Architectural Blueprint: This kit is currently a planned architectural blueprint.
The specifications, threat model, and rules below describe the planned implementation. Starter source containers and automated test suites are in active development.
Blueprint Specification
v1.0.0
LangGraph + LangChain + Qdrant + Ragas Evals (Blueprint)
Agentic RAG with cyclical state graph execution: agent decides dynamically when to retrieve, when to re-write queries, and when to invoke human-in-the-loop review with Ragas groundedness evals.
{'name': 'Python', 'version': '3.12+', 'registry': 'pypi.org'} {'name': 'LangGraph', 'version': '0.2+', 'registry': 'pypi.org'}
Agent Bootstrapping
Pre-Wired Rules
Single-source rules compiled to 6 native agent dialects. Transpiled deterministically from rules/source/.
• .cursor/rules/*.mdc
• CLAUDE.md
• .agents/rules/*.md
• .clinerules & CODEX.md
Security Guardrails
Automated Scanners
Invariants asserted in CI before merge. Scanned by Gitleaks and Semgrep AST checkers.
- Planned: Standard Security Scan Protocol Enabled
Docker Scaffolding
Turnkey Containers
Runnable containers with pre-wired environment variables, healthchecks, and dependency locks.
docker compose up -d
Agent Bootstrap Prompt
Feed this exact prompt into Claude Code, Cursor, Copilot, or Antigravity to anchor the agent:
You are building with the LangGraph + LangChain + Qdrant + Ragas Evals kit. Follow rules in AGENTS.md, DESIGN.md tokens, and steps/agent-build-steps.md.