agenticraft-llm

Production-grade LLM provider abstraction with intelligent routing, resilience, and cost optimization.

Why agenticraft-llm?

• 14 providers, one interface — OpenAI, Anthropic, Google, Azure, Bedrock, and 9 more
• Cost-aware routing — Thompson sampling finds the best cost/quality trade-off
• Built-in resilience — Circuit breakers, rate limiters, exponential backoff
• Key rotation — 5 strategies with automatic health tracking
• OpenAI-compatible gateway — Drop-in API server with SSE streaming
• Typed responses — LLMResponse with usage, cost, and latency

Quick Install

uv add "agenticraft-llm[openai]"

Quick Example

from agenticraft_llm import OpenAIProvider

provider = OpenAIProvider(api_key="sk-...")
response = await provider.complete([{"role": "user", "content": "Hello!"}])
print(response.content)

See the Quick Start guide for more.