|
+
previousSegmentId: event.previous_item_id,
|
|
83
|
+
});
|
|
84
|
+
return;
|
|
85
|
+
}
|
|
86
|
+
if (event.type === "input_audio_buffer.speech_started") {
|
|
87
|
+
this.emit("speech_started");
|
|
88
|
+
return;
|
|
89
|
+
}
|
|
90
|
+
if (event.type === "input_audio_buffer.speech_stopped") {
|
|
91
|
+
this.emit("speech_stopped");
|
|
92
|
+
return;
|
|
93
|
+
}
|
|
94
|
+
if (event.type === "conversation.item.input_audio_transcription.delta") {
|
|
95
|
+
const replaceDelta = this.transcriptionModel === "whisper-1";
|
|
96
|
+
const prev = this.partialByItemId.get(event.item_id) ?? "";
|
|
97
|
+
const next = replaceDelta ? event.delta : prev + event.delta;
|
|
98
|
+
this.partialByItemId.set(event.item_id, next);
|
|
99
|
+
this.emit("transcript", { segmentId: event.item_id, transcript: next, isFinal: false });
|
|
100
|
+
return;
|
|
101
|
+
}
|
|
102
|
+
if (event.type === "conversation.item.input_audio_transcription.completed") {
|
|
103
|
+
this.partialByItemId.set(event.item_id, event.transcript);
|
|
104
|
+
this.emit("transcript", { segmentId: event.item_id, transcript: event.transcript, isFinal: true });
|
|
105
|
+
return;
|
|
106
|
+
}
|
|
107
|
+
if (event.type === "error") {
|
|
108
|
+
const message = event.error?.message ?? "OpenAI realtime error";
|
|
109
|
+
fail(new Error(message));
|
|
110
|
+
}
|
|
111
|
+
});
|
|
112
|
+
ws.on("error", (err) => {
|
|
113
|
+
fail(err instanceof Error ? err : new Error(String(err)));
|
|
114
|
+
});
|
|
115
|
+
ws.on("close", () => {
|
|
116
|
+
this.logger.debug("OpenAI realtime websocket closed");
|
|
117
|
+
if (this.closing) {
|
|
118
|
+
return;
|
|
119
|
+
}
|
|
120
|
+
if (!resolved) {
|
|
121
|
+
fail(new Error("OpenAI realtime websocket closed before ready"));
|
|
122
|
+
return;
|
|
123
|
+
}
|
|
124
|
+
fail(new Error("OpenAI realtime websocket closed"));
|
|
125
|
+
});
|
|
126
|
+
});
|
|
127
|
+
return this.ready;
|
|
128
|
+
}
|
|
129
|
+
appendPcm16(pcm16le) {
|
|
130
|
+
if (!this.ws || this.ws.readyState !== WebSocket.OPEN) {
|
|
131
|
+
throw new Error("OpenAI realtime websocket not connected");
|
|
132
|
+
}
|
|
133
|
+
const base64Audio = pcm16le.toString("base64");
|
|
134
|
+
const event = { type: "input_audio_buffer.append", audio: base64Audio };
|
|
135
|
+
this.ws.send(JSON.stringify(event));
|
|
136
|
+
}
|
|
137
|
+
commit() {
|
|
138
|
+
if (!this.ws || this.ws.readyState !== WebSocket.OPEN) {
|
|
139
|
+
throw new Error("OpenAI realtime websocket not connected");
|
|
140
|
+
}
|
|
141
|
+
const event = { type: "input_audio_buffer.commit" };
|
|
142
|
+
this.ws.send(JSON.stringify(event));
|
|
143
|
+
}
|
|
144
|
+
clear() {
|
|
145
|
+
if (!this.ws || this.ws.readyState !== WebSocket.OPEN) {
|
|
146
|
+
return;
|
|
147
|
+
}
|
|
148
|
+
const event = { type: "input_audio_buffer.clear" };
|
|
149
|
+
this.ws.send(JSON.stringify(event));
|
|
150
|
+
}
|
|
151
|
+
close() {
|
|
152
|
+
try {
|
|
153
|
+
this.closing = true;
|
|
154
|
+
this.ws?.close();
|
|
155
|
+
}
|
|
156
|
+
catch {
|
|
157
|
+
// no-op
|
|
158
|
+
}
|
|
159
|
+
finally {
|
|
160
|
+
this.ws = null;
|
|
161
|
+
this.ready = null;
|
|
162
|
+
}
|
|
163
|
+
}
|
|
164
|
+
}
|
|
165
|
+
//# sourceMappingURL=realtime-transcription-session.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"realtime-transcription-session.js","sourceRoot":"","sources":["../../../../../../src/server/speech/providers/openai/realtime-transcription-session.ts"],"names":[],"mappings":"AACA,OAAO,SAAS,MAAM,IAAI,CAAC;AAC3B,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AA6D3C,MAAM,OAAO,kCACX,SAAQ,YAAY;IAgBpB,YAAY,MAOX;QACC,KAAK,EAAE,CAAC;QArBM,uBAAkB,GAAG,KAAK,CAAC;QAQnC,OAAE,GAAqB,IAAI,CAAC;QAC5B,UAAK,GAAyB,IAAI,CAAC;QACnC,YAAO,GAAG,KAAK,CAAC;QAChB,oBAAe,GAAG,IAAI,GAAG,EAAkB,CAAC;QAWlD,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC,MAAM,CAAC;QAC5B,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC,MAAM,CAAC,KAAK,CAAC,EAAE,QAAQ,EAAE,QAAQ,EAAE,SAAS,EAAE,wBAAwB,EAAE,CAAC,CAAC;QAC/F,IAAI,CAAC,kBAAkB,GAAG,MAAM,CAAC,kBAAkB,CAAC;QACpD,IAAI,CAAC,QAAQ,GAAG,MAAM,CAAC,QAAQ,CAAC;QAChC,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC,MAAM,CAAC;QAC5B,IAAI,CAAC,aAAa,GAAG,MAAM,CAAC,aAAa,IAAI,IAAI,CAAC;IACpD,CAAC;IAEM,KAAK,CAAC,OAAO;QAClB,IAAI,IAAI,CAAC,KAAK,EAAE,CAAC;YACf,OAAO,IAAI,CAAC,KAAK,CAAC;QACpB,CAAC;QAED,IAAI,CAAC,OAAO,GAAG,KAAK,CAAC;QACrB,IAAI,CAAC,KAAK,GAAG,IAAI,OAAO,CAAO,CAAC,OAAO,EAAE,MAAM,EAAE,EAAE;YACjD,MAAM,GAAG,GAAG,uDAAuD,CAAC;YACpE,MAAM,EAAE,GAAG,IAAI,SAAS,CAAC,GAAG,EAAE;gBAC5B,OAAO,EAAE;oBACP,aAAa,EAAE,UAAU,IAAI,CAAC,MAAM,EAAE;iBACvC;aACF,CAAC,CAAC;YACH,IAAI,CAAC,EAAE,GAAG,EAAE,CAAC;YAEb,IAAI,QAAQ,GAAG,KAAK,CAAC;YAErB,MAAM,IAAI,GAAG,CAAC,KAAY,EAAE,EAAE;gBAC5B,IAAI,QAAQ,EAAE,CAAC;oBACb,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,KAAK,CAAC,CAAC;oBAC1B,OAAO;gBACT,CAAC;gBACD,QAAQ,GAAG,IAAI,CAAC;gBAChB,MAAM,CAAC,KAAK,CAAC,CAAC;YAChB,CAAC,CAAC;YAEF,EAAE,CAAC,EAAE,CAAC,MAAM,EAAE,GAAG,EAAE;gBACjB,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,mDAAmD,CAAC,CAAC;gBACvE,MAAM,MAAM,GAAsB;oBAChC,IAAI,EAAE,gBAAgB;oBACtB,OAAO,EAAE;wBACP,IAAI,EAAE,eAAe;wBACrB,KAAK,EAAE;4BACL,KAAK,EAAE;gCACL,MAAM,EAAE,EAAE,IAAI,EAAE,WAAW,EAAE,IAAI,EAAE,KAAK,EAAE;gCAC1C,aAAa,EAAE;oCACb,KAAK,EAAE,IAAI,CAAC,kBAAkB;oCAC9B,GAAG,CAAC,IAAI,CAAC,QAAQ,CAAC,CAAC,CAAC,EAAE,QAAQ,EAAE,IAAI,CAAC,QAAQ,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;oCACrD,GAAG,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,MAAM,EAAE,IAAI,CAAC,MAAM,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