How mintplex-labs/anything-llm works
AnythingLLM is a multi-user AI application for chatting with documents, configuring model providers, and using agents. The main workflow starts with a user uploading or connecting documents, which the collector parses and the server embeds for workspace retrieval; chat requests then reach server-side chat handling and return model-generated responses to the frontend. The README also documents model routing, memories, scheduled tasks, agent tools, and integrations. Some subsystem links below rely on README descriptions because their implementation files were not sampled.
Main entry points: collector/index.js Collector API · apiChatHandler.js Chat handler · server/index.js HTTP application · router/index.js Model routing
Components 20
User interface
- Workspace chat
frontend/src/components/WorkspaceChat/index.jsx - Document management
frontend/src/components/Modals/ManageWorkspace/Documents/index.jsx - Connector settings
frontend/src/components/Modals/ManageWorkspace/DataConnectors/index.jsx
Server and chat
- HTTP application
server/index.js - Chat handler
server/utils/chats/apiChatHandler.js - Model routing
server/utils/router/index.js - Router configuration
server/endpoints/modelRouter.js
Document pipeline
- Collector API
collector/index.js - File conversion
collector/processSingleFile/index.js - Link processing
collector/processLink/index.js - Text normalization
collector/processRawText/index.js - Data connectors
collector/utils/extensions - Embedding worker
server/jobs/embedding-worker.js
AI and retrieval
- Vector database
server/utils/helpers/index.js - Model providers
server/utils/helpers/index.js
Workspace capabilities
- Agents and tools
server/utils/agents/aibitat - Workspace memories
server - Scheduled tasks
server
Other components
- User
- External sources
Connections 23
- User → Workspace chat: submits prompts
- User → Document management: uploads documents
- User → Connector settings: configures sources
- Workspace chat → HTTP application: sends chat requests
- HTTP application → Chat handler: routes chat
server/index.js - Chat handler → Model routing: selects model
server/utils/chats/apiChatHandler.js - Model routing → Model providers: dispatches inference
- Chat handler → Vector database: retrieves context
server/utils/chats/apiChatHandler.js - Chat handler → Agents and tools: invokes agents
server/utils/chats/apiChatHandler.js - Model routing → Router configuration: uses routing rules
server/endpoints/modelRouter.js - Document management → HTTP application: uploads files
- HTTP application → Collector API: requests processing
- Collector API → File conversion: processes files
collector/index.js - Collector API → Link processing: processes links
collector/index.js - File conversion → Text normalization: produces document text
- Link processing → Text normalization: produces document text
- Connector settings → Collector API: requests source loads
- External sources → Data connectors: provides content
- Data connectors → Collector API: loads source data
- Collector API → Embedding worker: hands off documents
- Embedding worker → Vector database: stores embeddings
server/jobs/embedding-worker.js - Chat handler → Workspace memories: supports personalization
server/utils/chats/apiChatHandler.js - Agents and tools → Scheduled tasks: runs scheduled prompts
server/utils/agents/aibitat/plugins/sql-agent/query.js