How sii-1111/ai-engineering-command-center works
AI Engineering Command Center is a scaffold for investigating engineering tasks and preparing evidence-backed changes, with human approval before risky mutations. The documented workflow starts in the web workspace, enters the API, and is orchestrated as a task graph: plan and investigate through GitHub tools, propose a change, pause for approval, then verify and review. Evaluation and observability capture outcomes. Source confirms the API, planner, GitHub MCP boundary, reviewer, evaluation, and state models; several broader capabilities are documented but not yet represented by sampled implementations, so their wiring is omitted.
Main entry points: graph.py Task Graph · main.py FastAPI Service · llm_planner.py Investigation Planner · github_investigator.py Research Agent
Components 19
User Workspace
- Web Command Center
apps/web/app/page.tsx - FastAPI Service
apps/api/app/main.py
Task Orchestration
- Task Graph
core/graph.py - Investigation Planner
agents/llm_planner.py - Research Agent
agents/github_investigator.py - Code Agent
agents/planner.py - Testing Agent
agents/tool_executor.py - Azure OpenAI Client
core/llm.py
Tooling and Changes
- MCP Client
core/mcp_client.py - GitHub MCP Tools
tooling/github/server.py - GitHub Client
tooling/github/client.py
Assurance and State
- Reviewer Agent
agents/reviewer.py - Engineering State
core/state/models.py - Memory Layer
core/memory - Task Evaluation
core/evaluation/metrics.py - Event Recording
core/observability.py
Other components
- Engineer
- GitHub
- Human Approval
Connections 24
- Engineer → Web Command Center: submits task
- Web Command Center → FastAPI Service: sends task
- FastAPI Service → Task Graph: invokes graph
- Task Graph → Investigation Planner: routes task
- Investigation Planner → Azure OpenAI Client: requests plan
- Investigation Planner → Research Agent: plans investigation
- Task Graph → Research Agent: dispatches research
- Task Graph → Code Agent: dispatches change work
- Task Graph → Testing Agent: dispatches verification
- Task Graph → Reviewer Agent: dispatches review
- Research Agent → MCP Client: calls GitHub tools
- Testing Agent → MCP Client: executes tools
- MCP Client → GitHub MCP Tools: calls tools
- GitHub MCP Tools → GitHub Client: invokes API client
- GitHub Client → GitHub: reads or writes
- Task Graph → Engineering State: updates task state
- Task Graph → Memory Layer: uses memory
- FastAPI Service → Human Approval: requests decision
- Human Approval → FastAPI Service: submits decision
- FastAPI Service → Task Evaluation: evaluates task
- Task Evaluation → Event Recording: records event
- FastAPI Service → Engineering State: reads task state
- FastAPI Service → Task Graph: resumes task
- Reviewer Agent → Azure OpenAI Client: requests review