# NovaFabric — Replay & Evidence Infrastructure for AI Agents > NovaFabric is open-source, self-hosted replay and evidence infrastructure for AI agents and agentic systems. It captures agent executions as portable Run Capsules for replay, behavioral/structural diff, lineage, cryptographic provenance, assurance, and audit. Source: · Author: Mohsen Seyedkazemi Ardebili · This is the Markdown twin of the HTML page; the HTML is canonical. --- open source · Apache-2.0 · self-hosted Replayable execution and evidence infrastructure. Open-source, self-hosted replay and evidence infrastructure for **AI agents and agentic systems**. It captures agent executions as **portable Run Capsules** with no application-code changes, then adds replay, behavioral/structural diff, lineage, cryptographic provenance, assurance and audit. The same capture layer can wrap general commands and supports local, Docker, Kubernetes and SLURM/HPC environments. Observability tells you _what happened_; NovaFabric focuses on whether a past execution can be _replayed, compared, and proven_. - INSTALL pip install novafabric - RUNS laptop → Docker → K8s → SLURM - OFFLINE no cloud · no account // the problem ## When an AI run breaks, every layer is suspect. A modern AI system isn't a single model — it's a drifting graph of prompts, models, tools (HTTP APIs, MCP servers, shell commands), memory/RAG context, datasets, environments, and execution targets (laptop, container, K8s pod, SLURM job). Each layer drifts at its own rate, and when something misbehaves you can't tell which layer moved. The ecosystem has excellent telemetry for _what happened_ — and almost nothing for _what would happen if I ran this again, under the same conditions, today._ Reproducibility isn't a nice-to-have: it's the precondition for debugging, regression detection, governance, and trust. // what it does ## Capture → Seal → Replay → Diff → Audit. ### Asset Registry The identity layer — `name@version` for models, agents, prompts, tools, datasets, evals, and deployments. ### Run Capsule A portable per-run snapshot — inputs, outputs, model calls, tool calls, a full environment lock, and the trace — written on success _and_ failure. A folder you own, not a row in someone's cloud. ### Replay Engine Reconstruct a past run in four honest modes — `exact`, `mocked`, `semantic`, and `forensic` — each making a specific, falsifiable promise. ### Lineage Graph Causation — which assets and prior runs contributed to this run, and what depends on it. ### Evidence Bundle A signed, audit-ready export — capsule + lineage + attestations + redaction proof — that a third party can verify _without NovaFabric installed_. ### Zero-code-change capture `nova capture python my_agent.py` — per-SDK hooks, wire-level hooks (httpx/requests/aiohttp/urllib3), and transparent HTTP/MCP proxies record any command. No application changes; all patches removed after the run. // the honest contract ## Replay modes with explicit, falsifiable semantics. ### exact Deterministic re-execution against the same local model, container, data, and environment — bit-identical where deterministic. ### mocked Re-run the command with every model/tool call served from the capsule cache — no model called, no tool invoked, fully deterministic. ### semantic Re-execute and compare _meaning_ (via a judge model), not tokens — for remote closed-weight LLMs whose weights and routing change without notice. ### forensic Inspect the captured run without re-executing anything — read-only, no network, no side effects. The taxonomy is deliberately honest: NovaFabric does **not** claim deterministic exact replay of remote closed-weight models. Replay modes make the promise explicit rather than pretending a changing provider is deterministic. // trust & governance ## Sealed so it can't be silently altered. ### NovaSeal — three independent proofs Every sealed capsule binds a DSSE signature (ECDSA P-256 / Ed25519), an RFC 3161 trusted timestamp (eIDAS-qualified path for the EU), and an append-only Merkle log with inclusion proofs. ### Pluggable signing backends Local keys, AWS KMS, Azure Key Vault, GCP KMS, or keyless Sigstore (Fulcio + Rekor). `nova verify` re-checks signature, timestamp, and log integrity. ### Policy & maker-checker An OPA/Rego policy engine (allow / deny / require-human / require-batch) and dual-approval with cryptographic separation of duties — proposer ≠ approver, enforced by key fingerprint. ### Compliance exporters WORM retention (S3 Object Lock, Azure immutable blob, GCS Bucket Lock) with legal holds, plus exporters for EU AI Act Annex IV, NIST AI RMF, GDPR RoPA, and CycloneDX AI-BOM. // standards-based, not reinvented ## Adopts open standards; invents only two formats. Only the Run Capsule and Evidence Bundle are new on-disk formats — everything else adopts an existing open standard, so NovaFabric feeds your observability stack rather than replacing it. **Design slogan:** local-first now, distributed-ready always, cluster-scale later — a one-node run is the smallest case of a distributed run. Built in Python 3.12+ with a Go collector tier. NovaFabric is **beta / pre-v1**: local capture, replay, diff and evidence workflows are the strongest path, while server and cluster-scale components remain less mature and on-disk formats are not frozen until v1.0. // more ## Explore the rest of the lab.