v2.1.0

What's Changed

• chore(deps): Update dependencies by @reneleonhardt in #42
• chore(deps): bump docker/metadata-action from 2a4836ac76fe8f5d0ee3a0d89aa12a80cc552ad3 to 902fa8ec7d6ecbf8d84d538b9b233a880e428804 by @dependabot in #43
• docs: update README.md by @eltociear in #48
• chore(deps): bump github.com/sashabaranov/go-openai from 1.19.4 to 1.38.1 by @dependabot in #47
• chore(ui): Switch to text-based Bun lock file to allow diffing by @richiejp in #50
• fix: correct image name, switch to sd-1.5-ggml as default by @mudler in #51
• chore(deps-dev): bump globals from 15.15.0 to 16.0.0 in /webui/react-ui by @dependabot in #45
• fix(githubreader): do not use pointers by @mudler in #53
• fix(github*): pass by correctly owner and repository by @mudler in #54
• feat(ui): Structured observability/status view by @richiejp in #40
• feat(browseragent): add browser agent runner action by @mudler in #55
• chore(deps): bump github.com/sashabaranov/go-openai from 1.38.1 to 1.38.2 by @dependabot in #63
• chore(deps): bump github.com/metoro-io/mcp-golang from 0.9.0 to 0.11.0 by @dependabot in #64
• chore(deps-dev): bump react-router-dom from 7.5.0 to 7.5.1 in /webui/react-ui by @dependabot in #65
• chore(deps-dev): bump eslint-plugin-react-hooks from 5.2.0 to 6.0.0 in /webui/react-ui by @dependabot in #67
• feat: add history metadata of agent browser by @mudler in #71
• chore: default to gemma-3-12b-it-qat by @mudler in #60
• chore(deps-dev): bump vite from 6.3.1 to 6.3.2 in /webui/react-ui by @dependabot in #69
• feat: allow the agent to perform things concurrently by @mudler in #74
• chore: better defaults for parallel jobs by @mudler in #76
• fix: Handle state on agent restart and update observables by @richiejp in #75
• feat: local MCP server support by @mudler in #61
• Update README.md by @richiejp in #80
• chore(deps): bump actions/checkout from 2 to 4 by @dependabot in #44
• feat(browseragent): post screenshot on slack by @mudler in #81
• fix(discord): automatically add 'Bot' prefix to token if missing by @mudler in #83
• fix(docker): Add mcpbox server to extended compose files by @richiejp in #84
• chore(mcpbox): use dind by @mudler in #85
• chore(mcpbox): use ubuntu:24.04 as base by @mudler in #86
• feat(ui): Action playground config and parameter forms by @richiejp in #16
• feat(ui,observability): Add summary details for observables by @richiejp in #78
• chore(defaults): Enable reasoning by default by @mudler in #89
• feat: add deep research action by @mudler in #91
• feat: specify timeout by @mudler in #97
• chore: Update all deps by @richiejp in #95
• feat: add option to strip thinking from output by @mudler in #100
• fix(core): Add recursive loop detection and move loop detection by @richiejp in #101
• fix: cleanup responses also when not picking any tool by @mudler in #102
• feat(connectors): add support for Matrix by @mudler in #82
• feat(github): add option to create PR from forks by @mudler in #105
• chore(deps): bump ubuntu from 22.04 to 24.04 by @dependabot in #107
• feat(github): add issue editor by @mudler in #106
• chore(deps): bump github.com/sashabaranov/go-openai from 1.38.2 to 1.38.3 by @dependabot in #93
• fix(actions): make sure to initialize a config by @mudler in #109
• feat(github): add action to list and search files in a repository by @mudler in #110
