OraClaw -- deterministic decision tools that complement LLM routing

[Whatsonyourmind]

Hey LiteLLM community! Sharing tools that complement LLM routing with deterministic decision-making.

The connection to LiteLLM

LiteLLM excels at routing requests across LLM providers. OraClaw adds a complementary layer: deterministic algorithmic decisions that don't need an LLM at all.

OraClaw provides 19 mathematical algorithms as a REST API and MCP server. When your application needs to make optimization decisions, the algorithms handle it in sub-25ms with exact, reproducible results -- saving LLM tokens for where they actually matter.

Relevant use cases with LiteLLM

Algorithm	LiteLLM Use Case
Multi-Armed Bandit	Optimize provider/model selection over time based on cost/quality
Anomaly Detection	Flag unusual latency, error rates, or cost patterns across providers
Time Series Forecast	Predict future token usage, costs, or traffic patterns
Calibration Score	Evaluate quality of model outputs across providers
Portfolio Risk	Assess risk of provider concentration

Setup

REST API (no auth needed):

curl -X POST https://oraclaw-api.onrender.com/api/v1/optimize/bandit \
  -H "Content-Type: application/json" \
  -d '{"arms": [{"id":"gpt4","pulls":100,"totalReward":88},{"id":"claude","pulls":80,"totalReward":76},{"id":"gemini","pulls":60,"totalReward":50}], "algorithm":"ucb1"}'

MCP Server: npx @oraclaw/mcp-server

MIT licensed, 1,072 tests.

Links: GitHub | npm | Live API