;iCAChD;gCACD,cAAc,EAAE,IAAI,CAAC,aAAa;6BACnC;yBACF;qBACF;iBACF,CAAC;gBACF,EAAE,CAAC,IAAI,CAAC,IAAI,CAAC,SAAS,CAAC,MAAM,CAAC,CAAC,CAAC;YAClC,CAAC,CAAC,CAAC;YAEH,EAAE,CAAC,EAAE,CAAC,SAAS,EAAE,CAAC,IAAI,EAAE,EAAE;gBACxB,MAAM,IAAI,GAAG,OAAO,IAAI,KAAK,QAAQ,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,IAAI,CAAC,QAAQ,CAAC,OAAO,CAAC,CAAC;gBACtE,IAAI,MAAe,CAAC;gBACpB,IAAI,CAAC;oBACH,MAAM,GAAG,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC;gBAC5B,CAAC;gBAAC,MAAM,CAAC;oBACP,OAAO;gBACT,CAAC;gBAED,MAAM,KAAK,GAAG,MAA2B,CAAC;gBAC1C,IACE,KAAK,CAAC,IAAI,KAAK,iBAAiB;oBAChC,KAAK,CAAC,IAAI,KAAK,iBAAiB,EAChC,CAAC;oBACD,IAAI,CAAC,QAAQ,EAAE,CAAC;wBACd,QAAQ,GAAG,IAAI,CAAC;wBAChB,OAAO,EAAE,CAAC;oBACZ,CAAC;oBACD,OAAO;gBACT,CAAC;gBAED,IAAI,KAAK,CAAC,IAAI,KAAK,8BAA8B,EAAE,CAAC;oBAClD,IAAI,CAAC,IAAI,CAAC,WAAW,EAAE;wBACrB,SAAS,EAAE,KAAK,CAAC,OAAO;wBACxB,iBAAiB,EAAE,KAAK,CAAC,gBAAgB;qBAC1C,CAAC,CAAC;oBACH,OAAO;gBACT,CAAC;gBAED,IAAI,KAAK,CAAC,IAAI,KAAK,mCAAmC,EAAE,CAAC;oBACvD,IAAI,CAAC,IAAI,CAAC,gBAAgB,CAAC,CAAC;oBAC5B,OAAO;gBACT,CAAC;gBAED,IAAI,KAAK,CAAC,IAAI,KAAK,mCAAmC,EAAE,CAAC;oBACvD,IAAI,CAAC,IAAI,CAAC,gBAAgB,CAAC,CAAC;oBAC5B,OAAO;gBACT,CAAC;gBAED,IAAI,KAAK,CAAC,IAAI,KAAK,mDAAmD,EAAE,CAAC;oBACvE,MAAM,YAAY,GAAG,IAAI,CAAC,kBAAkB,KAAK,WAAW,CAAC;oBAC7D,MAAM,IAAI,GAAG,IAAI,CAAC,eAAe,CAAC,GAAG,CAAC,KAAK,CAAC,OAAO,CAAC,IAAI,EAAE,CAAC;oBAC3D,MAAM,IAAI,GAAG,YAAY,CAAC,CAAC,CAAC,KAAK,CAAC,KAAK,CAAC,CAAC,CAAC,IAAI,GAAG,KAAK,CAAC,KAAK,CAAC;oBAC7D,IAAI,CAAC,eAAe,CAAC,GAAG,CAAC,KAAK,CAAC,OAAO,EAAE,IAAI,CAAC,CAAC;oBAC9C,IAAI,CAAC,IAAI,CAAC,YAAY,EAAE,EAAE,SAAS,EAAE,KAAK,CAAC,OAAO,EAAE,UAAU,EAAE,IAAI,EAAE,OAAO,EAAE,KAAK,EAAE,CAAC,CAAC;oBACxF,OAAO;gBACT,CAAC;gBAED,IAAI,KAAK,CAAC,IAAI,KAAK,uDAAuD,EAAE,CAAC;oBAC3E,IAAI,CAAC,eAAe,CAAC,GAAG,CAAC,KAAK,CAAC,OAAO,EAAE,KAAK,CAAC,UAAU,CAAC,CAAC;oBAC1D,IAAI,CAAC,IAAI,CAAC,YAAY,EAAE,EAAE,SAAS,EAAE,KAAK,CAAC,OAAO,EAAE,UAAU,EAAE,KAAK,CAAC,UAAU,EAAE,OAAO,EAAE,IAAI,EAAE,CAAC,CAAC;oBACnG,OAAO;gBACT,CAAC;gBAED,IAAI,KAAK,CAAC,IAAI,KAAK,OAAO,EAAE,CAAC;oBAC3B,MAAM,OAAO,GAAG,KAAK,CAAC,KAAK,EAAE,OAAO,IAAI,uBAAuB,CAAC;oBAChE,IAAI,CAAC,IAAI,KAAK,CAAC,OAAO,CAAC,CAAC,CAAC;gBAC3B,CAAC;YACH,CAAC,CAAC,CAAC;YAEH,EAAE,CAAC,EAAE,CAAC,OAAO,EAAE,CAAC,GAAG,EAAE,EAAE;gBACrB,IAAI,CAAC,GAAG,YAAY,KAAK,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,IAAI,KAAK,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC;YAC5D,CAAC,CAAC,CAAC;YAEH,EAAE,CAAC,EAAE,CAAC,OAAO,EAAE,GAAG,EAAE;gBAClB,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,kCAAkC,CAAC,CAAC;gBACtD,IAAI,IAAI,CAAC,OAAO,EAAE,CAAC;oBACjB,OAAO;gBACT,CAAC;gBACD,IAAI,CAAC,QAAQ,EAAE,CAAC;oBACd,IAAI,CAAC,IAAI,KAAK,CAAC,+CAA+C,CAAC,CAAC,CAAC;oBACjE,OAAO;gBACT,CAAC;gBACD,IAAI,CAAC,IAAI,KAAK,CAAC,kCAAkC,CAAC,CAAC,CAAC;YACtD,CAAC,CAAC,CAAC;QACL,CAAC,CAAC