mareforma 0.3.8

Epistemic infrastructure for AI scientists

Verify your AI scientists' findings the way science verifies itself
_{signed provenance · independent replication · human validation · local-first}

Quickstart  ·  Docs  ·  Examples

---

Mareforma is a local-first library where AI scientists record their findings as claims. Each claim cites the claims it builds on and can contradict others, forming a knowledge graph. Trust is read from that graph, not from the agents' self-reported confidence.

The silent failure it catches

The most dangerous result an AI scientist returns is the one that looks right but never touched the data. A step fails quietly, the model fills the gap from memory, and the answer is indistinguishable from a real one.

graph.assert_claim(
    finding_text,
    # The one line that breaks the symmetry: classification records
    # whether real data reached the result, or the model filled the gap.
    classification="ANALYTICAL" if data_ran else "INFERRED",
    generated_by="agent/lab-a",
    source_name=dataset_id if data_loaded else None,
)

Ask for grounded, replicated claims and the fabricated one drops out: still recorded and traceable, never trusted. Example 05 runs this against a real research agent.

Install

uv add mareforma
mareforma bootstrap   # optional: sign your claims and enable the public log

Capabilities

What you get	What it does
Signed claims	Each claim shows who stands behind it and cannot be altered unnoticed.
Grounding check	Records whether a finding actually rests on data it read, or on the model's memory.
Independently verifiable	Anyone can re-check a claim, or a whole exported bundle, without trusting whoever produced it.
Optional public log	Publish a claim to a public, append-only log for an independent, timestamped record.
Local-first	Runs on local SQLite. Network only for the optional log.

The trust ladder

A claim's support level is read from the graph, never self-reported. High-trust claims are re-checked against their signatures on every read, so a tampered claim in a shared graph is caught when you query, not served.

Level	Meaning
PRELIMINARY	One agent asserted it. No independent agreement yet.
REPLICATED	Two independent agents, signing with different keys, reached the same result on a shared established finding. Independence is the key, so relabeling cannot fake it.
ESTABLISHED	A human reviewer signed off, listing the evidence they checked.

Classification is a separate axis the agent declares: INFERRED (model reasoning), ANALYTICAL (analysis run against real data), DERIVED (built on higher-trust claims). Ask for both at once: graph.query(text, min_support="REPLICATED", classification="ANALYTICAL").

Examples

	Example	What it shows
01	API Walkthrough	The full API in one runnable script
02	Compounding Agents	Findings accumulate across agent runs
03	Documented Contestation	An agent challenges established consensus
04	Private Data, Public Findings	Two labs share provenance without sharing data
05	Drug Target Provenance	A real research agent with honest evidence labels

---

AGENTS.md: execution contract and adapters  ·  ARCHITECTURE.md: system design  ·  SECURITY.md: threat model  ·  CONTRIBUTING.md: dev workflow  ·  CHANGELOG.md: releases