Open-source VeriChron AI rebuilds past compliance states with bitemporal data, knowledge graphs and agents
LoksaiPalyam · reddit · 2026-09-29
Developer LoksaiPalyam released VeriChron AI, an open-source concept for enterprise compliance and audit intelligence built on bitemporal data, knowledge graphs, long-term memory, and AI agents. The core question: "What was our actual compliance posture at a specific point in the past — and what did we know at that time?"
Bitemporal compliance
- Valid Time: when something was actually true in the real world
- System Time: when the system became aware of it
Example: MFA was disabled April 12, the GRC system only detected it May 20, an audit finding was created June 3, fixed July 8. Asked about May 15, the system must reconstruct state without leaking future knowledge.
Architecture
- Next.js console → FastAPI backend → compliance agent pipeline (OpenClaw Gateway + Groq)
- Neo4j knowledge graph linking policies, controls, assets, evidence, incidents via GOVERNS, SATISFIEDBY, EVIDENCEDBY, VIOLATES, CAUSES relationships
- Hindsight Cloud memory layer preserving raw observations via retain → recall → reflect → creatementalmodel, instead of flattening to latest state
- A "Time Machine" UI lets auditors reconstruct compliance state at any past point
Ingestion flow: raw log → temporal extraction → Neo4j graph → Hindsight memory → recall + reflection → historical reconstruction → Groq → structured compliance assessment.
More from coding & agent
- 50 researchers drop 303-page guide: RLVR lets small open models match giants on coding — mdancho84 · 2026-09-29
- If AI Coding Agents Get Reliable Enough, Frameworks May Disappear Entirely — bendee983 · 2026-09-29
- Langfuse open-sources MCP Server to debug agent runs without raw telemetry — blaizedsouza · 2026-09-29
- The Claude Code creator's 30-min vibe-coding workshop beats 100 YouTube guides — HeyAmit_ · 2026-09-29
- One Prompt, One Chat: Generating a Brand Motion-Graphics Video with Claude Code — TawohAwa · 2026-09-29
- The overlooked async RL curation pitfall: task runtime differences skew sampling — auto_grad_ · 2026-09-29