• fix(reply): force replying without using tools by @mudler in #131
• feat(filters): Add configurable filters for incoming jobs by @richiejp in #99
• chore(deps): bump github.com/go-telegram/bot from 1.14.2 to 1.15.0 by @dependabot in #132
• chore(deps): bump github.com/sashabaranov/go-openai from 1.39.0 to 1.39.1 by @dependabot in #134
• chore(deps-dev): bump @types/react-dom from 19.1.2 to 19.1.3 in /webui/react-ui by @dependabot in #135
• feat(telegram): show thought process by @mudler in #140
• feat(telegram): handle correctly generated multimedia and links by @mudler in #141
• fix(telegram): upload of images by @mudler in #143
• fix(playground): convert fieldType number to string for json unmarshalling by @martianmr in #142
• feat(call_agents): allow to specify whitelist and blacklist agents by @mudler in #144
• feat(evaluation): add deep evaluation mechanism by @mudler in #145
• fix(telegram): split long messages by @mudler in #146
• feat(telegram): add action to send telegram message by @mudler in #147
• feat(agent): shared state, allow to track conversations globally by @mudler in #148
• fix(matrix): Stop Sync Go routine and correct logs by @richiejp in #154
• chore(deps-dev): bump @types/react-dom from 19.1.3 to 19.1.4 in /webui/react-ui by @dependabot in #153
• chore(deps-dev): bump vite from 6.3.3 to 6.3.5 in /webui/react-ui by @dependabot in #136
• chore(deps-dev): bump react-router-dom from 7.5.3 to 7.6.0 in /webui/react-ui by @dependabot in #152
• fix(core): Add prompt generated from KB to conv by @richiejp in #156
• chore(deps-dev): bump eslint from 9.25.1 to 9.26.0 in /webui/react-ui by @dependabot in #137
• feat: email connector by @AKSizov in #157
• fix(matrix),feat(core): Add KB observability and matrix improvements by @richiejp in #155
• fix: errored ssh output by @AKSizov in #158
• chore(docker-compose): update LocalAI images by @mudler in #160
• feat(sshbox): add sshbox to run commands by @mudler in #161
• chore(deps): bump github.com/metoro-io/mcp-golang from 0.11.0 to 0.12.0 by @dependabot in #151
• chore(deps): bump github.com/valyala/fasthttp from 1.61.0 to 1.62.0 by @dependabot in #150
• chore(deps-dev): bump @eslint/js from 9.25.1 to 9.26.0 in /webui/react-ui by @dependabot in #138
• ci: enable multi-arch Docker builds for amd64 and arm64 in image workflow by @TheGoddessInari in #164
• chore(deps-dev): bump @eslint/js from 9.26.0 to 9.27.0 in /webui/react-ui by @dependabot in #171
• chore(deps-dev): bump @types/react-dom from 19.1.4 to 19.1.5 in /webui/react-ui by @dependabot in #167
• chore(deps-dev): bump @types/react from 19.1.2 to 19.1.4 in /webui/react-ui by @dependabot in #169
• chore(deps-dev): bump eslint from 9.26.0 to 9.27.0 in /webui/react-ui by @dependabot in #165
• feat(reminders): add reminder system to perform long-term goals in the background by @mudler in #176
• feat(telegram): Add support for groups by @mudler in #183
• chore(deps): bump github.com/gofiber/fiber/v2 from 2.52.6 to 2.52.8 by @dependabot in #190
• chore(deps): bump github.com/metoro-io/mcp-golang from 0.12.0 to 0.13.0 by @dependabot in #189
• chore(deps-dev): bump react-router-dom from 7.6.0 to 7.6.1 in /webui/react-ui by @dependabot in #186
• chore(deps-dev): bump @vitejs/plugin-react from 4.4.1 to 4.5.0 in /webui/react-ui by @dependabot in https://github.com/mudler/LocalAG...

