# CC-RLM: Self-Improving Context Engine

> 70–80% fewer tokens, learned per session.

**Status:** Prototype

## Problem

Dumping a whole repo into the context window is slow, expensive, and noisy.

## What it does

A proxy layer for AI coding agents that replaces naive full-repo context injection with a live structural model (import graph, symbol index, diff state) and builds a sub-8K-token context pack per request. It learns which files matter by parsing which symbols the model actually cites, biasing future context toward them.

## Measured results

- **70–80%** token reduction
- **90%** recall
- **<200ms** latency

## Stack

Python, FastAPI, local LLM (Ollama), SQLite, BM25, AST walkers, Docker

## Topics

AI, Agents, Architecture

---

Part of the proof-of-work portfolio of Michael Wahl — VP, Software Engineering & AI Enablement.

- Canonical project page: https://connect.michaelwahl.org/project/cc-rlm-context-engine
- Full professional profile: https://connect.michaelwahl.org/
- Machine-readable context: https://connect.michaelwahl.org/llms.txt