[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-1972":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":10,"totalLinesOfCode":10,"stars":12,"forks":13,"watchers":14,"openIssues":15,"contributorsCount":16,"subscribersCount":16,"size":16,"stars1d":14,"stars7d":17,"stars30d":18,"stars90d":16,"forks30d":16,"starsTrendScore":19,"compositeScore":20,"rankGlobal":10,"rankLanguage":10,"license":21,"archived":22,"fork":22,"defaultBranch":23,"hasWiki":22,"hasPages":22,"topics":24,"createdAt":10,"pushedAt":10,"updatedAt":29,"readmeContent":30,"aiSummary":31,"trendingCount":16,"starSnapshotCount":16,"syncStatus":32,"lastSyncTime":33,"discoverSource":34},1972,"reflexio","ReflexioAI\u002Freflexio","ReflexioAI","Make your agents improve themselves. Reflexio is an AI agent self-improvement harness that enables your AI agents to continuously learn from real user interactions. ","https:\u002F\u002Freflexio.ai\u002F",null,"Python",293,39,15,1,0,29,152,45,4.81,"Apache License 2.0",false,"main",[25,26,27,28],"self-improvement","self-improving-agent","self-improving-ai","self-learning","2026-06-12 02:00:35","\u003Cp align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Freflexio-ai\u002Freflexio\">\n    \u003Cimg src=\"docs\u002Fimages\u002Fbanner.png\" width=\"800px\" alt=\"Reflexio - Make Your Agents Improve Themselves\">\n  \u003C\u002Fa>\n\u003C\u002Fp>\n\u003Cdiv align=\"center\">\n\n[![Python >= 3.12](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fpython-%3E%3D3.12-blue)](https:\u002F\u002Fwww.python.org\u002F)\n[![License: Apache 2.0](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Flicense-Apache%202.0-green)](LICENSE)\n[![PyPI](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fv\u002Freflexio-client)](https:\u002F\u002Fpypi.org\u002Fproject\u002Freflexio-client\u002F)\n[![Downloads](https:\u002F\u002Fstatic.pepy.tech\u002Fbadge\u002Freflexio-ai\u002Fmonth)](https:\u002F\u002Fpepy.tech\u002Fproject\u002Freflexio-ai)\n[![Search p50 57ms](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fsearch-57ms%20p50-brightgreen)](reflexio\u002Fbenchmarks\u002Fretrieval_latency\u002Fresults\u002Freport.md)\n[![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FReflexioAI\u002Freflexio)](https:\u002F\u002Fgithub.com\u002FReflexioAI\u002Freflexio\u002Fstargazers)\n\n[Quick Start](#quick-start) · [Features](#features) · [Integrations](#integrations) · [SDK](#sdk-usage) · [CLI](reflexio\u002Fcli\u002FREADME.md) · [Architecture](#architecture) · [Docs](https:\u002F\u002Fwww.reflexio.ai\u002Fdocs) · [Contributing](#contributing)\n\n\u003C\u002Fdiv>\n\n---\n\n\u003Cp align=\"center\">\n  \u003Cb>81% fewer planning steps\u003C\u002Fb>  ·  \u003Cb>72% less tokens\u003C\u002Fb>  ·  on real GDPVal knowledge-work tasks, \u003Cbr\u002F>\n  \u003Ci>on top of\u003C\u002Fi> what a SOTA self-improving Hermes agent already learns on its own.\u003Cbr\u002F>\n  \u003Ca href=\"benchmark\u002Fgdpval\u002FRESULTS.md\">\u003Cb>See the benchmark →\u003C\u002Fb>\u003C\u002Fa>\n\u003C\u002Fp>\n\n---\n\n### Migration from claude_code integration (removed in this release)\n\nThe `reflexio setup claude-code` command and its hook files have been removed.\nThe replacement is **[claude-smart](https:\u002F\u002Fgithub.com\u002FReflexioAI\u002Fclaude-smart)**,\na standalone Claude Code plugin distributed via npm.\n\n*This migration only removes the **hook\u002Fplugin installation** path. The local\n`claude-code` LLM provider routing (used to call Anthropic via the Claude Code\nCLI binary as a model backend) remains available — only remove obsolete hook\nentries, not your provider configuration.*\n\n**If you had the old integration installed**, your `.claude\u002Fsettings.json` (per-project)\nor `~\u002F.claude\u002Fsettings.json` (global) likely has hook entries referencing files that no longer exist.\nOpen the file and remove any `hooks` entries that reference paths under `reflexio\u002Fintegrations\u002Fclaude_code\u002F`\nor `integrations\u002Fclaude_code\u002F`. Then run `npx claude-smart install` (or use the Claude Code plugin marketplace)\nfor the modern equivalent.\n\n---\n\n## What is Reflexio?