How jacobbweber/companion-ai-os works
Companion is a local-first assistant and autonomous engineering workspace. A user enters through the web UI or CLI, sends a chat request, and receives a streamed response or tool-driven result; persistent memory and project state support that interaction. Separate product branches cover projects, calendar, learning and wiki, background routines, and governed engineering workflows. The graph includes the principal runtime and documented capabilities. Some domain modules were not sampled, so their internal call paths are not asserted.
Main entry points: agent.py Assistant agent · memory_store.py Memory store · web_server.py Web server · routines.py Routine manager
Components 22
User interfaces
- Web Studio
frontend/src/App.tsx - Web server
surfaces/web_server.py - Web API
surfaces/api_extensions.py - CLI entry
harness.py
Assistant runtime
- Assistant agent
companion/agent.py - LLM router
companion/llm.py - Tool execution
companion/tools.py - Response streaming
companion/streaming.py
Memory and projects
- Memory store
companion/memory_store.py - Project graph
companion/project_graph.py
Knowledge and planning
- Unified calendar
companion/calendar_engine.py - Knowledge wiki
companion/wiki_engine.py - Learning engine
companion/education_engine.py
Automation and governance
- DAG runtime
companion/dag_runtime.py - Workflow graph
companion/workflow_graph.py - Routine manager
companion/routines.py - Background jobs
companion/background_jobs.py - Approvals and trust
companion/archive_and_approvals.py - Safety controls
companion/safety.py - Credential vault
companion/vault.py - MCP hub
companion/mcp_hub.py
Other components
- User
Connections 26
- User → Web Studio: uses
README.md - Web Studio → Web server: API requests
- Web Studio → Web API: calls
frontend/src/App.tsx - Web server → Assistant agent: dispatches turns
surfaces/web_server.py - Web server → Web API: routes extensions
surfaces/web_server.py - CLI entry → Assistant agent: starts assistant
harness.py - Assistant agent → LLM router: requests completion
companion/agent.py - Assistant agent → Memory store: reads and writes
companion/agent.py - Assistant agent → Tool execution: dispatches tools
companion/agent.py - Assistant agent → Project graph: uses project context
companion/agent.py - Web server → Response streaming: streams responses
surfaces/web_server.py - LLM router → Response streaming: uses stream support
companion/llm.py - Project graph → Memory store: persists graph state
companion/project_graph.py - Unified calendar → Memory store: uses stored state
companion/calendar_engine.py - Learning engine → Memory store: stores mastery
companion/education_engine.py - Learning engine → Knowledge wiki: curates study notes
companion/education_engine.py - Routine manager → Background jobs: manages job routines
companion/routines.py - Routine manager → DAG runtime: runs DAG routines
companion/routines.py - Routine manager → Approvals and trust: requests approvals
companion/routines.py - DAG runtime → Memory store: uses persistent state
companion/dag_runtime.py - Workflow graph → Memory store: persists workflow state
- User → CLI entry: uses
README.md - Approvals and trust → Memory store: stores approval state
companion/archive_and_approvals.py - Safety controls → Assistant agent: guards actions
companion/agent.py - Credential vault → Tool execution: injects credentials
- MCP hub → Tool execution: provides integrations