DOCSEARCHbyAlgolia
Submit Docs

AnythingLLM Documentation

repository·main·Indexed 18 days ago

https://github.com/mintplex-labs/anythingllm-docs

Guides and technical documentation for AnythingLLM, focusing on Agent Flows and the flow builder. Includes detailed instructions on using blocks such as API Call, LLM Instruction, Web Scraper, and file operations (Read/Write File), as well as managing flow variables, JSON traversal, and debugging agentic workflows.

Tokens
82.4K
Snippets
127
Records
526
Agent score
60%

What's inside AnythingLLM

  1. Overview of AnythingLLM Features

    main

    AnythingLLM provides a comprehensive suite of features for managing LLM-powered applications. Key functional areas include:

    Core AI Capabilities

    • AI Agents: Autonomous agents for task execution.
    • Large Language Models (LLMs): Support for various LLM providers.
    • Embedding Models: Models for generating vector embeddings.
    • Transcription Models: Support for speech-to-text capabilities.
    • Vector Databases: Integration with vector stores for RAG (Retrieval-Augmented Generation).
    • Model Router: Logic for routing requests between different models.
    • Memories: Persistent context and memory management for AI interactions.
    • System Prompt Variables: Dynamic variables for customizing system instructions.

    Interface & Integration

    • API Access & Keys: Programmatic access to AnythingLLM capabilities.
    • Embedded Chat Widgets: Tools to embed chat interfaces into external websites.
    • Chat Modes: Different interaction styles for the chat interface.
    • Appearance Customization: UI/UX personalization options.

    Management & Security

    • Security & Access: Controls for user permissions and access.
    • Privacy & Data Handling: Protocols for managing sensitive information.
    • Chat Logs: History and auditing of chat interactions.
    • Event Logs: System-level logging for monitoring.

    Deployment

    • Cloud Deployment: Options for running AnythingLLM in cloud environments (e.g., via Docker).

  2. Overview of AnythingLLM Meeting Assistant

    main

    The Meeting Assistant is a free, desktop-exclusive feature available in AnythingLLM Desktop v1.10.0 and later. It allows users to record, transcribe, and summarize meetings locally on their device.

    Key capabilities include:

    • Recording & Transcription: Record live meetings or upload existing audio/video files (e.g., podcasts, YouTube videos).
    • Speaker Identification: Automatically identifies different speakers within the transcript.
    • Summarization: Generates summaries using your configured LLM provider (Cloud or Local). Summary prompts can be fully customized.
    • Agentic Follow-ups: Proposes actions based on the transcript/summary that can be executed via available agent skills or MCPs.
    • Search & Chat: Indexes meetings for semantic or keyword search (by topic, speaker, or date) and allows chatting directly with transcripts to ask specific questions.

    Note: This feature is currently available on MacOS and Windows. Linux support (x64 and ARM64) is coming soon.

  3. Overview of Scheduled Jobs

    main

    Scheduled Jobs allow you to automate AI agent tasks on a recurring schedule. You can define a prompt, a schedule (using cron expressions), and a specific subset of tools the agent is permitted to use.

    Key Features:

    • Unattended Execution: Agents run on a schedule without manual intervention.
    • Full Tracing: Every execution (Run) records the agent's thoughts, tool calls, and final responses.
    • File Persistence: Any files generated by the agent (e.g., charts, documents) are saved and available for download.
    • Automatic Workspace: AnythingLLM automatically creates a dedicated 'Scheduled Jobs' workspace to store results and allow you to continue job runs as normal conversation threads.
    • Push Notifications: Optional browser notifications can be configured to alert you when a job completes.

    ⚠️ Critical Requirements:

    • Version: Requires AnythingLLM v1.13.0 or later.
    • Mode: Available only in single-user mode. Self-hosted instances running in multi-user mode will not have access to the Scheduled Jobs settings page.

  4. Overview of AnythingLLM's Desktop Assistant

    main

    The AnythingLLM Desktop Assistant is an on-screen overlay available in AnythingLLM Desktop v1.11.0 and later. It provides a system-wide interface that allows you to chat with agents and interact with the context of any open application (such as Chrome, Slack, or VS Code) using a single keystroke.

    Key Capabilities:

    • System-wide access: Operates as an overlay across your entire OS.
    • Context Awareness: Can chat with any open application by leveraging screen context.
    • Extensibility: Leverages installed agent skills and Model Context Protocol (MCP) tools.
    • Model Flexibility: Works with both Cloud and Local LLM providers.

