Jargo

Changelog

• 34366d4dbc205b58e34ed3bec19d131806b97a46: Initial scaffold: jargo, a Go port of Pipecat (@fallais)
• cc5b923608249f2948f4c75f4523a102e0ffa97c: Merge pull request #1 from gojargo/security/fix-scanning-alerts (@fallais)
• 26e7525d0f5f625671323afa59b16cf549b540de: Merge pull request #2 from gojargo/feature/services (@fallais)
• 138524f8f362a2bd833165046a2ebe5edbc02af0: Merge pull request #3 from gojargo/feat/turns-providers-telephony (@fallais)
• 07f4f8b74a18cd291ee27e451ceb6f784332edeb: anthropic: add base URL, prompt-caching toggle, and extra request fields (@fallais)
• 9e341410cfdf8d802d282861eae4f2833112ce5c: assemblyai: expand STT config to full v3 streaming option surface (@fallais)
• b8cf872646cc3801227dca17ef6cfe8c77295abd: benchmarks: move the comparison to gojargo/jargo-benchmarks (@fallais)
• a61acd979adb1dcd77fdc01f9b5dd7f395acd1b0: cartesia: add language, generation config, audio format, version/URL (@fallais)
• dc2415cdf16668e0190bb08839bc1519c0676ce3: chore(release): prepare v0.0.1 (@fallais)
• 9dfa9156d1bf2905a214dbe417976394ce7c36fc: deepgram/lmnt/rime: make TTS sample rate configurable + Rime controls (@fallais)
• f14e303ff884cb53858d532520eea8f3898403ca: deepgram: expand STT config to full Deepgram option surface (@fallais)
• 83533f1a153820901c7a4b9b76126f579673fe0f: deps: update pion/opus to latest (CELT encoder fixes) (@fallais)
• b6ebf84cfb06361c813d9df321c25c24418653dd: elevenlabs: add base URL, latency/normalization/logging, pronunciation dicts (@fallais)
• 9b780283b68a8970e3115669605d9ab61e95dd80: elevenlabs: output_format mapping + native 48kHz, voice settings, language (@fallais)
• 4f302ff839b959a24b7532a408851217d02122f7: examples: fix gci import order after service->provider move (@fallais)
• b8dd439b3a257135c50f708f51293b026cabe9c4: feat(examples): per-provider voice examples; trim voicebot to a fixed stack (@fallais)
• 8ef69ed31357510fbd28d87a5d7aaefdbeee9ba1: feat(llm): add function/tool calling (@fallais)
• cad36795fa84c4c902bd044a268f9020e683eed3: feat(llm): add usage metrics, sampling params, and request timeout/retry (@fallais)
• 9541eb694c1617f9989b0b0a7da450e4f95c6f3b: feat(pipeline): add ParallelPipeline (@fallais)
• 93e37ce148db6b51b486724b4bddbf2f18179084: feat(pipeline): add ServiceSwitcher with manual and failover strategies (@fallais)
• 32b348bf0da82f1d28922c6e7b7835d07c9207c9: feat(resample): use libsoxr via cgo for high-quality resampling (@fallais)
• ae8b3294f9956d064d20b334e088002ea198093c: feat: OpenTelemetry tracing (@fallais)
• 00ba5b177a2762a92bb3a6df55f566f3cc71e4c6: feat: Pipecat turns/ subsystem, new providers, Twilio telephony, OSSF scorecard (@fallais)
• 5c6d31aadd308990c02a1d1c2d3cd92a33cf1ad0: feat: add Azure OpenAI and AWS Bedrock LLM providers (@fallais)
• 95648e0d02dcd0729559d9c5cef1b87b55a1d63b: feat: add OpenAI Realtime speech-to-speech service (@fallais)
• 0dfb098e79c3156e9c33a26c8cf723768db6b07f: feat: add local STT/TTS providers (whisper.cpp, Kokoro, Piper) (@fallais)
• 1d268b372d8141e3a9a929c19682243200dcebea: feat: automatic LLM context summarization (@fallais)
• f8e251386fb72eae88154f836a48c863c3d0e622: feat: broaden service catalog with shared LLM/STT/TTS bases (@fallais)
• e58d9fb4496c3f3f4dbf5de5d46dcece63dc04e1: feat: mem0 long-term memory (@fallais)
• 13cdd32bdbbfe654e8ef9aa82d2e42a8dc14c068: feat: pipeline metrics (in-band + OpenTelemetry) (@fallais)
• 4d6126c0ed5aa0f5fbee5a7d7f88a837f789e889: feat: turn-taking (Silero VAD + Smart Turn V3), benchmarks, docs (@fallais)
• 816c49d94443ff6de9e139a07ef55d912711ea1e: feat: voice pipeline engine, WebRTC transport, RTVI, and STT/LLM/TTS (@fallais)
• 9a42b05bc72e28dd42f75896e15d230fdc093b7b: frames: add TTSSpeakFrame + LLMRunFrame, and wire bot-tts-text (@fallais)
• e6a9ead77c4f64df30f0f5ccda7ae4e2c7635cbf: gladia: expand STT config to full session-init option surface (@fallais)
• d4a09ce5ca7fcf4ac02171413bda7907bbadaa31: google: add temperature, top_p, top_k, and extra generationConfig fields (@fallais)
• 664ee0a1d9b5f0ddde17c7b120e6e3ab14a28f92: google: implement tool-calling on the Gemini LLM (@fallais)
• c79a0acfaa36ed617ff722bdeaa115adddcea885: language: canonical Language enum + per-provider mappers (@fallais)
• aa1c7def824ed1205e35a618e34f4b5772867a19: lint: fix all golangci-lint findings (CI green) (@fallais)
• e87e185e4efab80261dc688cbc2e144e00dd88ca: llm-wrappers: align default models with current Pipecat defaults (@fallais)
• 2c0a543dd7dfc1aeb92d6da74f1272da460d01cb: openai: add STT language/prompt/temperature and TTS instructions/speed (@fallais)
• 4964c3c4df5aca47f052bb27f22210e53e0c2fdb: openai: add sampling params + extra fields to OpenAI-compatible LLM (@fallais)
• 409a008cb32aac561fea167e37a373efe28da3ff: openai: implement tool-calling on the OpenAI-compatible LLM (@fallais)
• c6483f988ca5a5889e2d286fedbc26e87bf01ac1: opus: add libopus (cgo) encoder behind a build tag (@fallais)
• 033152df5b62e20eb1812d36c88306d6fb04dfc0: perf(turn): replace DFT with gonum FFT and sparse mel filterbank (@fallais)
• 5d5fa513c1d1440bb0e15d834c5ba03a9ae3eb83: pionrtc: pace audio output to real time (@fallais)
• 887cec92d6c4374678861e45723bb1f1e899dbb4: refactor: explicit config + validation; Viper example (@fallais)
• 828c9be7b5bf60d775994d53598d7f177f4aad37: refactor: split service/ into service contracts + provider implementations (@fallais)
• f8d1003d487c02f30c87ba83065df00c0d1e3e8c: security: clear code-scanning alerts (deps, action pinning, Dockerfile) (@fallais)
• f0d1d1e0a146aff838847b46c0fa0a366ce8508d: test(benchmarks): add pipeline plumbing and concurrency benchmarks (@fallais)