# PHIROK

> PHIROK by KRONS Systems is a platform for verifiable memory and persistent evidence for AI systems.

PHIROK records evidence and context, with cryptographic receipts, append-only history, tenant-scoped separation, non-generative GraphRAG retrieval and deterministic reconstruction of stored memory state.

## Read the site

- [PHIROK — Verifiable memory for AI in production](https://phirok-evidence.lferro.chatgpt.site/): Verifiable memory and persistent evidence for AI systems. Explore cryptographic receipts, tenant-scoped records and deterministic replay.
- [Platform — PHIROK](https://phirok-evidence.lferro.chatgpt.site/platform/): Explore evidence-first history, cryptographic receipts, append-only records, tenant separation, traceable retrieval and replay.
- [Architecture — PHIROK](https://phirok-evidence.lferro.chatgpt.site/architecture/): Follow authenticated evidence capture into append-only storage, cryptographic lineage, retrieval and review.
- [Trust model — PHIROK](https://phirok-evidence.lferro.chatgpt.site/trust/): Understand the boundaries of integrity, provenance, tenant isolation and replay. Clear mechanisms and explicit limits.
- [Evidence playground — PHIROK](https://phirok-evidence.lferro.chatgpt.site/playground/): Compare sample decision records and test a local cryptographic integrity check. Inspect evidence and understand what a hash can prove.
- [Developers — PHIROK](https://phirok-evidence.lferro.chatgpt.site/developers/): Read the PHIROK evidence model, integration boundaries and evaluation criteria. Plain HTML and Markdown for people and agents.
- [Evaluate PHIROK — Get started](https://phirok-evidence.lferro.chatgpt.site/access/): Prepare an evaluation brief for a consequential AI workflow, or continue to the PHIROK application for account access.
- [Site privacy — PHIROK](https://phirok-evidence.lferro.chatgpt.site/privacy/): How the PHIROK website handles local preferences, demonstration input, downloads and links to its application.

## Boundaries

- A receipt does not make a source factually true.
- Replay concerns retained memory state, not a guarantee of identical model generations or access to hidden reasoning.
- The current semantic layer uses deterministic test fixtures. Commercial production embeddings are not advertised.
- No mathematical hallucination prevention, live contradiction detection or regulatory certification is claimed.
- Interactive examples on this site are illustrative. The local SHA-256 example is not PHIROK's production BLAKE3 receipt implementation.
- This site does not publish an API endpoint or a versioned API request contract. Obtain deployment-specific documentation before integrating.

## Technical brief

- [Full plain-text overview](https://phirok-evidence.lferro.chatgpt.site/llms-full.txt)