\nReflexio is an **AI agent self-improvement harness** that enables your AI agents to continuously learn from real user interactions. It turns user corrections into persisted behavioral improvements for agents and capturing successful execution paths for reuse.  \n\nWhat one user teaches, every user benefits from.  \n\nAs your agent is used more, it becomes smarter, faster, and more effective at solving domain-specific tasks.\nThe moat for AI agents is what your agent learns from every interaction it handles.  \n\nOur vision is that AI systems should get better with every interaction.\n\n> **Benchmarked on GDPVal**: on 4 of 5 real knowledge-work tasks from OpenAI's public GDPVal benchmark, Reflexio cuts a **median −81% planning steps and −72% tokens** on a Hermes agent running `minimax\u002FMiniMax-M2.7` — measured against a *warm baseline*: the same agent re-running the task after it has already learned from itself. In other words, Reflexio's savings come **on top of** what a SOTA self-improving agent has learnt on its own. See the full writeup → [benchmark\u002Fgdpval\u002FRESULTS.md](benchmark\u002Fgdpval\u002FRESULTS.md).\n\n```mermaid\nflowchart LR\n    A[AI Agent] -->|conversations| B[Reflexio]\n    G[Human Expert] -->|ideal responses| B\n    B --> C[User Profiles]\n    B --> D[Playbook Extraction]\n    D --> E[Playbook Aggregation]\n    B --> F[Success Evaluation]\n```\n\nPublish conversations from your agent, and Reflexio closes the self-improvement loop:\n\n- **Never Repeat the Same Mistake**: Transforms user corrections and interaction signals into improved decision-making processes — so agents adapt their behavior and avoid repeating the same mistakes.\n- **Lock In What Works**: Persists successful strategies and workflows so your agent reuses proven paths instead of starting from scratch.\n- **Transfer Learning Across Users**: What one user teaches, every user benefits from — corrections and successful strategies from one individual propagate to improve the agent for everyone, no retraining required.\n- **Learn from Human Experts**: Publish expert-provided ideal responses alongside agent responses — Reflexio automatically extracts actionable playbooks from the differences.\n\n> **For developers**: See [developer.md](developer.md) for project structure, environment setup, testing, and coding guidelines.\n\n## Table of Contents\n\n- [Demo](#demo)\n- [Quick Start](#quick-start)\n- [Features](#features)\n- [Integrations](#integrations)\n- [SDK Usage](#sdk-usage)\n- [Architecture](#architecture)\n- [Documentation](#documentation)\n- [Contributing](#contributing)\n- [Star History](#star-history)\n- [License](#license)\n\n## Demo\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"docs\u002Fimages\u002Freflexio_example.gif\" width=\"800px\" alt=\"Reflexio example experience\">\n\u003C\u002Fp>\n\n## Quick Start\n\n### Prerequisites\n\n| Tool | Description |\n| --- | --- |\n| [uv](https:\u002F\u002Fdocs.astral.sh\u002Fuv\u002Fgetting-started\u002Finstallation\u002F) | Python package manager |\n| [Node.js](https:\u002F\u002Fnodejs.org\u002F) >= 18 | Frontend runtime |\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"docs\u002Fimages\u002Fdemo.gif\" width=\"800px\" alt=\"Reflexio Demo\">\n\u003C\u002Fp>\n\n### Setup\n\n**Option A — Install from PyPI** (fastest, for users):\n\n```shell\npip install reflexio-ai\n\n# start\u002Fstop services. data saved under ~\u002F.reflexio\nreflexio services start           # API (8081), Docs (8082), SQLite storage\nreflexio services stop            # Stop all services\n```\n\n**Option B — Clone from source** (for contributors):\n\n```shell\n# clone the repo\ngit clone https:\u002F\u002Fgithub.com\u002FReflexioAI\u002Freflexio.git\ncd reflexio\n\n# configure: copy env template, then set at least one LLM API key (OpenAI, Anthropic, etc.)\ncp .env.example .env\n\n# install dependencies\nuv sync                                    # Python (includes workspace packages)\nnpm --prefix docs install                  # API docs\n\n# start\u002Fstop services. data saved under ~\u002F.reflexio\nuv run reflexio services start             # API (8081), Docs (8082), SQLite storage\nuv run reflexio services stop              # Stop all services\n```\n\n> Alternative: `python -m reflexio.cli services start` or `.