langgraph stride.stub

---

id: langgraph-stride name: LangGraph_STRIDE repo_path: ~/Documents/sw30labs/repos/LangGraph_STRIDE remote_url: https://github.com/sw30labs/LangGraph_STRIDE.git primary_language: Python framework: LangGraph + OpenAI-compatible (vLLM on DGX Spark / oMLX on Apple Silicon) category: ai-security stacks: [langgraph, agentic, pentest, dgx, mlx] ingested: 2026-08-28 last_commit_date: 2026-08-12

LangGraph_STRIDE

Purpose

Turns application inventory into STRIDE/DREAD threat models, then into CVE, CTI, and cyber-threat scenario reports. One codebase targets vLLM on NVIDIA DGX Spark and oMLX on Apple Silicon through the same OpenAI-compatible client. Platform detection changes defaults, never workflow logic. Personal research tool; not employer-affiliated.

Architecture

• src/config.py — host detection, env resolution, Pydantic config tree.
• src/inference_client.py — URL normalization, model discovery, thinking removal, truncation policy, shared client cache.
• src/threat_model.py — six-node STRIDE/DREAD generation, QA retry, final QA, persistence.
• src/scenario.py — eight-node CVE/CTI/scenario selection, generation, approval, persistence.
• src/gui.py + src/web/ — local job console, report browser, logs.
• src/workspace.py — workspace CMDB, attack-chain library, Application → model → scenario navigator.
• src/maintain/ — Maintain studio: novel + KEV-gap chain forge, clone autopsy, force-layout viz, Neo4j projection (optional, dedicated compose on :7475/:7688 — separate from TwinOps).

Dependencies

Notable: httpx>=0.27, langgraph>=0.4, neo4j>=5.14, prometheus-client>=0.20, pydantic>=2.7, python-dotenv>=1.0, pyyaml>=6.0, rich>=13.7, typer>=0.12. Optional graph extra for Neo4j driver.

CI/CD

gate.sh plus platform setup scripts (setup_and_run_macos.sh, setup_and_run_dgx.sh). Compose files for DGX Spark and Neo4j.

Entities and topics

• stride-lite — sibling STRIDE/DREAD + Vault research tool (oMLX-first)
• tslit — integrity-testing lineage (collapsed); different threat axis
• Tags: [langgraph, threat-modeling, stride, dread, dgx, mlx, cti]