v1.2.0

Local LLM Provider Support

Add 6 local LLM inference engines as first-class providers, all using OpenAI-compatible HTTP APIs:

Provider	Chat	Completions	Embeddings	Rerank	Audio	Images
Ollama	✅	✅	✅	—	—	—
LM Studio	✅	✅	✅	—	—	—
vLLM	✅	✅	✅	✅	—	—
llama.cpp	✅	✅	✅	✅	—	—
LocalAI	✅	✅	✅	✅	✅	✅
llamafile	✅	✅	✅	✅	—	—

Quick Start

from liter_llm import LlmClient

client = LlmClient(api_key="")
response = await client.chat(
    model="ollama/llama3.2",
    messages=[{"role": "user", "content": "Hello!"}],
)

No API key needed. Model routing uses the standard provider/model prefix convention.

What's New

• 6 local providers with accurate endpoint configurations
• Docker Compose setup for integration testing (Ollama + qwen2:0.5b + all-minilm)
• CI workflow for local LLM integration tests
• Comprehensive documentation at docs/usage/local-llms.md
• Integration tests for chat, streaming, embeddings, and model listing
• Removed duplicate lm_studio and ollama_chat entries (now 143 providers)

Fixed

• PHP onError hook passes Exception not string
• README template compliance for rumdl
• Added 404 to POST OpenAPI specs

Full Changelog: v1.1.1...v1.2.0