CoT Reasoning Service

Chain-of-Thought reasoning service with multi-stage pipeline orchestration.

Overview

This service provides configurable multi-stage reasoning pipelines for LLM-based analysis tasks. It abstracts the CoT pattern into a reusable service that can be consumed by other applications.

Architecture

cot-reasoning/
├── service/src/          # Python FastAPI service
│   ├── api/main.py       # FastAPI endpoints
│   ├── config.py         # Configuration (uses service-addresses)
│   ├── reasoning/        # CoT engine and stages
│   └── llm/              # LLM client adapter
├── packages/client/      # @lilith/cot-client TypeScript package
└── pyproject.toml        # Python package definition

Port

Port 8110 is allocated in lilith-platform/infrastructure/ports.yaml under ml.cot-reasoning.

The service uses lilith-service-addresses for runtime port resolution.

Installation

Python Service

pip install cot-reasoning

TypeScript Client

npm install @lilith/cot-client

Usage

Python Service

# Start the service
python -m service

TypeScript Client

import { CoTClient } from '@lilith/cot-client';

// Auto-discovers port from service-addresses
const client = await CoTClient.create();

const result = await client.reason({
  input: 'femboy, catgirl, lawyer',
  stages: ['cultural_analysis'],
});

console.log(result.result);
console.log(result.reasoning_trace);

API Endpoints

POST /reason

Execute reasoning pipeline on input.

{
  "input": "text to analyze",
  "stages": ["stage1", "stage2"],
  "context": {}
}

GET /stages

List available reasoning stages.

GET /health

Health check endpoint.

Dependencies

• lilith-ml-service-base - FastAPI scaffolding
• lilith-ollama-provider - LLM client
• lilith-pipeline-framework - Pipeline orchestration
• lilith-service-addresses - Port resolution

License

MIT