This plugin integrates AI Large Language Model (LLM) capabilities directly into Mattermost workspaces, allowing users to run local or cloud-based models. It targets Mattermost users and administrators, offering benefits like automated summarization, action item extraction, and semantic search, while maintaining data control and flexible deployment options.

How It Works

The plugin functions as a Mattermost extension, providing a bridge to various LLM providers, including local options (Ollama, vLLM) and cloud services (OpenAI, Anthropic, Azure), or any OpenAI-compatible API. It enables features such as thread and channel summarization, action item identification, meeting transcription, and semantic search, the latter requiring a PostgreSQL database with the pgvector extension. This flexible approach allows organizations to choose their preferred LLM infrastructure and maintain data privacy.

Quick Start & Requirements

• Installation: Download the latest release, upload, and enable via the Mattermost System Console.
• Prerequisites: Mattermost Server v11.9.0+, PostgreSQL with pgvector extension (for semantic search), network access to LLM provider.
• Docs: Comprehensive guides available in the /docs directory.
• Experimental Build: Latest master build available at https://github.com/mattermost/mattermost-plugin-agents/releases/tag/latest-master.

Highlighted Details

• Supports multiple AI assistants with specialized personalities.
• Enables key features: Thread & Channel Summarization, Action Item Extraction, Meeting Transcription, Semantic Search, Smart Reactions, Direct Conversations.
• Offers flexible LLM support: local, cloud, and OpenAI-compatible APIs.
• Includes an Integration Bridge Client (Go library) for extending AI capabilities to other Mattermost plugins.
• Features benchmark tests for streaming performance and multi-provider evaluation support.

Maintenance & Community

No specific details on contributors, sponsorships, or dedicated community channels (e.g., Discord/Slack) are provided in the README.

Licensing & Compatibility

The repository is licensed under Apache-2, with the server/enterprise directory subject to the Mattermost Source Available License. Users should review the latter for specific commercial use or linking restrictions.

Limitations & Caveats

Semantic search functionality is dependent on a PostgreSQL database equipped with the pgvector extension. The "latest-master" build is explicitly marked as experimental.