v2.0.0

🚀 LocalAGI v2.0.0: The AI Agent Orchestration Platform Reborn! 🚀

I am so absolutely thrilled to announce the arrival of LocalAGI v2.0.0! 🎉 This isn't just an update; it's a complete rebirth of our vision for a powerful, local-first AI agent framework.

From Experiment to Platform

LocalAGI started as a pioneering Python experiment back when "AI Agents" were just emerging concepts (see the original announcement!). It showcased the potential of local, coordinated AI task execution.

Remember the Original? (Python Demos)

Searching the internet (interactive mode):

search.mp4

Planning a road trip (batch mode):

planner.mp4

Building on that foundation and inspired by the community's enthusiasm, we've rebuilt LocalAGI from the ground up in Go!

What is LocalAGI v2?

LocalAGI v2 is a modern, robust AI Agent Orchestration Platform designed for the next generation of AI workflows. It allows you to easily configure, deploy, manage, and monitor intelligent AI agents and teams that collaborate to achieve complex goals – all while running locally and integrating seamlessly with LocalAI.

✨ Key Features in v2.0.0 ✨

• 🚀 Built with Go: Rewritten for performance, concurrency, and maintainability.
• 🎯 OpenAI Responses API Compatible: Acts as a drop-in replacement, allowing seamless integration with LocalAI or even OpenAI's API, enhanced with powerful agentic capabilities.
• 🎨 Intuitive No-Code WebUI: Configure agents, build teams, define goals, manage integrations, and monitor progress through a user-friendly interface. Huge thanks to @richiejp and the community for their massive contributions here!
• 🛡️ Privacy-First Design: Your data, models, and agent interactions stay entirely on your hardware.
• 🔌 Instant Integrations: Built-in support for interacting with external services like Slack, Telegram, Discord, GitHub Issues, IRC, and more.
• ⚡ Extensible & Multimodal: Supports multiple LLMs (via LocalAI), vision models, and custom tool/function definitions.
• 🧠 Works with LocalRecall: Integrates with our dedicated memory component, LocalRecall, for persistent agent memory.

Check Out the WebUI!

Managing sophisticated agent workflows has never been easier:

The Local AI Stack

LocalAGI is a core component of our open-source, privacy-first AI ecosystem:

Component	Description
LocalAI	The free, Open Source OpenAI alternative for local inferencing (LLMs, Vision, Audio, etc.).
LocalAGI (You are here!)	The agent orchestration platform for managing complex AI workflows.
LocalRecall	A RESTful API providing persistent memory and knowledge base capabilities for agents.

Getting Started

Ready to unleash the power of local AI agents?

1. Ensure you have LocalAI running. LocalAGI relies on it for model inferencing.
2. Check out the Installation Guide in our README.md for instructions using Docker, Go, or other methods.
3. Explore the WebUI and start configuring your first agent!

We're constantly improving documentation, so please check back often or ask questions in the Discussions section!

Join the Community! ❤️

LocalAGI is built by the community, for the community. It's not backed by VCs, just passionate open-source developers donating their time.

• Star the Repo! ⭐ It helps visibility immensely!
• Report Bugs & Request Features: Use the GitHub Issues tab.
• Contribute Code: Check out CONTRIBUTING.md.
• Spread the Word: Tell your friends and colleagues about LocalAGI!

A massive thank you to everyone who contributed code, ideas, bug reports, and encouragement to make v2.0.0 possible!

We can't wait to see the incredible things you build with LocalAGI!

– The LocalAI/LocalAGI Team

What's Changed

• Pin openai API to pre v1.0 by @richiejp in #8
• Update README.md by @scenaristeur in #6
• Add langchain-community requirement to make it build again by @greygoo in #11
• LocalAGI port in Go by @mudler in #14
• Update README.md by @richiejp in #23
• fix(pick_action): improve action pickup by using only the assistant thought process by @mudler in #25
• feat(github): add actions to comment and read PRs by @mudler in #26
• Add Github reviewer and improve reasoning by @mudler in #27
• chore: cleanup, identify goal from conversation when evaluting achievement by @mudler in #29
• fix(ui): Don't try to pass unserializable Go objects to status UI by @richiejp in #28
• fix(ui): Proxy avatars endpoint in dev mode by @richiejp in #32
• fix(test): Encourage LLM to plan multiple searches by @richiejp in #36
• chore: return more results to the LLM of the action that was done by @mudler in #38
• feat(github): add action to open up a PR and get all repository content by @mudler in #39

New Contributors

• @richiejp made their first contribution in #8
• @scenaristeur made their first contribution in #6
• @greygoo made their first contribution in #11
• @mudler made their first contribution in #14

Full Changelog: https://github.com/mudler/LocalAGI/commits/v2.0.0