\u002Frun_services.sh`\n\nOnce running, open **[http:\u002F\u002Flocalhost:8082](http:\u002F\u002Flocalhost:8082)** to interactively browse and try out the API.\n\u003Cp align=\"center\">\n  \u003Cimg src=\"docs\u002Fimages\u002Fdoc_website.png\" width=\"800px\" alt=\"Reflexio Doc Website\">\n\u003C\u002Fp>\n\n### Try it in 30 seconds (CLI)\n\nReflexio ships a first-class CLI — the fastest way to see the loop end-to-end with no code. Publish a real multi-turn conversation where the user **corrects** the agent (that's the signal Reflexio learns from), then search for what was extracted:\n\n```shell\nuv run reflexio publish --user-id alice --wait --data '{\n  \"interactions\": [\n    {\"role\": \"user\",      \"content\": \"Deploy the new service.\"},\n    {\"role\": \"assistant\", \"content\": \"Starting deployment to us-east-1...\"},\n    {\"role\": \"user\",      \"content\": \"Wait — we never deploy production to us-east-1. Always use us-west-2.\"},\n    {\"role\": \"assistant\", \"content\": \"Understood. Switching to us-west-2.\"}\n  ]\n}'\n\n# Search the extracted profiles and playbooks\nuv run reflexio search \"deployment region\"\n```\n\nOne conversation, two artifacts: a user profile (`production region is us-west-2`) and an agent playbook (`confirm region before deploying`). See the [CLI reference](reflexio\u002Fcli\u002FREADME.md) for all input modes (inline JSON, `--file`, `--stdin`) and the full command list.\n\n### Integrate with the Python SDK\n\n```python\nimport reflexio\n\nclient = reflexio.ReflexioClient(\n    url_endpoint=\"http:\u002F\u002Flocalhost:8081\u002F\"\n)\n\n# Publish a multi-turn conversation where the user corrects the agent —\n# Reflexio extracts a profile (\"prod region = us-west-2\") and a playbook\n# (\"confirm region before deploying\").\nclient.publish_interaction(\n    user_id=\"alice\",\n    interactions=[\n        {\"role\": \"user\",      \"content\": \"Deploy the new service.\"},\n        {\"role\": \"assistant\", \"content\": \"Starting deployment to us-east-1...\"},\n        {\"role\": \"user\",      \"content\": \"Wait — we never deploy production to us-east-1. Always use us-west-2.\"},\n        {\"role\": \"assistant\", \"content\": \"Understood. Switching to us-west-2.\"},\n    ],\n)\n```\n\nReflexio will automatically generate profiles and extract playbooks in the background.\n\n## Features\n\n### Profile Generation\n\n- Extracts behavioral profiles from conversations using configurable extractors\n- Supports versioning (current → pending → archived) with upgrade\u002Fdowngrade workflows\n- Multiple extractors run in parallel with independent windows and strides\n\n[Read more about user profiles →](https:\u002F\u002Fwww.reflexio.ai\u002Fdocs\u002Fconcepts\u002Fuser-profiles)\n\n### Playbook Extraction & Aggregation\n\n- Extracts playbooks from user behavior patterns\n- Clusters similar entries and aggregates with LLM (with change detection to skip unchanged clusters)\n- Approval workflow: review and approve\u002Freject agent playbooks\n\n[Read more about agent playbooks →](https:\u002F\u002Fwww.reflexio.ai\u002Fdocs\u002Fconcepts\u002Fagent-playbook)\n\n### Expert Learning\n\n- Publish human-expert ideal responses alongside agent responses via the `expert_content` field\n- Reflexio automatically compares agent vs. expert responses, focusing on substantive differences (missing info, incorrect approach, reasoning gaps) while ignoring stylistic ones\n- Generates actionable playbooks as trigger\u002Finstruction\u002Fpitfall SOPs that teach the agent what to do differently\n\n[Read more about interactions & expert content →](https:\u002F\u002Fwww.reflexio.ai\u002Fdocs\u002Fconcepts\u002Finteractions#5-expert-content-for-learning-from-experts)\n\n### Agent Success Evaluation\n\n- Session-level evaluation triggered automatically (10 min after last request)\n- Shadow comparison mode: A\u002FB test regular vs shadow agent responses\n- Tool usage analysis for blocking issue detection\n- **Causal measurement of Reflexio's impact** — session-level A\u002FB grouping on the Evaluation page driven by `Request.metadata.reflexio_retrieval_enabled` (with per-turn shadow comparison in