,CAAC;QAEH,OAAO,IAAI,CAAC,KAAK,CAAC;IACpB,CAAC;IAEM,WAAW,CAAC,OAAe;QAChC,IAAI,CAAC,IAAI,CAAC,EAAE,IAAI,IAAI,CAAC,EAAE,CAAC,UAAU,KAAK,SAAS,CAAC,IAAI,EAAE,CAAC;YACtD,MAAM,IAAI,KAAK,CAAC,yCAAyC,CAAC,CAAC;QAC7D,CAAC;QACD,MAAM,WAAW,GAAG,OAAO,CAAC,QAAQ,CAAC,QAAQ,CAAC,CAAC;QAC/C,MAAM,KAAK,GAAsB,EAAE,IAAI,EAAE,2BAA2B,EAAE,KAAK,EAAE,WAAW,EAAE,CAAC;QAC3F,IAAI,CAAC,EAAE,CAAC,IAAI,CAAC,IAAI,CAAC,SAAS,CAAC,KAAK,CAAC,CAAC,CAAC;IACtC,CAAC;IAEM,MAAM;QACX,IAAI,CAAC,IAAI,CAAC,EAAE,IAAI,IAAI,CAAC,EAAE,CAAC,UAAU,KAAK,SAAS,CAAC,IAAI,EAAE,CAAC;YACtD,MAAM,IAAI,KAAK,CAAC,yCAAyC,CAAC,CAAC;QAC7D,CAAC;QACD,MAAM,KAAK,GAAsB,EAAE,IAAI,EAAE,2BAA2B,EAAE,CAAC;QACvE,IAAI,CAAC,EAAE,CAAC,IAAI,CAAC,IAAI,CAAC,SAAS,CAAC,KAAK,CAAC,CAAC,CAAC;IACtC,CAAC;IAEM,KAAK;QACV,IAAI,CAAC,IAAI,CAAC,EAAE,IAAI,IAAI,CAAC,EAAE,CAAC,UAAU,KAAK,SAAS,CAAC,IAAI,EAAE,CAAC;YACtD,OAAO;QACT,CAAC;QACD,MAAM,KAAK,GAAsB,EAAE,IAAI,EAAE,0BAA0B,EAAE,CAAC;QACtE,IAAI,CAAC,EAAE,CAAC,IAAI,CAAC,IAAI,CAAC,SAAS,CAAC,KAAK,CAAC,CAAC,CAAC;IACtC,CAAC;IAEM,KAAK;QACV,IAAI,CAAC;YACH,IAAI,CAAC,OAAO,GAAG,IAAI,CAAC;YACpB,IAAI,CAAC,EAAE,EAAE,KAAK,EAAE,CAAC;QACnB,CAAC;QAAC,MAAM,CAAC;YACP,QAAQ;QACV,CAAC;gBAAS,CAAC;YACT,IAAI,CAAC,EAAE,GAAG,IAAI,CAAC;YACf,IAAI,CAAC,KAAK,GAAG,IAAI,CAAC;QACpB,CAAC;IACH,CAAC;CACF"}
|
|
@@ -0,0 +1,27 @@
|
|
|
1
|
+
import type { Logger } from "pino";
|
|
2
|
+
import type { SpeechToTextProvider, TextToSpeechProvider } from "../../speech-provider.js";
|
|
3
|
+
import type { RequestedSpeechProviders } from "../../speech-types.js";
|
|
4
|
+
import { type OpenAiSpeechProviderConfig } from "./config.js";
|
|
5
|
+
export type OpenAiSpeechAvailability = {
|
|
6
|
+
stt: boolean;
|
|
7
|
+
tts: boolean;
|
|
8
|
+
dictationStt: boolean;
|
|
9
|
+
};
|
|
10
|
+
export type SpeechServices = {
|
|
11
|
+
sttService: SpeechToTextProvider | null;
|
|
12
|
+
ttsService: TextToSpeechProvider | null;
|
|
13
|
+
dictationSttService: SpeechToTextProvider | null;
|
|
14
|
+
};
|
|
15
|
+
export declare function getOpenAiSpeechAvailability(openaiConfig: OpenAiSpeechProviderConfig | undefined): OpenAiSpeechAvailability;
|
|
16
|
+
export declare function validateOpenAiCredentialRequirements(params: {
|
|
17
|
+
providers: RequestedSpeechProviders;
|
|
18
|
+
openaiConfig: OpenAiSpeechProviderConfig | undefined;
|
|
19
|
+
logger: Logger;
|
|
20
|
+
}): void;
|
|
21
|
+
export declare function initializeOpenAiSpeechServices(params: {
|
|
22
|
+
providers: RequestedSpeechProviders;
|
|
23
|
+
openaiConfig: OpenAiSpeechProviderConfig | undefined;
|
|
24
|
+
existing: SpeechServices;
|
|
25
|
+
logger: Logger;
|
|
26
|
+
}): SpeechServices;
|
|
27
|
+
