How offloadmemory/agentfork works
AgentFork is a multi-tenant platform for authoring, evaluating, and delivering AI agents. Teams configure agents and workflows, connect models and knowledge sources, validate behavior in a playground or evaluation suite, then publish versioned agents for inference through APIs and chat channels. Knowledge ingestion and retrieval, analytics, and access controls support this lifecycle. The tree and README establish several capabilities; sampled source confirms selected runtime paths, not every UI-to-API connection or subsystem implementation.
Main entry points: inference/route.ts Inference API · knowledge-bases/route.ts Knowledge base API · llm-provider-service.ts LLM provider management · agents/route.ts Agent management API
Components 23
Agent authoring
- Agent studio
apps/web-ui/app/(dashboard)/agents/page.tsx - Agent management API
apps/web-ui/app/api/agents/route.ts - Workflow engine
libs/shared/src/workflow/workflow-engine.ts - Agent playground
apps/web-ui/app/(dashboard)/agents/playground/page.tsx
Knowledge and retrieval
- Knowledge base workspace
apps/web-ui/app/(dashboard)/knowledge-bases/page.tsx - Knowledge base API
apps/web-ui/app/api/knowledge-bases/route.ts - Web crawler
libs/knowledge-base/src/web-crawler/crawlee-engine.ts
Inference and delivery
- Inference API
apps/web-ui/app/api/v1/inference/route.ts - Inference sessions
libs/shared/src/services/inference-session-service.ts - Embeddable widget
apps/sdk/src/components/smc-chat-widget/smc-chat-widget.tsx - Widget API client
apps/sdk/src/services/api.service.ts - Channel management
apps/web-ui/app/(dashboard)/connectors/page.tsx
Evaluation and operations
- Evaluation suite
apps/web-ui/app/(dashboard)/evaluation/experiments/page.tsx - Analytics dashboards
apps/web-ui/app/(dashboard)/analytics/page.tsx - Dashboard query service
libs/shared/src/services/dashboard-query-service.ts - Agent operations console
apps/mission-control/app/(console)/dashboard/page.tsx
Platform services
- LLM provider management
libs/shared/src/services/llm-provider-service.ts - MCP server management
apps/web-ui/app/(dashboard)/mcp-servers/page.tsx - Platform database
Other components
- Platform team
- LLM providers
- Agent end user
- External tools
Connections 28
- Platform team → Agent studio: designs agents
- Agent studio → Agent management API: manages agents
- Platform team → Knowledge base workspace: configures sources
- Knowledge base workspace → Knowledge base API: manages knowledge
- Knowledge base API → Web crawler: starts crawls
- Platform team → Agent playground: tests agents
- Agent playground → Agent management API: tests configuration
- Platform team → Evaluation suite: evaluates behavior
- Platform team → LLM provider management: configures models
- LLM provider management → Platform database: stores provider records
- LLM provider management → LLM providers: discovers models
- Agent management API → Platform database: stores agent data
- Knowledge base API → Platform database: stores knowledge data
- Inference API → Inference sessions: manages sessions
- Inference sessions → Platform database: persists session state
- Agent end user → Embeddable widget: chats through
- Embeddable widget → Widget API client: sends messages
- Widget API client → Inference API: requests inference
- Agent end user → Inference API: calls REST inference
- Inference API → LLM providers: generates responses
- Inference API → Workflow engine: resolves workflow turns
- Inference API → MCP server management: loads tool configuration
- Inference API → External tools: invokes tools
- Platform team → Channel management: configures delivery
- Platform team → Analytics dashboards: reviews insights
- Analytics dashboards → Dashboard query service: queries metrics
- Dashboard query service → Platform database: reads tenant data
- Platform team → Agent operations console: operates agent runtime