How tencentcloud/tencentdb-agent-memory works
This repository combines a layered agent-memory service with a knowledge hub, a cross-agent proxy, and a management panel. The main memory workflow captures conversations and distills them through L0–L3; agents can then recall memory and use skills, while the hub adds Wiki and CodeGraph capabilities. The graph includes documented product branches and only source-level relationships visible in the excerpts or SOURCE INDEX; several unsampled subsystem links are therefore omitted.
Main entry points: server.ts Knowledge API · manager.ts Wiki ingestion and search · auto-capture.ts Conversation capture · l0-recorder.ts L0 recording
Components 20
Memory Core
- Conversation capture
MemoryCore/src/core/hooks/auto-capture.ts - L0 recording
MemoryCore/src/core/conversation/l0-recorder.ts - L1 extraction
MemoryCore/src/core/record/l1-extractor.ts - L2 scenes
MemoryCore/src/core/scene/scene-extractor.ts - L3 persona
MemoryCore/src/core/persona/persona-generator.ts - Memory recall
MemoryCore/src/core/hooks/auto-recall.ts - Skill library
MemoryCore/src/core/skill/skill-core.ts - Memory storage
MemoryCore/src/core/storage/adapter.ts - Memory gateway
MemoryCore/src/gateway/server.ts
Knowledge Hub
- Knowledge API
MemoryKnowledge/src/server.ts - Wiki ingestion and search
MemoryKnowledge/src/engines/wiki/manager.ts - Wiki index
MemoryKnowledge/src/engines/wiki/index-db.ts - External Wiki import
MemoryKnowledge/src/wiki-source/routes.ts - CodeGraph
MemoryKnowledge/src/engines/code/index.ts
Agent Proxy
- Request injection
MemoryProxy/src/injection/pipeline.ts - Proxy service
MemoryProxy/src/server.ts
Management Panel
- Wiki management routes
MemoryPanel/src/panel/http/routes/knowledge/wiki-routes.ts - Knowledge client
MemoryPanel/src/panel/kernel/adapters/http-knowledge-client.ts
Other components
- Agent clients
- Team members
Connections 17
- Agent clients → Conversation capture: send conversations
README.md - Conversation capture → L0 recording: record messages
README.md - L0 recording → L1 extraction: distill atoms
README.md - L1 extraction → L2 scenes: derive scenes
README.md - L2 scenes → L3 persona: synthesize persona
README.md - Agent clients → Memory recall: request recall
README.md - Agent clients → Skill library: use skills
README.md - Memory gateway → Memory storage: adapt storage
- Knowledge API → Wiki ingestion and search: serve Wiki
MemoryKnowledge/src/server.ts - Knowledge API → CodeGraph: serve CodeGraph
MemoryKnowledge/src/server.ts - Wiki ingestion and search → Wiki index: read index
MemoryKnowledge/src/engines/wiki/manager.ts - External Wiki import → Wiki ingestion and search: import pages
MemoryKnowledge/src/wiki-source/routes.ts - Proxy service → Request injection: run injection
MemoryProxy/src/server.ts - Agent clients → Proxy service: send requests
README.md - Team members → Wiki management routes: manage Wiki
README.md - Wiki management routes → Knowledge client: call client
MemoryPanel/src/panel/http/routes/knowledge/wiki-routes.ts - Knowledge client → Knowledge API: call API