v1.0.0

1.0.0 (2026-05-17)

Documentation

• add COMPATIBILITY.md — declare v1.0 API stability (054fa38)

v0.8.0

0.8.0 (2026-05-17)

Features

• providers/vertexanthropic + providers/azureserverless (cb04731)
• router: magic model x platform resolution (63dde6d)

Bug Fixes

• azureanthropic: drop unused fmt import + lint issues (a238f58)
• deepgram: tolerate variant channel shapes (closes #9) (dbfe592)

v0.7.0

0.7.0 (2026-05-16)

Features

• azureanthropic: Claude on Azure AI Foundry (0b90b4a)

v0.6.0

0.6.0 (2026-05-16)

⚠ BREAKING CHANGES

• forward non-transcript STT events; add Type + SpeechFinal

Features

• forward non-transcript STT events; add Type + SpeechFinal (7fc0b30)

v0.5.0

0.5.0 (2026-05-16)

Features

• cohere,voyage,together: Rerank for RAG retrieval refinement (e0a0e81)
• fireworks,deepseek: embeddings via OpenAI-compatible delegation (eba469b)
• foundation for v0.5 — Reranker, ResponseSchema, Realtime tool use (e4833ba)
• geminilive: Gemini Live — bidirectional WS for low-latency voice agents (eea620a)
• openai,anthropic,vertex: structured outputs via ChatRequest.ResponseSchema (065433d)

v0.4.0

0.4.0 (2026-05-16)

Features

• anthropic: NewRecommendedEmbedder — Voyage AI shim for Claude users (7b3519b)
• cartesia: SpeakRealtime — WebSocket TTS with multi-turn context (e087263)
• deepgram: WebSocket streaming STT via wss://api.deepgram.com/v1/listen (3ef9c6d)
• elevenlabs: SpeakRealtime — WebSocket TTS via /stream-input (1928f78)
• openairealtime: OpenAI Realtime API (gpt-4o-realtime, WebSocket) (425b42a)
• typed tool-result message (Message.ToolCallID + ToolResultMessage helper) (0594af0)

Bug Fixes

• drop unused struct fields flagged by staticcheck (331c0eb)

v0.3.0

0.3.0 (2026-05-15)

Features

• add Speaker, Transcriber, Embedder interfaces + root types (a8482c3)
• bedrock: embeddings via InvokeModel for Titan + Cohere (de3ad80)
• cartesia: TTS (Speaker) for real-time voice agents (10d7301)
• cohere,mistral,together: embeddings (e28ba12)
• deepgram: STT via Nova-3 (Transcriber) (db17bed)
• elevenlabs: TTS (Speaker) + STT (Transcriber) (91ebbf7)
• gemini,vertex: audio (TTS+STT) + embeddings (1228a66)
• groq: Whisper STT via OpenAI-compatible audio endpoint (0881281)
• openai,azureopenai: audio (TTS+STT) + embeddings (2af970e)
• voyage: Voyage AI embeddings (Embedder) (a4a3f6a)

v0.2.0

0.2.0 (2026-05-15)

Features

• azureopenai: Azure OpenAI Service provider (91bd56d)
• bedrock: AWS Bedrock provider via Converse Stream API (ce26369)
• cohere: Cohere v2 Chat provider (4fbe4b0)
• gemini: Google AI Studio Gemini provider (HTTP, no SDK) (f6a8a02)
• mistral: Mistral AI chat provider (8db32aa)
• openrouter: OpenRouter provider with attribution headers (1fbca4b)
• providers: thin OpenAI-compatible helpers for 7 vendors (3917dec)
• tool use, thinking output, prompt caching, multimodal helpers (488f4b1)
• vertex: Google Vertex AI Gemini provider (18d0711)

Bug Fixes

• vertex: skip TestNew_Succeeds when ADC unavailable (5e3068e)

v0.1.1 — test suite (no API change)

Adds a 330-subtest table-driven test suite across the public API and internal helpers of both providers. 100% statement coverage on the root llmrouter package and 90.4% on each of the openai and anthropic providers. No source or API changes.

Install: go get github.com/elloloop/llmrouter@v0.1.1

v0.1.0 — OpenAI + Anthropic

First public release. Streaming-first Go client for LLM providers with one OpenAI-shaped API surface.

Providers:

• providers/openai — OpenAI + any OpenAI-API-compatible endpoint (Together, Groq, OpenRouter, vLLM, Ollama, self-hosted) via WithBaseURL
• providers/anthropic — full request + SSE translation Anthropic /v1/messages ↔ OpenAI delta format

Install: go get github.com/elloloop/llmrouter@v0.1.0

Roadmap (v0.2): Azure OpenAI, AWS Bedrock, Google Vertex AI.

See README.