Summary

coding-agent-session-search (cass) unifies and indexes chat histories from numerous AI coding agents into a single, high-performance, local-first TUI. It enables developers to quickly search across fragmented agent data, surfacing past solutions and institutional knowledge, thereby preventing redundant problem-solving and enhancing productivity.

How It Works

cass employs a dual-storage architecture: a normalized SQLite database acts as the source of truth, while a Tantivy full-text index provides sub-60ms search latency. Edge n-grams are pre-computed for rapid prefix matching, and connectors normalize diverse agent data formats (JSONL, SQLite, Markdown) into a unified schema. An interactive TUI or a machine-readable CLI facilitates querying, filtering, and result exploration, with optional local semantic search available via FastEmbed.

Quick Start & Requirements

Installation is handled via a provided script:

• Linux/macOS: curl -fsSL https://raw.githubusercontent.com/Dicklesworthstone/coding_agent_session_search/main/install.sh | bash -s -- --easy-mode --verify
• Windows (PowerShell): irm https://raw.githubusercontent.com/Dicklesworthstone/coding_agent_session_search/main/install.ps1 | iex; install.ps1 -EasyMode -Verify No non-default prerequisites are explicitly listed for core functionality.

Highlighted Details

• Sub-60ms Search Latency: Achieved via edge n-gram indexing, prefix caching, and optimized Tantivy queries.
• Optional Local Semantic Search: Integrates with FastEmbed and MiniLM models for meaning-based retrieval without network calls.
• Rich Terminal UI (TUI): Features a three-pane layout, live indexing status, multi-open queues, and extensive keyboard navigation.
• Multi-Machine Search: Supports aggregating session data from remote machines via SSH/rsync.
• AI/Automation Mode: Designed for programmatic use with forgiving syntax, structured JSON output, and self-documenting API.
• Broad Agent Support: Connectors for Codex, Claude Code, Gemini CLI, Cursor, ChatGPT, Aider, Pi-Agent, and more.

Maintenance & Community

The README does not detail specific contributors, community channels (like Discord/Slack), or a public roadmap.

Licensing & Compatibility

Released under the MIT License, permitting commercial use and integration into closed-source projects.

Limitations & Caveats

Encrypted ChatGPT v2/v3 conversations are not indexed due to encryption key access limitations. Performance may degrade significantly with over 500,000 messages. Semantic search requires manual download of model files.