How ld-singh/ai-sdlc-architecture-review-agent works
AI-SDLC is a human-in-the-loop architecture review agent: a user submits an issue or PR change, repository context is retrieved, specialist agents assess delivery risks, and the system emits an auditable decision pack and approval recommendation. The main review path is supported by the README and sampled CLI, retrieval, agent, and graph sources. Optional vector retrieval, live GitHub input, model comparison, and evaluation are shown as separate capabilities; unsampled wiring is not inferred beyond documented or indexed references.
Main entry points: graph.py Agent Graph · llm.py LLM Client · main.py Review CLI · context_agent.py Context Agent
Components 29
Inputs and CLI
- Issue or PR
- Review CLI
src/ai_sdlc/main.py - Diff Parser
src/ai_sdlc/tools/diff_parser.py
Repository Context
- Review Repository
- File Loader
src/ai_sdlc/retrieval/file_loader.py - Chunk and Index
src/ai_sdlc/retrieval/indexer.py - Context Retrieval
src/ai_sdlc/retrieval/retriever.py - Embeddings
src/ai_sdlc/retrieval/embeddings.py - Vector Store
src/ai_sdlc/retrieval/vector_store.py
Review Agents
- Agent Graph
src/ai_sdlc/graph.py - Context Agent
src/ai_sdlc/agents/context_agent.py - Architecture Impact
src/ai_sdlc/agents/architecture_agent.py - CI/CD Risk
src/ai_sdlc/agents/cicd_agent.py - Security Governance
src/ai_sdlc/agents/security_agent.py - Testing Strategy
src/ai_sdlc/agents/testing_agent.py - ADR and PR Comment
src/ai_sdlc/agents/adr_agent.py
Runtime Services
- Review State
src/ai_sdlc/models.py - LLM Client
src/ai_sdlc/llm.py - Output Guardrails
src/ai_sdlc/guardrails.py - Run Metrics
src/ai_sdlc/observability/metrics.py - Token Estimator
src/ai_sdlc/tools/token_estimator.py - Markdown Renderer
src/ai_sdlc/tools/markdown.py
Decision and Evaluation
- Decision Pack
- Golden-case Eval
src/ai_sdlc/evals/runner.py - Model Comparison
src/ai_sdlc/benchmarks.py
Other components
- Engineer
- GitHub CLI
- LLM Provider
- Named Reviewer
Connections 41
- Engineer → Issue or PR: submits
README.md - Issue or PR → Review CLI: reviews
README.md - GitHub CLI → Review CLI: fetches read-only
README.md - Review CLI → Diff Parser: parses diff
src/ai_sdlc/main.py - Review CLI → Agent Graph: runs review
src/ai_sdlc/main.py - Review Repository → File Loader: loads files
README.md - File Loader → Chunk and Index: chunks files
src/ai_sdlc/retrieval/indexer.py - Context Retrieval → File Loader: loads files
src/ai_sdlc/retrieval/retriever.py - Context Retrieval → Chunk and Index: ranks chunks
src/ai_sdlc/retrieval/retriever.py - Context Retrieval → Embeddings: embeds chunks
src/ai_sdlc/retrieval/retriever.py - Context Retrieval → Vector Store: searches vectors
src/ai_sdlc/retrieval/retriever.py - Agent Graph → Context Agent: dispatches
src/ai_sdlc/graph.py - Agent Graph → Architecture Impact: dispatches
src/ai_sdlc/graph.py - Agent Graph → CI/CD Risk: dispatches
src/ai_sdlc/graph.py - Agent Graph → Security Governance: dispatches
src/ai_sdlc/graph.py - Agent Graph → Testing Strategy: dispatches
src/ai_sdlc/graph.py - Agent Graph → ADR and PR Comment: dispatches
src/ai_sdlc/graph.py - Agent Graph → Review State: threads state
src/ai_sdlc/models.py - Context Agent → Context Retrieval: retrieves context
src/ai_sdlc/agents/context_agent.py - Context Agent → Architecture Impact: provides context
README.md - Architecture Impact → CI/CD Risk: feeds findings
README.md - CI/CD Risk → Security Governance: feeds findings
README.md - Security Governance → Testing Strategy: feeds findings
README.md - Testing Strategy → ADR and PR Comment: feeds findings
README.md - Context Agent → LLM Client: completes prompts
src/ai_sdlc/agents/context_agent.py - Architecture Impact → LLM Client: requests findings
src/ai_sdlc/agents/architecture_agent.py - CI/CD Risk → LLM Client: requests analysis
src/ai_sdlc/agents/cicd_agent.py - Security Governance → LLM Client: requests analysis
src/ai_sdlc/agents/security_agent.py - Testing Strategy → LLM Client: requests analysis
src/ai_sdlc/agents/testing_agent.py - ADR and PR Comment → LLM Client: requests synthesis
src/ai_sdlc/agents/adr_agent.py - LLM Client → LLM Provider: calls provider
src/ai_sdlc/llm.py - LLM Client → Output Guardrails: validates output
src/ai_sdlc/llm.py - LLM Client → Token Estimator: estimates usage
src/ai_sdlc/llm.py - Agent Graph → Run Metrics: records metrics
src/ai_sdlc/graph.py - Context Agent → Markdown Renderer: renders report
src/ai_sdlc/agents/context_agent.py - Agent Graph → Decision Pack: writes decision pack
README.md - Decision Pack → Named Reviewer: requests approval
README.md - Review CLI → Golden-case Eval: runs eval
src/ai_sdlc/main.py - Golden-case Eval → Agent Graph: evaluates reviews
src/ai_sdlc/evals/runner.py - Review CLI → Model Comparison: compares models
src/ai_sdlc/main.py - Model Comparison → Agent Graph: runs comparisons
src/ai_sdlc/benchmarks.py