Update README.md
Browse files
README.md
CHANGED
|
@@ -10,62 +10,56 @@ 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/
|
| 14 |
---
|
|
|
|
| 15 |
# RFTSystems β Agent Forensics Suite (START HERE)
|
| 16 |
|
| 17 |
-
|
| 18 |
-
Audit, prove, replay, and diff
|
| 19 |
|
| 20 |
## Why this exists
|
| 21 |
-
AI is being
|
| 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 |
-
|
| 26 |
|
| 27 |
-
##
|
| 28 |
-
**learn β generate proof β record reality β seal it β
|
| 29 |
|
| 30 |
-
##
|
| 31 |
-
1. Open **
|
| 32 |
-
2.
|
| 33 |
-
3.
|
| 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 |
-
- **
|
| 44 |
-
-
|
| 45 |
-
|
| 46 |
-
- **TimelineDiff Differential Reproducibility** β First divergence: align two run bundles and pinpoint exactly where/why they split.
|
| 47 |
-
- **TrustStack Console** β Audit cockpit: inspect runs, compare state, and trace exactly what changed and why.
|
| 48 |
-
- **Coherent Compute Engine** β Verification-first benchmark: live throughput + stability/energy behaviour + downloadable receipt.
|
| 49 |
-
|
| 50 |
-
## Design principle
|
| 51 |
-
We donβt βimprisonβ agents. We measure drift from declared intent and produce evidence. Enforcement remains an operator decision; this suite is the instrumentation layer.
|
| 52 |
|
| 53 |
-
|
| 54 |
-
|
|
|
|
| 55 |
|
| 56 |
-
|
|
|
|
|
|
|
| 57 |
|
| 58 |
-
-
|
| 59 |
-
|
| 60 |
-
|
| 61 |
-
- Berne Convention β Arts.5(2) & 6bis
|
| 62 |
-
- TRIPS Agreement β Arts.9β14
|
| 63 |
|
| 64 |
-
|
|
|
|
|
|
|
| 65 |
|
| 66 |
-
|
|
|
|
|
|
|
| 67 |
|
| 68 |
-
|
|
|
|
|
|
|
| 69 |
|
| 70 |
-
|
| 71 |
-
|
|
|
|
| 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/
|
| 14 |
---
|
| 15 |
+
|
| 16 |
# RFTSystems β Agent Forensics Suite (START HERE)
|
| 17 |
|
| 18 |
+
Evidence-first instrumentation for AI agents and safety decisions.
|
| 19 |
+
Audit, prove, replay, and diff runs β turning βtrust meβ into verification.
|
| 20 |
|
| 21 |
## Why this exists
|
| 22 |
+
AI is being deployed faster than teams can reproduce failures, attribute responsibility, or prove what changed.
|
| 23 |
+
The unacceptable part isnβt that systems fail β itβs that **nobody can prove what happened**.
|
|
|
|
| 24 |
|
| 25 |
+
This suite exists to make runs **inspectable and independently verifiable**.
|
| 26 |
|
| 27 |
+
## The workflow
|
| 28 |
+
**learn β generate proof β record reality β seal it β replay β diff β audit β benchmark**
|
| 29 |
|
| 30 |
+
## Quick start (60 seconds)
|
| 31 |
+
1. Open **AuditPlane** and generate a baseline suite (signed receipts + Merkle proofs).
|
| 32 |
+
2. Replay the same suite and confirm **drift diffs**.
|
| 33 |
+
3. Export the offline bundle so anyone can verify it without trusting you.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 34 |
|
| 35 |
## The labs
|
| 36 |
+
- **AuditPlane β LLM Decision Proofs**
|
| 37 |
+
Ed25519-signed receipts + hash-chained runs + replay + drift diffs + Merkle proofs
|
| 38 |
+
https://huggingface.co/spaces/RFTSystems/AuditPlane__LLM_Decision_Proofs
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 39 |
|
| 40 |
+
- **ReplayProof β Agent POV Verified Replay**
|
| 41 |
+
Deterministic run bundles you can replay and verify anywhere
|
| 42 |
+
https://huggingface.co/spaces/RFTSystems/ReplayProof__Agent_POV__Verified_Replay
|
| 43 |
|
| 44 |
+
- **Agent Flight Recorder**
|
| 45 |
+
Hash-chained event logging across prompts, tools, outputs, and memory reads/writes
|
| 46 |
+
https://huggingface.co/spaces/RFTSystems/Agent_Flight_Recorder
|
| 47 |
|
| 48 |
+
- **RFT Memory Receipt Engine**
|
| 49 |
+
Generate and verify tamper-evident receipts for agent state/memory
|
| 50 |
+
https://huggingface.co/spaces/RFTSystems/RFT_Memory_Receipt_Engine
|
|
|
|
|
|
|
| 51 |
|
| 52 |
+
- **TimelineDiff β Differential Reproducibility**
|
| 53 |
+
Align two bundles and pinpoint the first divergence
|
| 54 |
+
https://huggingface.co/spaces/RFTSystems/TimelineDiff__Differential_Reproducibility
|
| 55 |
|
| 56 |
+
- **TrustStack Console**
|
| 57 |
+
Audit cockpit: inspect runs, compare state, trace what changed and why
|
| 58 |
+
https://huggingface.co/spaces/RFTSystems/TrustStack_Console
|
| 59 |
|
| 60 |
+
- **Coherent Compute Engine**
|
| 61 |
+
Verification-first benchmark with downloadable receipts
|
| 62 |
+
https://huggingface.co/spaces/RFTSystems/Coherent_Compute_Engine
|
| 63 |
|
| 64 |
+
## Principle
|
| 65 |
+
No receipts β no claims.
|