lovstudio/Ataru

What it solves

Ataru provides a high-performance local memory retrieval system for AI conversations. It solves the problem of valuable information—such as troubleshooting steps, critical commands, and architectural decisions—becoming lost across fragmented local session files from various AI agents (like Claude and Codex), making it difficult to relocate specific context without manual searching.

How it works

Ataru creates a unified, searchable index of local AI session transcripts. It uses a modular architecture built with Rust and Tauri, employing Tantivy and Jieba for full-text keyword search and an optional SQLite-based vector store for semantic retrieval. The system normalizes diverse session formats into a consistent hierarchy of Projects, Sessions, Runs, and Turns. It operates as a local-first service that can be accessed via a desktop GUI, a JSON CLI, or as "Agent Skills" that allow other AI agents to programmatically search and retrieve historical context.

Who it’s for

It is designed for developers and power users who use multiple AI agents for coding and project management and need a reliable way to recover specific historical context and decisions from their local conversation logs.

Highlights

• Unified Indexing: Normalizes session formats from sources like Claude CLI, App, and Web, and Codex.
• Hybrid Search: Combines keyword search (default/offline) with optional semantic recall using RRF fusion.
• Granular Retrieval: Allows switching between Turn, Run, Session, and Project levels of granularity to find the exact piece of information needed.
• Agent-Ready: Provides a stable API and dedicated "Skills" so AI agents can autonomously search their own history.
• Local-First Privacy: Runs offline by default; original transcripts are read-only and never uploaded to the cloud unless a remote embedding provider is explicitly configured.