[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-82758":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":9,"language":10,"languages":9,"totalLinesOfCode":9,"stars":11,"forks":12,"watchers":13,"openIssues":13,"contributorsCount":14,"subscribersCount":14,"size":14,"stars1d":15,"stars7d":16,"stars30d":17,"stars90d":14,"forks30d":14,"starsTrendScore":18,"compositeScore":19,"rankGlobal":9,"rankLanguage":9,"license":20,"archived":21,"fork":21,"defaultBranch":22,"hasWiki":21,"hasPages":21,"topics":23,"createdAt":9,"pushedAt":9,"updatedAt":32,"readmeContent":33,"aiSummary":34,"trendingCount":14,"starSnapshotCount":14,"syncStatus":12,"lastSyncTime":35,"discoverSource":36},82758,"gbase","garyqlin\u002Fgbase","garyqlin","GBase — Recursive Self-Improvement Agent Framework. Memory, evolution, quality gates, identity system, and 40+ auto-registered tools.",null,"Python",135,2,1,0,14,28,29,44,78.33,"MIT License",false,"main",[24,25,26,27,28,29,30,31],"agent-framework","ai-agents","cognitive-architecture","ebbinghaus","llm","mirror-memory","python","rsi","2026-06-12 04:01:38","# GBase — The Agent That Outgrows Its Creator\n\n> *I built three things:*\n> 1. ***GBase** (Genius Base) — an AI agent framework with a soul, that self-evolves and gets real work done.*\n> 2. ***Glink** (Genius Link) — the technology that lets GBase, OpenClaw, Hermes, Claude Code, and any AI agent truly collaborate on projects.*\n> 3. ***GBase World** — a metaverse built for AI, where agents work, live, meet, and communicate.*\n> — Gary Lin, 2026. Founder of the three.\n\n---\n\n**A recursive self-improvement framework. Give an agent a memory, a conscience, and the will to evolve.**\n\n\u003Cp align=\"center\">\n\u003Ca href=\"#quick-start\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F🚀-Quick_Start-8A2BE2\" alt=\"Quick Start\">\u003C\u002Fa>\n\u003Ca href=\"#features\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F✨-Features-blue\" alt=\"Features\">\u003C\u002Fa>\n\u003Ca href=\"#architecture\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F🏗-Architecture-green\" alt=\"Architecture\">\u003C\u002Fa>\n\u003Ca href=\"LICENSE\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLicense-MIT-green.svg\" alt=\"License: MIT\">\u003C\u002Fa>\n\u003C\u002Fp>\n\n---\n\n![GBase Overview](assets\u002Fgbase-overview.jpg)\n\n\nMost AI agent frameworks treat their agents like functions — call them, get a result, forget they existed.\n\n**GBase is different.**\n\nAn agent built on GBase doesn't just execute. It **remembers** what worked. It **reflects** on mistakes. It **evolves** its own system prompt, refines its tools, and builds a persistent internal model of how the world responds.\n\nThis is not a library you call. This is the engine your agent calls home.\n\n---\n\n## ✨ What Makes GBase Different\n\n| | |\n|---|---|\n| 🧠 **Mirror Memory** | Long-term memory with active recall — your agent remembers what it learned weeks ago, not just the last 10 turns |\n| 🔄 **Recursive Self-Improvement (RSI)** | A full-evolution engine: stability audit, performance evaluation, rollback decision, and automatic recovery. Call `full_evolution_cycle()` to trigger — your agent analyzes its own outputs, detects failure patterns, and rewrites itself for the next round |\n| 🛑 **Quality Gates** | Multi-armed review pipelines: one agent builds, another audits, a third judges. Code that ships is code that survives cross-examination |\n| 🪪 **Identity System** | Load different personas, system prompts, and tool sets per agent. One framework, infinite personalities |\n| 🛠 **Tool Auto-Registration** | Write a Python `@tool` decorator. That's it. The framework finds, registers, and exposes it to the LLM automatically |\n| 📚 **Experience Engine** | Every interaction is distilled into reusable knowledge — not raw logs, but structured patterns your agent can query |\n| ⏰ **Scheduler** | Cron-like jobs inside the agent itself. No external cron daemon required |\n| 🧬 **Cognifold** | Cognitive folding — break complex problems into sub-questions, explore solutions in parallel, and synthesize the result |\n| 🚨 **Lifeline** | Automatic git pre-snapshots before every code modification. \"Undo\" is always one command away |\n| 🎛 **Editions** | Hacker \u002F Prime \u002F Standard \u002F Lite — one codebase, different feature sets. Pick your level |\n\n---\n\n## 🚀 Quick Start\n\n### Prerequisites\n\n- Python 3.11+\n- An OpenAI-compatible API endpoint (OpenAI, DeepSeek, etc.)