sly-codechum

sly-codechum / chum-mem

Public

Better memory for your AI agents. (Karpathy + Graphify + PCKC)

17
1
100% credibility
Found Apr 18, 2026 at 17 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Rust
AI Summary

chum-mem builds dual knowledge graphs from code repositories and AI coding sessions to enable precise, proof-backed memory retrieval for better agent performance.

How It Works

1
🔍 Discover chum-mem

You find chum-mem, a smart memory helper that makes your AI coding buddy remember your project's code structure and past chats perfectly.

2
🚀 Start the memory service

With one easy command, you launch the background service that watches your code and conversations.

3
🔌 Connect to your AI tool

You add a simple plugin to your favorite AI coder like Claude or Gemini so it can use the memory.

4
💭 Code and chat naturally

As you work with your AI, it automatically builds a map of your code, decisions, and sessions without you lifting a finger.

5
📊 Explore your memory map

Open the web view to see a 3D graph of your code relationships, claims, and session history.

6
🔍 Search and recall instantly

Ask your AI or search the dashboard to pull up exact facts, fixes, or past ideas with proof.

✅ AI remembers like a teammate

Your coding sessions flow smoothly as the AI recalls code details, decisions, and history, saving time and reducing errors.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 17 to 17 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is chum-mem?

chum-mem builds structured memory for AI coding agents, turning raw repo code and session events into dual knowledge graphs with verifiable claims, proofs, and temporal validity. Agents get precise answers to "what calls function X?", session continuity for resuming refactors, and contradiction-aware context packs—all via automatic hooks into Claude, Codex, or Gemini on every turn. Rust-powered Docker stack includes a 3D web dashboard for graph viz, hybrid search, and community detection.

Why is it gaining traction?

It fixes RAG's blind spots for coding: structural queries, decision history, and proof-tracked claims prevent model prose leaks into durable memory. Hybrid retrieval (full-text + vectors + graph proximity) with typed partitions delivers better memory retention than chunk-based similarity. Benchmarks hit 83% pass rate on agent tasks, with sub-100ms latencies even on 68k-node graphs.

Who should use this?

AI agent power users in Claude.dev or Cursor, refactoring large codebases across 19 languages. Teams needing better memory for OpenClaw agents, where session history and repo structure must persist without manual prompts. Devs frustrated by agents forgetting prior decisions or hallucinating call graphs.

Verdict

Promising for better AI agent memory (beats basic RAG), with easy Docker setup, plugin hooks, and rich dashboard—but low maturity (17 stars, 1.0% credibility score) means test on non-critical projects first. Strong docs and benchmarks make it worth a spin.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.