garyqlin

garyqlin / gbase

Public

GBase — Recursive Self-Improvement Agent Framework. Memory, evolution, quality gates, identity system, and 40+ auto-registered tools.

agent-framework ai-agents cognitive-architecture ebbinghaus llm
21
0
85% credibility
Found May 28, 2026 at 21 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

GBase is an AI agent framework that transforms a basic AI assistant into a persistent, self-improving companion. It gives the assistant long-term memory so it remembers past conversations, an experience engine that learns from successes and failures, and the ability to analyze and improve its own behavior. The framework supports multiple personalities, automatic code snapshots for safety, and can run as a single assistant or a coordinated team of specialized agents.

How It Works

1
🤖 You discover GBase

You hear about an AI assistant that actually remembers things between conversations and can learn from its mistakes.

2
⚙️ You set up your assistant

You install the framework and connect it to an AI service of your choice, giving your assistant its own identity.

3
🧠 Your assistant starts learning

As you chat, your assistant automatically stores important facts, lessons, and patterns it discovers in its long-term memory.

4
🔄 Your assistant improves itself

When you ask, your assistant analyzes its own past work, spots patterns in its mistakes, and rewrites its own approach for next time.

5
You can work alone or with a team
🧑‍💻
Solo mode

Your single assistant handles everything with its full memory and skills

🏗️
Team mode

Multiple specialized assistants (like a coder, a designer, a reviewer) work together on complex projects

6
🛡️ Everything stays safe

Before making any changes, your assistant automatically saves a snapshot so you can always undo if something goes wrong.

✨ Your AI assistant becomes truly capable

You have an AI partner that remembers what worked, learns from what didn't, and gets better at helping you over time.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 21 to 21 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 gbase?

GBase is a Python framework for building AI agents that don't just execute tasks -- they remember, learn, and evolve. Unlike simple chat wrappers, it gives agents long-term memory with active recall, a multi-agent identity system, and built-in quality gates where one agent builds and another audits. The framework runs in CLI mode for interactive sessions or as an HTTP server with a REST API, and supports spawning specialized sub-agents (called "arms") that collaborate on complex workflows. It works with any OpenAI-compatible API, so you're not locked into a single provider.

Why is it gaining traction?

The hook is recursive self-improvement: agents can analyze their own outputs, detect failure patterns, and trigger an evolution cycle that rewrites themselves for the next round. Quality gates enforce multi-agent review pipelines -- code doesn't ship until it survives cross-examination. The @tool decorator auto-registers any Python function as an LLM-accessible tool, which is a surprisingly ergonomic pattern. There's also a DAG-based orchestration engine for deterministic multi-step workflows, and a lifeline system that auto-snapshots git state before every code modification.

Who should use this?

Backend developers building multi-agent pipelines who need persistent memory and self-improvement. Teams tired of agents that forget everything after each session. Projects requiring audit trails where code changes are reviewed before shipping. Anyone running Python-based agent workflows that need structure beyond "call LLM, get response."

Verdict

At 21 stars with alpha status, gbase is early-stage and the 0.85% credibility score reflects that -- there's limited community validation, no visible test suite, and the documentation leans toward ambition over polish. The feature set is genuinely interesting for a framework this young, but you'd be adopting it before it's proven. Worth watching, but probably not ready for production workloads unless you're comfortable with the bleeding edge.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.