//# sourceMappingURL=runtime.d.ts.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"runtime.d.ts","sourceRoot":"","sources":["../../../../../../src/server/speech/providers/openai/runtime.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,MAAM,EAAE,MAAM,MAAM,CAAC;AAEnC,OAAO,KAAK,EAAE,oBAAoB,EAAE,oBAAoB,EAAE,MAAM,0BAA0B,CAAC;AAC3F,OAAO,KAAK,EAAE,wBAAwB,EAAE,MAAM,uBAAuB,CAAC;AACtE,OAAO,EAGL,KAAK,0BAA0B,EAChC,MAAM,aAAa,CAAC;AAWrB,MAAM,MAAM,wBAAwB,GAAG;IACrC,GAAG,EAAE,OAAO,CAAC;IACb,GAAG,EAAE,OAAO,CAAC;IACb,YAAY,EAAE,OAAO,CAAC;CACvB,CAAC;AAEF,MAAM,MAAM,cAAc,GAAG;IAC3B,UAAU,EAAE,oBAAoB,GAAG,IAAI,CAAC;IACxC,UAAU,EAAE,oBAAoB,GAAG,IAAI,CAAC;IACxC,mBAAmB,EAAE,oBAAoB,GAAG,IAAI,CAAC;CAClD,CAAC;AAaF,wBAAgB,2BAA2B,CACzC,YAAY,EAAE,0BAA0B,GAAG,SAAS,GACnD,wBAAwB,CAO1B;AAED,wBAAgB,oCAAoC,CAAC,MAAM,EAAE;IAC3D,SAAS,EAAE,wBAAwB,CAAC;IACpC,YAAY,EAAE,0BAA0B,GAAG,SAAS,CAAC;IACrD,MAAM,EAAE,MAAM,CAAC;CAChB,GAAG,IAAI,CAkCP;AAED,wBAAgB,8BAA8B,CAAC,MAAM,EAAE;IACrD,SAAS,EAAE,wBAAwB,CAAC;IACpC,YAAY,EAAE,0BAA0B,GAAG,SAAS,CAAC;IACrD,QAAQ,EAAE,cAAc,CAAC;IACzB,MAAM,EAAE,MAAM,CAAC;CAChB,GAAG,cAAc,CAwEjB"}
|
|
@@ -0,0 +1,103 @@
|
|
|
1
|
+
import { DEFAULT_OPENAI_REALTIME_TRANSCRIPTION_MODEL, DEFAULT_OPENAI_TTS_MODEL, } from "./config.js";
|
|
2
|
+
import { OpenAIRealtimeTranscriptionSession } from "./realtime-transcription-session.js";
|
|
3
|
+
import { OpenAISTT } from "./stt.js";
|
|
4
|
+
import { OpenAITTS } from "./tts.js";
|
|
5
|
+
function resolveOpenAiCredentials(openaiConfig) {
|
|
6
|
+
const openaiApiKey = openaiConfig?.apiKey;
|
|
7
|
+
return {
|
|
8
|
+
openaiSttApiKey: openaiConfig?.stt?.apiKey ?? openaiApiKey,
|
|
9
|
+
openaiTtsApiKey: openaiConfig?.tts?.apiKey ?? openaiApiKey,
|
|
10
|
+
openaiDictationApiKey: openaiApiKey,
|
|
11
|
+
};
|
|
12
|
+
}
|
|
13
|
+
export function getOpenAiSpeechAvailability(openaiConfig) {
|
|
14
|
+
const credentials = resolveOpenAiCredentials(openaiConfig);
|
|
15
|
+
return {
|
|
16
|
+
stt: Boolean(credentials.openaiSttApiKey),
|
|
17
|
+
tts: Boolean(credentials.openaiTtsApiKey),
|
|
18
|
+
dictationStt: Boolean(credentials.openaiDictationApiKey),
|
|
19
|
+
};
|
|
20
|
+
}
|
|
21
|
+
export function validateOpenAiCredentialRequirements(params) {
|
|
22
|
+
const { providers, logger, openaiConfig } = params;
|
|
23
|
+
const openAiCredentials = resolveOpenAiCredentials(openaiConfig);
|
|
24
|
+
const missingOpenAiCredentialsFor = [];
|
|
25
|
+
if (providers.voiceStt.provider === "openai" && !openAiCredentials.openaiSttApiKey) {
|
|
26
|
+
missingOpenAiCredentialsFor.push("voice.stt");
|
|
27
|
+
}
|
|
28
|
+
if (providers.voiceTts.provider === "openai" && !openAiCredentials.openaiTtsApiKey) {
|
|
29
|
+
missingOpenAiCredentialsFor.push("voice.tts");
|
|
30
|
+
}
|
|
31
|
+
if (providers.dictationStt.provider === "openai" &&
|
|
32
|
+
!openAiCredentials.openaiDictationApiKey) {
|
|
33
|
+
missingOpenAiCredentialsFor.push("dictation.stt");