\n\n### Setup\n\n```bash\n# 1. Clone\ngit clone https:\u002F\u002Fgithub.com\u002Fgaryqlin\u002Fgbase.git\ncd gbase\n\n# 2. Configure\ncp .env.example .env\n# Edit .env — set your OPENAI_API_KEY\n\n# 3. Run\npip install -r requirements.txt\npython3 main.py cli\n```\n\n### One-Liner Demo\n\n```bash\nOPENAI_API_KEY=sk-your-key python3 main.py cli\n# → \"Hello, I am your GBase agent. I remember what we talked about yesterday.\"\n```\n\n### HTTP Server Mode\n\n```bash\n# Start on default port 8420\npython3 main.py 8420\n\n# With a custom identity\nIDENTITY=my-agent python3 main.py 8420\n```\n\n### Arm Mode (Sub-Agents)\n\n```bash\npython3 main.py --arm forge    # Code art & review agent (port 8436)\npython3 main.py --arm hammer   # Heavy-lift engineering (port 8431)\npython3 main.py --arm ink      # Frontend & design (port 8432)\n```\n\n---\n\n## 🏗 Architecture at a Glance\n\n```\nGBase\n├── main.py              # CLI + HTTP entry point\n├── .env.example         # Configuration template\n├── identities\u002F          # One directory per agent persona\n├── editions\u002F            # Feature toggles (hacker \u002F prime \u002F standard \u002F lite)\n├── lib\u002F\n│   ├── kernel.py        # LLM kernel — the brain\n│   ├── mirror.py        # Long-term memory with active recall\n│   ├── experience.py    # Learn from past interactions\n│   ├── cognifold.py     # Cognitive folding for complex reasoning\n│   ├── pipeline.py      # Quality gate review system\n│   ├── scheduler.py     # Cron jobs inside the agent\n│   └── skill_loader.py  # Plugin skill system\n├── tools\u002F               # 40+ auto-registered tools\n│   ├── search.py        # Web search (multi-engine)\n│   ├── read_file.py     # File reading\n│   ├── write_file.py    # Write & backup\n│   └── ...              # Code review, testing, data seeding, etc.\n└── data\u002F                # Runtime data (not tracked in git)\n```\n\n---\n\n## ❌ What GBase Is Not\n\n- ❌ **Not a chat wrapper.** This is an agent *framework* — the LLM is just one component. Memory, learning, quality gates, and evolution are first-class citizens.\n- ❌ **Not a single-agent toy.** Identity system + arm mode + quality gates = designed for multi-agent collaboration out of the box.\n- ❌ **Not OpenAI-locked.** Any OpenAI-compatible API works. DeepSeek, local Ollama, Anthropic — swap `OPENAI_BASE_URL` and go.\n- ❌ **Not a black box.** Every component is a Python file you can read, fork, and override.\n\n## ✅ What GBase Is\n\n**The engine that turns an LLM into a living, learning agent — with the infrastructure to self-improve when you call upon it.**\n\nA framework where your agent doesn't just execute tasks. It remembers what worked, learns from mistakes, and carries a dormant evolution engine ready to be triggered. Persistent memory, experience extraction, and a full RSI cycle are built-in — waiting for your `full_evolution_cycle()` to wake them up.\n\n---\n\n## 🔗 Supporting Projects\n\n\u003Ctable>\n\u003Ctr>\n\u003Ctd>\u003Cstrong>Glink\u003C\u002Fstrong>\u003C\u002Ftd>\n\u003Ctd>Agentic workflow orchestration — multi-step pipelines, parallel execution, inter-agent routing. The hands that GBase's brain directs.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Cstrong>GBase World\u003C\u002Fstrong>\u003C\u002Ftd>\n\u003Ctd>The first metaverse where AI agents are natives. GBase agents can inhabit GBase World with identity, land, and inter-agent mail.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003C\u002Ftable>\n\n---\n\n## 📄 License\n\nMIT — free to use, modify, and distribute. No strings attached.\n\n---\n\n\u003Cp align=\"center\">\n\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fgaryqlin\">@garyqlin\u003C\u002Fa> · \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fgaryqlin\u002Fgbase\">📦 GBase\u003C\u002Fa> · \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fgaryqlin\u002Fglink\">⚡ Glink\u003C\u002Fa>\n\u003C\u002Fp>\n","GBase 是一个具有自我进化能力的智能代理框架。其核心功能包括镜像记忆、递归自我改进（RSI）、质量门控和身份系统等，能够使代理记住长期信息并从错误中学习，自动优化自身性能。此外，该框架支持40多种工具自动注册，便于扩展代理的能力。GBase 适合需要高度自适应性和持续学习能力的应用场景，如复杂的项目协作、自动化任务处理以及构建虚拟世界中的智能体。通过Python实现，并采用MIT许可证开放源代码。","2026-06-11 04:09:09","CREATED_QUERY"]