easy-agent-mem 0.5.2

CLI-first persistent memory for AI coding sessions with direct Obsidian note writing and local fallback storage.

agent-mem

Automatic context compression and persistent memory for AI coding agents

agent-mem watches your work, detects meaningful progress, and gives you a one-paste handoff prompt that tells your IDE agent to summarize, save memory, and start fresh with minimal context.

---

Features

• Smart watch mode with file + git + idle detection
• One-paste handoff prompts (Groq powered, optional)
• Rich Obsidian-first storage with wiki-links and frontmatter
• Local fallback mode (.agent-memory/)
• New: graph command - builds an Obsidian-native knowledge graph from code, chat history, and memory

---

Badges

---

Quick Start

pip install easy-agent-mem

agent-mem init
agent-mem configure-groq          # optional but recommended
agent-mem watch                   # start automatic handoff mode

New in v0.5+:
agent-mem graph build -> creates a full knowledge graph in agent-mem-output/

---

New: Knowledge Graph (agent-mem graph)

agent-mem graph build              # Basic deterministic graph
agent-mem graph build --enrich     # Enrich with Groq (inferred relationships)
agent-mem graph build --compact    # Compact mode for large projects

Output: Clean Obsidian-ready Markdown files in agent-mem-output/

• Index.md - beautiful dashboard
• Code structure, decisions, blockers, concepts, sessions
• Clear EXTRACTED vs INFERRED labeling with confidence scores

Open agent-mem-output/Index.md in Obsidian for full graph navigation and backlinks.

---

Commands Overview

Command	Description
agent-mem init	Setup Obsidian / fallback + IDE rules
agent-mem configure-groq	Set Groq API key
agent-mem watch	Automatic one-paste handoff watcher
agent-mem graph build	Build knowledge graph (new)
agent-mem summarize	Manual summary
agent-mem recall <query>	Search memory

---

Project Links

• PyPI: https://pypi.org/project/easy-agent-mem/
• GitHub: https://github.com/atharvavdeo/agent-mem

---

Made for developers who want their AI agent to actually remember things.

---

License: MIT

---

How to Update README

1. Open your README.md
2. Replace the entire content with the text above
3. Save and commit:

git add README.md
git commit -m "docs: update README with badges and graph feature"
git push

---

What Part Requires Improvement (Specific Feedback)

Current State (Good):

• Core graph building works
• Compact mode and enrichment are implemented
• Output is Obsidian-friendly

Areas that still need improvement (Priority order):

1. Index.md is still a bit basic
   -> Can be made more dashboard-like with better stats layout and visual separation.

2. Error handling when Groq key is missing/invalid
   -> Currently it fails silently or shows confusing messages. Should give clear user-friendly message.

3. Documentation in README
   -> The new graph command section is missing detailed examples and flags explanation.

4. Performance on very large projects
   -> Scanning thousands of files can be slow. Compact mode helps but can be optimized further.