  5. What is the AnythingLLM Community Hub?

    main

    The AnythingLLM Community Hub (https://hub.anythingllm.com) is a platform and marketplace designed for AnythingLLM users to share and discover various configurations and capabilities. It allows users to share items, skills, and workflows either publicly or privately with the community.

    Currently in beta, the hub supports the following item types:

    • Agent Skills: Specialized capabilities for AI agents.
    • System Prompts: Pre-configured instructions for LLM behavior.
    • Slash Commands: Custom commands used within the AnythingLLM interface.

  6. What is included in AnythingLLM Pro?

    main

    AnythingLLM Pro provides unlimited daily usage of all Magic features (system-wide AI tools) and additional professional perks.

    Magic Features

    Magic features are AI tools that work system-wide across your entire computer, privately and on-device. While they have a free tier, Pro removes all daily usage limits.

    • Magic Echo: Smart dictation that cleans your words and understands screen context anywhere on your computer.
    • Magic Beacon: Allows you to highlight any content on your screen to instantly ask questions, revise text, or perform research without switching applications.
    • Magic Tab: An autocomplete feature that finishes sentences as you type in any application. You can press Tab to accept suggestions; only accepted text counts toward usage.

    Additional Pro Perks

    • Watermark-Free Documents: AI-generated documents are produced without AnythingLLM branding, making them ready for professional sharing.
    • Future Pro Features: All new Pro features released in the future are automatically included in your subscription.

  7. What is AnythingLLM

    main

    AnythingLLM is an all-in-one AI application designed for Retrieval Augmented Generation (RAG), AI Agents, and more. It is built to provide a zero-setup, private environment for local LLMs, or a fully-customizable, private enterprise solution for businesses.

    Key use cases include:

    • Personal/Local Use: A zero-setup, private application for local LLMs and RAG.
    • Business/Organization Use: A customizable, private application that functions like a permissioned ChatGPT, supporting any LLM, embedding model, or vector database.

  8. Capabilities of the Google Calendar Agent

    main

    The Google Calendar Agent provides tools for viewing, creating, and managing calendar events. Actions marked with ✏️ will require explicit user approval in AnythingLLM.

    1. Calendars (Read-only)

    • List Calendars: List all owned or subscribed calendars (names, IDs, time zones).
    • Get Calendar: Get specific details for a calendar by ID.

    2. View Events (Read-only)

    • Get Event: Retrieve full details for a single event by ID.
    • Get Events for Day: List all events for a specific date.
    • Get Events: Search for events within a date range or by keyword.
    • Get Upcoming Events: List events starting from the current time.

    3. Create Events (Requires Approval ✏️)

    • Quick Add: Create events using natural language (e.g., "add a meeting with John tomorrow at 3pm").
    • Create Event: Full control over properties including timed, all-day, multi-day, recurring events, guests, locations, and descriptions.

    4. Update Events (Requires Approval ✏️)

    • Update Event: Modify existing event titles, descriptions, locations, times, or guest lists. Only provided fields are changed.

    5. RSVP (Requires Approval ✏️)

    • Set My Status: Manage attendance with statuses: YES, NO, MAYBE, or INVITED.

  10. Configure Transcription Models in AnythingLLM

    main

    AnythingLLM allows you to configure custom audio transcription providers to convert audio into text. You can choose between local providers that run on your own hardware or cloud-based providers that use external APIs.

    Local Transcription Model Providers

    Use local providers if you want to keep audio processing on your machine without sending data to the cloud.

    • Built-in (Xenova): A local option provided directly within AnythingLLM.

    Cloud Transcription Model Providers

    Use cloud providers for high-performance transcription via external APIs.

    • OpenAI: Uses OpenAI's transcription services (e.g., Whisper) via API.

  12. Features of AnythingLLM Mobile

    main

    AnythingLLM Mobile provides a full local and synced LLM experience on Android:

    • Local SLM Chat: Chat with small language models (SLMs) directly on your device, supporting both reasoning and non-reasoning models.
    • On-device RAG: Locally process documents using a small embedding model and a local vector database for fully offline RAG with citations.
    • Model Swapping: Change models on the fly.
    • Organization: Create and manage workspaces and threads.
    • Agentic Tools: Access tools like web search, web scraping, deep research, and cross-app interactions (e.g., drafting emails or managing calendars).
    • Syncing: Sync all data with Desktop or Cloud instances.