4175e2070c
feat: migrate to typed messages, drop base64
...
CI / Lint (pull_request) Failing after 58s
CI / Test (pull_request) Failing after 1m22s
CI / Notify (pull_request) Successful in 1s
CI / Release (pull_request) Has been skipped
CI / Docker Build & Push (pull_request) Has been skipped
- Switch OnMessage → OnTypedMessage with natsutil.Decode[messages.ChatRequest]
- Return *messages.ChatResponse / *messages.ChatStreamChunk (not map[string]any)
- Audio as raw []byte in msgpack (25% wire savings vs base64)
- Remove strVal/boolVal/intVal helpers
- Add .dockerignore, GOAMD64=v3 in Dockerfile
- Update tests for typed structs (9 tests pass)
2026-02-20 07:10:43 -05:00
609b44de83
feat: add e2e tests + benchmarks, fix config API
...
- e2e_test.go: full pipeline tests (LLM-only, RAG, TTS, timeout)
- main.go: fix config field->method references (EmbeddingsURL() etc.)
- Benchmarks: LLMOnly 136µs/op, RAGFlow 496µs/op
2026-02-20 06:45:21 -05:00
adcdb87b9a
feat: rewrite chat-handler in Go
...
Replace Python chat handler with Go for smaller container images.
Uses handler-base Go module for NATS, health, telemetry, and service clients.
- RAG pipeline: embed → Milvus → rerank → LLM
- Streaming response chunks
- Optional TTS synthesis
- Custom response_subject support for companions-frontend
2026-02-19 17:58:52 -05:00
