Update README.md
Browse files
README.md
CHANGED
|
@@ -1,13 +1,71 @@
|
|
| 1 |
---
|
| 2 |
title: START HERE Agent Forensics Suite
|
| 3 |
-
emoji:
|
| 4 |
-
colorFrom:
|
| 5 |
-
colorTo:
|
| 6 |
sdk: gradio
|
| 7 |
sdk_version: 6.3.0
|
| 8 |
app_file: app.py
|
| 9 |
-
pinned:
|
| 10 |
short_description: 'Start here: verify agent runs with receipts, replays, diffs.'
|
|
|
|
|
|
|
|
|
|
| 11 |
---
|
|
|
|
| 12 |
|
| 13 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
title: START HERE Agent Forensics Suite
|
| 3 |
+
emoji: 🔥
|
| 4 |
+
colorFrom: indigo
|
| 5 |
+
colorTo: purple
|
| 6 |
sdk: gradio
|
| 7 |
sdk_version: 6.3.0
|
| 8 |
app_file: app.py
|
| 9 |
+
pinned: true
|
| 10 |
short_description: 'Start here: verify agent runs with receipts, replays, diffs.'
|
| 11 |
+
license: other
|
| 12 |
+
thumbnail: >-
|
| 13 |
+
https://cdn-uploads.huggingface.co/production/uploads/685edcb04796127b024b4805/nX59TmzGS6YkzPd5j1kyW.png
|
| 14 |
---
|
| 15 |
+
# RFTSystems — Agent Forensics Suite (START HERE)
|
| 16 |
|
| 17 |
+
**Evidence-first instrumentation for AI agents.**
|
| 18 |
+
Audit, prove, replay, and diff agent runs — turning “trust me” into verification.
|
| 19 |
+
|
| 20 |
+
## Why this exists
|
| 21 |
+
AI is being shipped into real systems faster than teams can reliably reproduce or explain agent behaviour. When an agent fails, too many postmortems still rely on screenshots, partial logs, and opinions — not evidence.
|
| 22 |
+
|
| 23 |
+
The deeper risk is that **nobody can prove what happened**: what the agent saw, what it called, what it wrote, and where the run diverged. When failures are unreproducible, accountability collapses.
|
| 24 |
+
|
| 25 |
+
RFTSystems exists to make agent behaviour **inspectable and independently verifiable**. This suite produces artifacts you can share and validate: hash-chained timelines, tamper-evident receipts, deterministic replays, and first-divergence diffs. You don’t need to trust the author — you can verify the evidence.
|
| 26 |
+
|
| 27 |
+
## Workflow
|
| 28 |
+
**learn → generate proof → record reality → seal it → diff it → audit it → benchmark it**
|
| 29 |
+
|
| 30 |
+
### Quick start (60 seconds)
|
| 31 |
+
1. Open **ReplayProof** and run a deterministic session.
|
| 32 |
+
2. Export the run bundle (receipts + hashes).
|
| 33 |
+
3. Upload the bundle to verify integrity, then share it — anyone can replay it.
|
| 34 |
+
|
| 35 |
+
### Full pipeline (real systems)
|
| 36 |
+
1. **Record reality** (Agent Flight Recorder)
|
| 37 |
+
2. **Seal it** into receipts (RFT Memory Receipt Engine)
|
| 38 |
+
3. **Diff** two runs and find first divergence (TimelineDiff)
|
| 39 |
+
4. **Audit** state transitions and governance evidence (TrustStack)
|
| 40 |
+
5. **Benchmark** verifiable performance signals (Coherent Compute Engine)
|
| 41 |
+
|
| 42 |
+
## The labs
|
| 43 |
+
- **ReplayProof Agent POV Verified Replay** — Fast proof: generate a deterministic run bundle you can verify and replay anywhere.
|
| 44 |
+
- **Agent Flight Recorder** — Chain-of-custody logging: hash-chained events across prompts, tools, outputs, and memory reads/writes.
|
| 45 |
+
- **RFT Memory Receipt Engine** — Proof layer: generate/download tamper-evident receipts; upload to independently verify integrity.
|
| 46 |
+
- **TimelineDiff Differential Reproducibility** — First divergence: align two run bundles and pinpoint exactly where/why they split.
|
| 47 |
+
- **TrustStack Console** — Audit cockpit: inspect runs, compare state, and trace exactly what changed and why.
|
| 48 |
+
- **Coherent Compute Engine** — Verification-first benchmark: live throughput + stability/energy behaviour + downloadable receipt.
|
| 49 |
+
|
| 50 |
+
## Design principle
|
| 51 |
+
We don’t “imprison” agents. We measure drift from declared intent and produce evidence. Enforcement remains an operator decision; this suite is the instrumentation layer.
|
| 52 |
+
|
| 53 |
+
## Licence / Rights Notice
|
| 54 |
+
All materials contained in or associated with this repository — including but not limited to text, code, algorithms, equations, figures, datasets, and documentation — are original works authored by Liam Grinstead and form part of the Rendered Frame Theory (RFT) research framework.
|
| 55 |
+
|
| 56 |
+
These works are protected under the following laws and treaties:
|
| 57 |
+
|
| 58 |
+
- Copyright, Designs and Patents Act 1988 (UK) — ss.1–103 and ss.77–89
|
| 59 |
+
- Trade Secrets (Enforcement etc.) Regulations 2018 (UK) — Regs.2–6
|
| 60 |
+
- Copyright and Rights in Databases Regulations 1997 (UK) — Regs.14–24
|
| 61 |
+
- Berne Convention — Arts.5(2) & 6bis
|
| 62 |
+
- TRIPS Agreement — Arts.9–14
|
| 63 |
+
|
| 64 |
+
All rights are reserved.
|
| 65 |
+
|
| 66 |
+
No part of this work may be copied, reproduced, distributed, performed, displayed, trained upon by AI systems, reverse-engineered, or used to create derivative works without the author’s explicit written consent.
|
| 67 |
+
|
| 68 |
+
Verification: Each record is timestamped through the Zenodo/DataCite registry and may reference the master DOI: https://doi.org/10.5281/zenodo.17460107 as the consolidated legal and authorship archive.
|
| 69 |
+
|
| 70 |
+
© 2025 Liam Grinstead — All Rights Reserved.
|
| 71 |
+
Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
|