410979729

410979729 / scope-recall

Public

Hermes current-turn memory provider with SQLite truth storage, LanceDB semantic companion, and strong scope isolation.

agent ai-agent ai-memory hermes hermes-agent
10
0
89% credibility
Found May 18, 2026 at 14 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

Scope-recall is a memory plugin for the Hermes AI assistant that gives your AI a reliable, persistent memory. It stores important facts you share (preferences, project details, operations notes) permanently across all your conversations, while keeping temporary chat notes isolated to the current session. The system uses two storage layers: a local SQLite database as the authoritative record and an optional semantic search companion for finding memories by meaning. When you chat, the AI automatically recalls relevant memories from your past conversations without injecting stale information from unrelated topics. Everything stays on your computer—your memories are never sent to external services.

How It Works

1
💭 You want your AI to remember things

You've been chatting with your AI assistant and noticed it forgets important details when you start a new conversation.

2
🔌 You install the memory plugin

You add scope-recall to your Hermes AI setup. The plugin lives entirely on your computer, keeping your memories private.

3
✨ Your AI starts remembering

Now when you mention your project preferences or past decisions, the AI automatically recalls what you told it before.

4
📝 You ask the AI to remember important things

You can tell the AI to store key facts—like your preferred workflow or project details—and it remembers them permanently.

5
Your memories are organized automatically
🌟
Important memories follow you

Your preferences, project facts, and ops notes stay with you across all conversations and windows.

📄
Temporary notes stay local

General chat notes remain in the current conversation only, so they don't clutter future chats.

6
🔍 You can search your memory history

The AI searches both what you said and what things mean to find the most relevant memories for your current question.

🎉 Your AI has a reliable memory

Important information is never lost, temporary chatter stays organized, and your AI answers questions using everything you've ever told it.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 14 to 10 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 scope-recall?

scope-recall is a Python plugin for the Hermes AI agent framework that solves two persistent memory problems: the agent forgetting everything when you open a new chat, and local scratch context bleeding into unrelated conversations. It uses a dual-layer approach with SQLite as the authoritative storage and LanceDB for semantic search. Memories tagged as user preferences, project facts, or ops notes persist across windows; raw conversation captures stay locked to the current chat. The plugin hooks into Hermes at turn start, handles automatic capture with rules-based filtering, and surfaces relevant memories through a prefetch mechanism. It ships with tools for storing, searching, forgetting, updating, and exporting memories, plus maintenance utilities for deduplication and hygiene checks.

Why is it gaining traction?

The hook is scope isolation done right. Most agent memory systems either share everything globally or forget everything between sessions. scope-recall draws a deliberate boundary: durable facts follow you, scratch notes do not. The hybrid retrieval combining lexical search with vector similarity means searches catch both exact matches and semantic variations. The governance layer catches duplicates and noisy captures before they hit storage, keeping the SQLite truth store auditable.

Who should use this?

Hermes agent operators running multi-session workflows who want durable context without cross-contamination. If you switch between projects or topics and are tired of re-explaining basics, this adds the missing persistence layer. Not for teams needing production-tested infrastructure or comprehensive documentation.

Verdict

scope-recall addresses a real pain point with a thoughtfully layered design. The SQLite-plus-vector approach is architecturally sound and the scope model is intuitive. However, with a 0.9% credibility score and only 10 stars, this is early-stage software lacking real-world validation. Install it for local experimentation or personal agents, but hold off on production deployments until field testing accumulates. The fundamentals are there; the track record is not.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.