Skip to content
Yistar.
← All Projects
Active development
Local-firstSession ArchiveDashboard

CodexJournal-Lite

Making AI sessions searchable work memory

GitHub

01 / Context

Why It Exists

When using multiple AI coding tools daily, session content is scattered across different tools' local storage. It's hard to recall 'the approach we discussed last time' or 'what the agent did before.' CodexJournal-Lite consolidates these fragments into a searchable local knowledge base.

Problem

AI coding tool session records are scattered in different locations with various formats. Without a unified search and review mechanism, important context is easily lost. Repeated discussions and tasks lack records, reducing efficiency.

Solution

CodexJournal-Lite automatically scans and archives session records from multiple AI tools, parsing them into unified formats with full-text search, task extraction, and dashboard visualization. Outputs structured Markdown/JSON; archive/analysis/verification/dashboard make no external service calls.

02 / System

Core Capabilities

  • Auto-archive local records from Codex, Claude Code, Gemini CLI, OpenCode, JetBrains/IDEA
  • Output structured Markdown/JSON to journal/, data/tasks.json, reports
  • Full-text search across session history
  • Localhost-only dashboard (127.0.0.1:7777) showing work timeline and stats
  • Browse by project, date, and tool

03 / Workflow

Workflow

  1. Configure AI tool session directories to monitor
  2. Auto-scan and archive new session records
  3. Browse history via search or categories
  4. View dashboard for work timeline
  5. Extract important tasks and decisions to local notes

04 / Boundaries

Security & Privacy

All data stored locally on the filesystem; no network, no uploads. Archive/analysis/verification/dashboard make no external service calls. No upload, no telemetry, zero third-party npm dependencies. Personal output must not enter public repositories.

Current Limitations

  • Depends on each tool's local log format and directory
  • Gemini/OpenCode may not be enabled by default
  • JetBrains is currently an inventory probe
  • Search is full-text index based; no semantic search

05 / Roadmap

Roadmap

01

Completed

  • Multi-tool session auto-archival
  • Structured Markdown/JSON output
  • Full-text search
  • Localhost-only dashboard
02

In Progress

  • Task extraction and tagging system
  • More tool format adaptation
03

Exploring

  • Semantic search support
  • Integration with note-taking tools

06 / Participate

How to Participate

Published as an npm package, available on GitHub and npm. Try it out and share feedback. If your AI tool's format is not supported, submit a format sample in an issue.

GitHub