development)\n\n[Read more about evaluation →](https:\u002F\u002Fwww.reflexio.ai\u002Fdocs\u002Fexamples\u002Fagent-evaluation)\n\n### Search & Retrieval\n\n- Hybrid search (vector + full-text) across profiles and playbooks\n- LLM-powered query rewriting for improved recall\n- Unified search across all entity types in parallel\n- **Fast at scale**: unified search across ~3,000 indexed rows (profile + user playbook + agent playbook, ~1,000 rows each, queried in parallel) runs at **~57 ms p50 \u002F ~73 ms p95** — measured service-layer with local SQLite on an Apple Silicon MacBook, 30 trials × 20 fixed queries. See the [full benchmark report](reflexio\u002Fbenchmarks\u002Fretrieval_latency\u002Fresults\u002Freport.md) or reproduce with [`reflexio.benchmarks.retrieval_latency`](reflexio\u002Fbenchmarks\u002Fretrieval_latency\u002FREADME.md).\n\n### Multi-Provider LLM Support\n\n- OpenAI, Anthropic, Google Gemini, OpenRouter, Azure, MiniMax, and custom endpoints\n- Powered by LiteLLM — configure your preferred provider via API keys or custom endpoints\n\n## SDK Usage\n\nFor detailed API documentation, see the [full API reference](https:\u002F\u002Fwww.reflexio.ai\u002Fdocs\u002Fapi-reference).\n\nInstall the client:\n\n```shell\npip install reflexio-client\n```\n\n### Basic usage\n\n```python\nimport reflexio\n\nclient = reflexio.ReflexioClient(\n    url_endpoint=\"http:\u002F\u002Flocalhost:8081\u002F\"\n)\n\n# Publish interactions\nclient.publish_interaction(\n    user_id=\"user-123\",\n    interactions=[\n        {\"role\": \"user\",      \"content\": \"...\"},\n        {\"role\": \"assistant\", \"content\": \"...\"},\n    ],\n    agent_version=\"v1\",       # optional: track agent versions\n    session_id=\"session-abc\", # optional: group requests into sessions\n)\n\n# Search profiles\nprofiles = client.search_user_profiles(\n    reflexio.SearchUserProfileRequest(query=\"deployment region preference\")\n)\n\n# Search agent playbooks\nplaybooks = client.get_agent_playbooks(\n    reflexio.GetAgentPlaybooksRequest(agent_version=\"v1\")\n)\n```\n\n### Configuration\n\n```python\n# Update org configuration\nclient.set_config(reflexio.SetConfigRequest(\n    config=reflexio.Config(\n        api_key_config=reflexio.APIKeyConfig(openai=\"sk-...\"),\n        profile_extractor_config=reflexio.ProfileExtractorConfig(...),\n        user_playbook_extractor_config=reflexio.PlaybookConfig(...),\n    )\n))\n```\n\n## Integrations\n\nReflexio integrates with popular AI agent frameworks out of the box:\n\n- **[LangChain](reflexio\u002Fintegrations\u002Flangchain\u002FREADME.md)** -- Drop-in callbacks for LangChain chains and agents.\n- **[OpenClaw](reflexio\u002Fintegrations\u002Fopenclaw\u002FREADME.md)** -- Native integration with the OpenClaw agent framework.\n\n## Architecture\n\n```\nClient (SDK \u002F Web UI)\n  → FastAPI Backend\n    → Reflexio Orchestrator\n      → GenerationService\n        ├─ ProfileGenerationService  → Extractor(s) → Deduplicator → Storage\n        ├─ PlaybookGenerationService → Extractor(s) → Deduplicator → Storage\n        └─ GroupEvaluationScheduler  → Evaluator(s) → Storage (deferred 10 min)\n```\n\nSee [developer.md](developer.md) for project structure, supported LLM providers, and development setup.\n\n## Documentation\n\nFor comprehensive guides, examples, and API reference, visit the **[Reflexio Documentation](https:\u002F\u002Fwww.reflexio.ai\u002Fdocs)**.\n\nFor coding agents adding Reflexio to another agent, see **[Integrating an AI Agent with Reflexio](AI_AGENT_INTEGRATION.md)**.\n\n## Contributing\n\nWe welcome contributions! Please see [developer.md](developer.md) for guidelines.\n\n## Star History\n\n[![Star History Chart](https:\u002F\u002Fapi.star-history.com\u002Fsvg?repos=ReflexioAI\u002Freflexio&type=Date)](https:\u002F\u002Fstar-history.com\u002F#ReflexioAI\u002Freflexio&Date)\n\n## License\n\nThis project is currently licensed under [Apache License 2.0](LICENSE).\n","Reflexio 是一个用于AI代理自我改进的工具，它能够使你的AI代理通过真实的用户交互不断学习。其核心功能包括将用户的修正转化为持久的行为改进，并记录成功的执行路径以供重用。技术上基于Python（版本≥3.12），具有高效的检索延迟性能和较低的规划步骤与令牌消耗。适用于需要持续优化用户体验、提高解决特定领域任务效率的场景，如客户服务自动化、内容生成等。通过不断的学习，AI代理在处理更多交互后变得更加智能、快速且高效。",2,"2026-06-11 02:47:07","CREATED_QUERY"]