Skip to content
RawbeewPublic

About

Multi-provider LLM failover router + agent harness. Adaptive provider ordering, retry w/ backoff, semantic cache, quota-aware routing, key rotation, 9 guarded tools, agent benchmark suite. Stdlib-only core. 173 tests.

Topics

Resources

Security policy

Stars

0 stars

Watchers

0 watching

Forks

Latest commit

 

History

75 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Flippy

Flippy routes language-model requests across providers and offers an OpenAI-compatible API. It includes retries, quota tracking, key rotation and an agent with configurable tools.

Get started

Requires Python 3.11 or later. Set a provider key in your environment, then start the server:

export GROQ_KEY="your-api-key"
python src/server.py

The server listens on 127.0.0.1:8080 by default. See .env.example for supported provider settings; the server reads environment variables and does not automatically load that file.

curl http://127.0.0.1:8080/v1/models

curl http://127.0.0.1:8080/v1/chat/completions \
  -H 'Content-Type: application/json' \
  -d '{"messages":[{"role":"user","content":"Hello"}]}'

Set FLIPPY_AUTH_TOKEN to require a bearer token. A non-loopback server requires authentication; keep keys out of source files and version control.

Providers and costs

Flippy supports Groq, NVIDIA NIM, OpenRouter, FreeInference, Cloudflare Workers AI, Ollama and other compatible endpoints. Available models and quotas depend on the provider and account. Pinning a model limits routing to providers configured for that model.

Flippy does not guarantee free usage. Check your provider plan, model pricing, limits and terms before enabling a route. A free-tier label in the code is not a billing control.

Hermes

Flippy exposes an OpenAI-compatible endpoint that can be configured as a custom provider in Hermes. Choose an exact model ID from /v1/models and verify its availability before use. A configured model does not by itself prove that inference, tool calls or failover will work with your provider.

Agent

python -m src.loomweaver agent "Summarize the files in this directory"

Tools are scoped to the task; grant additional tools explicitly with --tools when needed. See ARCHITECTURE.md and SECURITY.md for details.

License

MIT

About

Multi-provider LLM failover router + agent harness. Adaptive provider ordering, retry w/ backoff, semantic cache, quota-aware routing, key rotation, 9 guarded tools, agent benchmark suite. Stdlib-only core. 173 tests.

Topics

Resources

Security policy

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages