Add llm-async

[sonic182]

Project

llm_async

Checklist

• One project per PR
• PR title format: Add llm-async
• Entry format: * [llm_async](https://github.com/sonic182/llm-async) - ... .
• Description is concise and short

Why This Project Is Awesome

Which criterion does it meet? (pick one)

• Industry Standard — The go-to tool for a specific use case
• Rising Star — 5000+ stars in < 2 years, significant adoption
• Hidden Gem — Exceptional quality, solves niche problems elegantly

Explain:
llm_async elegantly solves the pain of building truly async LLM clients in Python: unified multi-provider interface (OpenAI, Claude, Google Gemini, and OpenRouter), real end-to-end streaming over asyncio, tool-calling with consistent round-trips, and structured outputs with JSON schema (where supported), all focused on low latency via aiosonic.

How It Differs

If similar entries exist, what makes this one unique?

• Async-first + real performance: built on aiosonic (low-overhead async HTTP) and designed for streaming and fast round-trips.
• Single API for multiple providers: same messages/tools/streaming pattern for OpenAI, Claude, Gemini, and OpenRouter.
• Integrated tool-calling round-trips: helpers to execute tool calls and forward results in the next turn consistently.
• Structured Outputs with schema validation: supported by OpenAI, Gemini, and OpenRouter (not Claude).
• Production-ready extras: HTTP/2 support in the underlying client and request() for calling direct endpoints (GET/POST/etc) on any provider.
• OpenAI Responses API + prompt caching: support for previous_response_id + prompt_cache_key for multi-turn conversations without resending full history.