|
|
34
|
+
}
|
|
35
|
+
if (missingOpenAiCredentialsFor.length > 0) {
|
|
36
|
+
logger.error({
|
|
37
|
+
requestedProviders: {
|
|
38
|
+
dictationStt: providers.dictationStt.provider,
|
|
39
|
+
voiceStt: providers.voiceStt.provider,
|
|
40
|
+
voiceTts: providers.voiceTts.provider,
|
|
41
|
+
},
|
|
42
|
+
missingOpenAiCredentialsFor,
|
|
43
|
+
}, "Invalid speech configuration: OpenAI provider selected but credentials are missing");
|
|
44
|
+
throw new Error(`Missing OpenAI credentials for configured speech features: ${missingOpenAiCredentialsFor.join(", ")}`);
|
|
45
|
+
}
|
|
46
|
+
}
|
|
47
|
+
export function initializeOpenAiSpeechServices(params) {
|
|
48
|
+
const { providers, openaiConfig, existing, logger } = params;
|
|
49
|
+
const openAiCredentials = resolveOpenAiCredentials(openaiConfig);
|
|
50
|
+
let sttService = existing.sttService;
|
|
51
|
+
let ttsService = existing.ttsService;
|
|
52
|
+
let dictationSttService = existing.dictationSttService;
|
|
53
|
+
const needsOpenAiStt = !sttService && providers.voiceStt.provider === "openai";
|
|
54
|
+
const needsOpenAiTts = !ttsService && providers.voiceTts.provider === "openai";
|
|
55
|
+
const needsOpenAiDictation = !dictationSttService && providers.dictationStt.provider === "openai";
|
|
56
|
+
if ((needsOpenAiStt || needsOpenAiTts || needsOpenAiDictation) &&
|
|
57
|
+
(openAiCredentials.openaiSttApiKey ||
|
|
58
|
+
openAiCredentials.openaiTtsApiKey ||
|
|
59
|
+
openAiCredentials.openaiDictationApiKey)) {
|
|
60
|
+
logger.info("OpenAI speech provider initialized");
|
|
61
|
+
if (needsOpenAiStt && openAiCredentials.openaiSttApiKey) {
|
|
62
|
+
const { apiKey: _sttApiKey, ...sttConfig } = openaiConfig?.stt ?? {};
|
|
63
|
+
sttService = new OpenAISTT({
|
|
64
|
+
apiKey: openAiCredentials.openaiSttApiKey,
|
|
65
|
+
...sttConfig,
|
|
66
|
+
}, logger);
|
|
67
|
+
}
|
|
68
|
+
if (needsOpenAiTts && openAiCredentials.openaiTtsApiKey) {
|
|
69
|
+
const { apiKey: _ttsApiKey, ...ttsConfig } = openaiConfig?.tts ?? {};
|
|
70
|
+
ttsService = new OpenAITTS({
|
|
71
|
+
apiKey: openAiCredentials.openaiTtsApiKey,
|
|
72
|
+
voice: "alloy",
|
|
73
|
+
model: DEFAULT_OPENAI_TTS_MODEL,
|
|
74
|
+
responseFormat: "pcm",
|
|
75
|
+
...ttsConfig,
|
|
76
|
+
}, logger);
|
|
77
|
+
}
|
|
78
|
+
const dictationApiKey = openAiCredentials.openaiDictationApiKey;
|
|
79
|
+
if (needsOpenAiDictation && dictationApiKey) {
|
|
80
|
+
dictationSttService = {
|
|
81
|
+
id: "openai",
|
|
82
|
+
createSession: ({ logger: sessionLogger, language, prompt }) => new OpenAIRealtimeTranscriptionSession({
|
|
83
|
+
apiKey: dictationApiKey,
|
|
84
|
+
logger: sessionLogger,
|
|
85
|
+
transcriptionModel: openaiConfig?.realtimeTranscriptionModel ??
