Agent Session Mirror

CodeAware Studio: Intelligent Session Analysis for AI Coding Agents

[](https://makoy-daniot2001.github.io/agent-session-mirror/)

**Transform your coding agent's raw search sessions into actionable insights with hybrid retrieval, self-benchmarking, and cross-encoder reranking.**

[](https://opensource.org/licenses/MIT) [](https://nodejs.org) [](https://openai.com) [](https://anthropic.com) [](https://github.com)

Your Coding Agent's Sixth Sense

Imagine giving your AI coding assistant the ability to remember not just *what* it found, but *how* it found it—and to get measurably better at finding things tomorrow. That is the core philosophy behind CodeAware Studio.

While most code search tools are like a library catalog—functional but static—CodeAware Studio is more like a detective's casebook. It logs every search session, every retrieval pathway, and every filter that led your agent to the correct snippet. It then uses this historical data to **self-benchmark** and improve its own performance, creating a perpetual cycle of search intelligence.

This repository provides the high-level session observer and analyzer layer. It consumes the retrieval output from tools like `@sivru/cli` and adds a layer of deep observability. It is the difference between asking your agent "find a function" and your agent saying "I found it using this exact strategy, which was 23% more efficient than last week's attempt."

Core Architecture: The Observation Loop

The following Mermaid diagram illustrates how CodeAware Studio i