How snowlights2022/maibot works
MaiBot is a conversational LLM agent designed for lifelike interaction in group chats, with persistent personal memory, extensibility through plugins, and a browser-based management dashboard. The documented user workflow is incoming conversation → agent reasoning and response → conversation platform; memory ingestion and retrieval support that interaction. The graph also preserves dashboard management and monitoring capabilities. Source excerpts are partial and do not expose the host’s full message-dispatch wiring, so uncertain connections between the host, reasoning engine, and memory runtime are omitted.
Main entry points: router.tsx Dashboard app · ingest_service.py Memory ingestion · reasoning_engine.py Reasoning engine · aggregate_query_service.py Memory retrieval
Components 19
Conversation Agent
- Bot host
bot.py - Reasoning engine
src/maisaka/reasoning_engine.py - LLM service
src/services/llm_service.py
Memory System
- Memory ingestion
src/A_memorix/core/runtime/services/ingest_service.py - Memory retrieval
src/A_memorix/core/utils/aggregate_query_service.py - Memory storage
src/A_memorix/core/storage/metadata_store.py - Vector storage
src/A_memorix/core/storage/vector_store.py - Embedding service
src/services/embedding_service.py - Retrieval tuning
src/A_memorix/core/utils/retrieval_tuning_manager.py
Extensions
- Plugin runtime
src/plugin_runtime/dependency_pipeline.py
Management Dashboard
- Dashboard app
dashboard/src/router.tsx - Chat workspace
dashboard/src/routes/chat/index.tsx - Memory management
dashboard/src/routes/resource/knowledge-base.tsx - Plugin management
dashboard/src/routes/plugins/PluginMarketplacePage.tsx - Runtime monitoring
dashboard/src/routes/monitor/index.tsx
Other components
- Chat participant
- Chat platform
- LLM provider
- MaiBot backend
Connections 21
- Chat participant → Chat platform: sends messages
- Chat platform → Bot host: delivers messages
- Bot host → Reasoning engine: triggers turns
- Reasoning engine → LLM service: requests generation
- LLM service → LLM provider: dispatches requests
- Reasoning engine → Chat platform: produces replies
- Memory ingestion → Memory storage: writes metadata
- Memory ingestion → Vector storage: writes vectors
- Memory ingestion → Embedding service: requests embeddings
- Memory retrieval → Memory storage: reads metadata
- Memory retrieval → Vector storage: searches vectors
- Retrieval tuning → Memory retrieval: configures retrieval
- Dashboard app → Chat workspace: routes to
- Dashboard app → Memory management: routes to
- Dashboard app → Plugin management: routes to
- Dashboard app → Runtime monitoring: routes to
- Chat workspace → MaiBot backend: uses backend APIs
- Memory management → MaiBot backend: uses memory APIs
- Plugin management → MaiBot backend: uses plugin APIs
- Runtime monitoring → MaiBot backend: uses monitoring APIs
- MaiBot backend → Plugin runtime: manages plugins