|
|
86
|
+
DEFAULT_OPENAI_REALTIME_TRANSCRIPTION_MODEL,
|
|
87
|
+
...(language ? { language } : {}),
|
|
88
|
+
...(prompt ? { prompt } : {}),
|
|
89
|
+
turnDetection: null,
|
|
90
|
+
}),
|
|
91
|
+
};
|
|
92
|
+
}
|
|
93
|
+
}
|
|
94
|
+
else if (needsOpenAiStt || needsOpenAiTts || needsOpenAiDictation) {
|
|
95
|
+
logger.warn("OpenAI speech providers are configured but credentials are missing");
|
|
96
|
+
}
|
|
97
|
+
return {
|
|
98
|
+
sttService,
|
|
99
|
+
ttsService,
|
|
100
|
+
dictationSttService,
|
|
101
|
+
};
|
|
102
|
+
}
|
|
103
|
+
//# sourceMappingURL=runtime.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"runtime.js","sourceRoot":"","sources":["../../../../../../src/server/speech/providers/openai/runtime.ts"],"names":[],"mappings":"AAIA,OAAO,EACL,2CAA2C,EAC3C,wBAAwB,GAEzB,MAAM,aAAa,CAAC;AACrB,OAAO,EAAE,kCAAkC,EAAE,MAAM,qCAAqC,CAAC;AACzF,OAAO,EAAE,SAAS,EAAE,MAAM,UAAU,CAAC;AACrC,OAAO,EAAE,SAAS,EAAE,MAAM,UAAU,CAAC;AAoBrC,SAAS,wBAAwB,CAC/B,YAAoD;IAEpD,MAAM,YAAY,GAAG,YAAY,EAAE,MAAM,CAAC;IAC1C,OAAO;QACL,eAAe,EAAE,YAAY,EAAE,GAAG,EAAE,MAAM,IAAI,YAAY;QAC1D,eAAe,EAAE,YAAY,EAAE,GAAG,EAAE,MAAM,IAAI,YAAY;QAC1D,qBAAqB,EAAE,YAAY;KACpC,CAAC;AACJ,CAAC;AAED,MAAM,UAAU,2BAA2B,CACzC,YAAoD;IAEpD,MAAM,WAAW,GAAG,wBAAwB,CAAC,YAAY,CAAC,CAAC;IAC3D,OAAO;QACL,GAAG,EAAE,OAAO,CAAC,WAAW,CAAC,eAAe,CAAC;QACzC,GAAG,EAAE,OAAO,CAAC,WAAW,CAAC,eAAe,CAAC;QACzC,YAAY,EAAE,OAAO,CAAC,WAAW,CAAC,qBAAqB,CAAC;KACzD,CAAC;AACJ,CAAC;AAED,MAAM,UAAU,oCAAoC,CAAC,MAIpD;IACC,MAAM,EAAE,SAAS,EAAE,MAAM,EAAE,YAAY,EAAE,GAAG,MAAM,CAAC;IACnD,MAAM,iBAAiB,GAAG,wBAAwB,CAAC,YAAY,CAAC,CAAC;IAEjE,MAAM,2BAA2B,GAAa,EAAE,CAAC;IACjD,IAAI,SAAS,CAAC,QAAQ,CAAC,QAAQ,KAAK,QAAQ,IAAI,CAAC,iBAAiB,CAAC,eAAe,EAAE,CAAC;QACnF,2BAA2B,CAAC,IAAI,CAAC,WAAW,CAAC,CAAC;IAChD,CAAC;IACD,IAAI,SAAS,CAAC,QAAQ,CAAC,QAAQ,KAAK,QAAQ,IAAI,CAAC,iBAAiB,CAAC,eAAe,EAAE,CAAC;QACnF,2BAA2B,CAAC,IAAI,CAAC,WAAW,CAAC,CAAC;IAChD,CAAC;IACD,IACE,SAAS,CAAC,YAAY,CAAC,QAAQ,KAAK,QAAQ;QAC5C,CAAC,iBAAiB,CAAC,qBAAqB,EACxC,CAAC;QACD,2BAA2B,CAAC,IAAI,CAAC,eAAe,CAAC,CAAC;IACpD,CAAC;IAED,IAAI,2BAA2B,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;QAC3C,MAAM,CAAC,KAAK,CACV;YACE,kBAAkB,EAAE;gBAClB,YAAY,EAAE,SAAS,CAAC,YAAY,CAAC,QAAQ;gBAC7C,QAAQ,EAAE,SAAS,CAAC,QAAQ,CAAC,QAAQ;gBACrC,QAAQ,EAAE,SAAS,CAAC,QAAQ,CAAC,QAAQ;aACtC;YACD,2BAA2B;SAC5B,EACD,oFAAoF,CACrF,CAAC;QACF,MAAM,IAAI,KAAK,CACb,8DAA8D,2BAA2B,CAAC,IAAI,CAAC,IAAI,CAAC,EAAE,CACvG,CAAC;IACJ,CAAC;AACH,CAAC;AAED,MAAM,UAAU,8BAA8B,CAAC,MAK9C;IACC,MAAM,EAAE,SAAS,EAAE,YAAY,EAAE,QAAQ,EAAE,MAAM,EAAE,GAAG,MAAM,CAAC;IAC7D,MAAM,iBAAiB,GAAG,wBAAwB,CAAC,YAAY,CAAC,CAAC;IAEjE,IAAI,UAAU,GAAG,QAAQ,CAAC,UAAU,CAAC;IACrC,IAAI,UAAU,GAAG,QAAQ,CAAC,UAAU,CAAC;IACrC,IAAI,mBAAmB,GAAG,QAAQ,CAAC,mBAAmB,CAAC;IAEvD,MAAM,cAAc,GAAG,CAAC,UAAU,IAAI,SAAS,CAAC,QAAQ,CAAC,QAAQ,KAAK,QAAQ,CAAC;IAC/E,MAAM,cAAc,GAAG,CAAC,UAAU,IAAI,SAAS,CAAC,QAAQ,CAAC,QAAQ,KAAK,QAAQ,CAAC;IAC/E,MAAM,oBAAoB,GACxB,CAAC,mBAAmB,IAAI,SAAS,CAAC,YAAY,CAAC,QAAQ,KAAK,QAAQ,CAAC;IAEvE,IACE,CAAC,cAAc,IAAI,cAAc,IAAI,oBAAoB,CAAC;QAC1D,CAAC,iBAAiB,CAAC,eAAe;YAChC,iBAAiB,CAAC,eAAe;YACjC,iBAAiB,CAAC,qBAAqB,CAAC,EAC1C,CAAC;QACD,MAAM,CAAC,IAAI,CAAC,oCAAoC,CAAC,CAAC;QAElD,IAAI,cAAc,IAAI,iBAAiB,CAAC,eAAe,EAAE,CAAC;YACxD,MAAM,EAAE,MAAM,EAAE,UAAU,EAAE,GAAG,SAAS,EAAE,GAAG,YAAY,EAAE,GAAG,IAAI,EAAE,CAAC;YACrE,UAAU,GAAG,IAAI,SAAS,CACxB;gBACE,MAAM,EAAE,iBAAiB,CAAC,eAAe;gBACzC,GAAG,SAAS;aACb,EACD,MAAM,CACP,CAAC;QACJ,CAAC;QAED,IAAI,cAAc,IAAI,iBAAiB,CAAC,eAAe,EAAE,CAAC;YACxD,MAAM,EAAE,MAAM,EAAE,UAAU,EAAE,GAAG,SAAS,EAAE,GAAG,YAAY,EAAE,GAAG,IAAI,EAAE,CAAC;YACrE,UAAU,GAAG,IAAI,SAAS,CACxB;gBACE,MAAM,EAAE,iBAAiB,CAAC,eAAe;gBACzC,KAAK,EAAE,OAAO;gBACd,KAAK,EAAE,wBAAwB;gBAC/B,cAAc,EAAE,KAAK;gBACrB,GAAG,SAAS;aACb,EACD,MAAM,CACP,CAAC;QACJ,CAAC;QAED,MAAM,eAAe,GAAG,iBAAiB,CAAC,qBAAqB,CAAC;QAChE,IAAI,oBAAoB,IAAI,eAAe,EAAE,CAAC;YAC5C,mBAAmB,GAAG;gBACpB,EAAE,EAAE,QAAQ;gBACZ,aAAa,EAAE,CAAC,EAAE,MAAM,EAAE,aAAa,EAAE,QAAQ,EAAE,MAAM,EAAE,EAAE,EAAE,CAC7D,IAAI,kCAAkC,CAAC;oBACrC,MAAM,EAAE,eAAe;oBACvB,MAAM,EAAE,aAAa;oBACrB,kBAAkB,EAChB,YAAY,EAAE,0BAA0B;wBACxC,2CAA2C;oBAC7C,GAAG,CAAC,QAAQ,CAAC,CAAC,CAAC,EAAE,QAAQ,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;oBACjC,GAAG,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,MAAM,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;oBAC7B,aAAa,EAAE,IAAI;iBACpB,CAAC;aACL,CAAC;QACJ,CAAC;IACH,CAAC;SAAM,IAAI,cAAc,IAAI,cAAc,IAAI,oBAAoB,EAAE,CAAC;QACpE,MAAM,CAAC,IAAI,CAAC,oEAAoE,CAAC,CAAC;IACpF,CAAC;IAED,OAAO;QACL,UAAU;QACV,UAAU;QACV,mBAAmB;KACpB,CAAC;AACJ,CAAC"}
|