Skip to content

Releases

Project-M v0.1.1 — Ambient Context Engine, FTS5 Search & Agent Skill Installer

Choose a tag to compare

@adityadebnath009 adityadebnath009 released this 16 Nov 06:28

What's New in v0.1.1

A major update adding high-performance historical search, automated Agent Skill installation, and 102 verified tests!

Key Additions in v0.1.1

  • SQLite FTS5 with Okapi BM25 Ranking (search_context):
    • Sub-3ms lexical search across historical conversation turns from Claude, Antigravity, and Codex.
    • Porter stemming (tokenize = 'porter unicode61') and zero-lag C-level database triggers.
    • Zero cloud embedding bills or external vector databases.
  • 1-Click Agent Skill Installer:
    • project-m setup all now automatically installs the frontier SKILL.md into your global AI assistant directories (~/.claude/skills, ~/.gemini/antigravity/skills, ~/.codex/skills).
  • 8 FastMCP Tools Registered:
    • get_project_board, record_decision, update_task, publish_context, get_recent_context, get_turn, get_project_summary, and search_context.
  • 102 Verified Tests Passing:
    • 100% green automated test suite covering SQLite WAL concurrency, FTS5 BM25 search, and multi-agent end-to-end workflows.

Quick Installation

1. Instant Run (Zero-Install):

uvx --from git+https://github.com/adityadebnath009/Project-M.git project-m

2. Auto-Config Claude Desktop, Antigravity & Codex:

git clone https://github.com/adityadebnath009/Project-M.git
cd Project-M
uv sync
uv run project-m setup all

3. Install Globally as Native CLI:

uv tool install git+https://github.com/adityadebnath009/Project-M.git

🛡️ Verified & Local-First

  • 100% Local SQLite WAL: Zero cloud subscriptions, zero data leaks.
  • M8ven MCP Trust Index: Vetted for local context isolation and clean stdio transport.
  • MIT License: Free and open-source for personal and commercial workflows.