Add --mareforma-export post-run flag for signed claim receipts#1
Open
felipeyanez wants to merge 1 commit into
Open
Add --mareforma-export post-run flag for signed claim receipts#1felipeyanez wants to merge 1 commit into
felipeyanez wants to merge 1 commit into
Conversation
Adds a single --mareforma-export [RUN_DIR] flag to launch.py. When set, walks RUN_DIR for autoscientists_submission/research_insights.md files and signs each one (Ed25519, content-addressable) into a local mareforma graph at RUN_DIR/.mareforma/graph.db. Content-keyed idempotency: re-runs on unchanged files are no-ops; editing a file produces a new claim and preserves the old one for audit. The flag is opt-in, the mareforma import is lazy, and the normal bootstrap path is untouched when the flag is absent. Requires: pip install mareforma[clawinstitute]>=0.3.3 Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Hi. I read your AutoScientists paper and noticed how every
research_insights.mdends with some version of the "single-run, no independent replication" caveat. I've been working on an epistemic substrate (mareforma) where agents write signed, cross-referenced claims, and I think it can help here.This PR is intentionally small: a single
--mareforma-export [RUN_DIR]post-run mode added tolaunch.py. When set, it walksRUN_DIRforautoscientists_submission/research_insights.mdfiles and signs each one (Ed25519, content-addressable) into a local graph atRUN_DIR/.mareforma/graph.db. When a second lab runs the same task and signs with their own enrolled key, the substrate flagsREPLICATEDwhen both converge on the same finding. Today "independent validation" is a phrase in your paper; with this in place it becomes a query:graph.query(text, min_support="REPLICATED").Tested against all 24 reference submissions in
task-biomlbench/: 24 of 24 signed on first run, 0 of 24 re-signed on the second pass (content-keyed idempotency, so editing a file produces a new claim and preserves the old one for audit). Round-trips viamareforma askreturn findings withPRELIMINARYtrust level and per-task source attribution.43 lines added to
launch.py. Single file changed. Opt-in flag, lazy import, no required deps. The normal bootstrap path is untouched.The full integration also signs
autoscientists.pyandautoscientists_submission.csvfor tamper-evidence across the whole submission directory, runs as a live event hook during the experiment (instead of post-run only), and supports validator enrollment for theREPLICATED -> ESTABLISHEDpromotion. Happy to send a follow-up if this direction is useful. Otherwise this stands alone.Cheers.