🚀 llm-fiber v1.0.1 — First stable release

Summary

✨ A minimal, async‑first client with wire‑speed streaming, consistent ergonomics across providers, and built‑in observability. One API, multiple providers, zero boilerplate.

Highlights

• ⚡ Async‑first with clean sync mirrors
• 🔀 Multi‑provider: OpenAI, Anthropic, Gemini — auto‑routes by model name
• 🌊 Wire‑speed streaming with simple, ordered events
• 🧰 Normalized tool/function calling across providers
• 🛡️ Reliability built‑in: retries with jitter, granular timeouts, typed errors
• 💸 Cost controls: token/cost ceilings with clear failure modes
• 🧠 Caching: deterministic keys with in‑memory LRU+TTL
• 🧵 Batch helpers: concurrency caps, fail‑fast/return‑exceptions, summaries
• 🔭 Observability: structured logs + metrics (latency/TTFB/retries/tokens/cache)
• 🧩 Optional extras: OpenTelemetry, StatsD, structlog

Install

• Base: pip install llm-fiber
• Extras: pip install "llm-fiber[otel]", "[statsd]", "[structlog]", or "[all]"
• Python: 3.8–3.12

Docs

• 📚 Full docs & feature guides: https://github.com/aimlesx/llm-fiber/blob/main/docs/index.md
• 🚦 Quickstart in README: https://github.com/aimlesx/llm-fiber

Notes

• ✅ First stable release; public APIs targeted for long‑term stability
• 🔒 Use only the env vars you need: OPENAI_API_KEY, ANTHROPIC_API_KEY, GEMINI_API_KEY
• 💡 Quick taste: Fiber.from_env().sync.ask("Hello!", model="gpt-4o-mini")