Repository navigation
test: bind every toolchain contract to the shared acceptance vectors - #2000
Merged
Merged
Conversation
State owns schema/toolchain-vectors-v1.json; this copy is what the offline Python tests and the Worker's vitest suite bind to, and CI diffs it against protected State main. The completion builder, the replay orchestrator, the live replay schemas, the Worker's completion decoder and the submission view decoder now all accept and reject the same toolchain strings. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
lean-eval-state is private, so neither this repository's CI nor lean-eval's can fetch its copy anonymously. The public copy here is canonical; State mirrors it with a drift check in its own CI. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Schema patterns are evaluated as ECMA-262 regular expressions, where \$ is the end of input, so the trailing-newline vector pins the difference from Python's lenient \$. The Worker's State event validator, the historical replay schemas that state the general grammar, and the historical scripts' own patterns are bound too; only the fixed-toolchain public replay smoke stays out. Edits to the vectors now also trigger the Worker's check and deploy workflow. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
scripts/historical_replay_controller.py loads every JSON file in schemas/ as a JSON Schema resource, and the vectors are data, not a schema. They live in contracts/ instead. Co-Authored-By: Claude Fable 5.1 <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.
This PR adds
contracts/toolchain-vectors-v1.json, the canonical list of benchmark toolchain strings every lean-eval contract must accept or reject identically, and tests binding the completion builder, the replay orchestrator and the other scripts with their own toolchain pattern, every schema that states the general toolchain grammar, and the Worker's evaluation-completion decoder, State event validator and submission-view decoder to it (schema patterns are evaluated as ECMA-262 regular expressions, so the trailing-newline vector pins the difference from Python's lenient$). This repository is public, so it holds the canonical copy; https://github.com/leanprover/lean-eval-state (private) will mirror it in a reviewed State contract bump, and leanprover/lean-eval#661 checks the benchmark's pinned toolchain against it, so a pin that any contract rejects fails in lean-eval's own CI. Edits to the vectors also trigger the Worker's check and deploy workflow.🤖 Prepared with Claude Code