claude-bot
caedf2e6f4
- extract_section.py: slice a book section out of a pdftotext -layout dump and list its labeled Theorem/Proposition/Lemma/Corollary/Definition items. - formalize.py: LLM-formalize a section into Lean 4 candidate statements (preferring Mathlib vocabulary) plus a definitions manifest. - check_statements.py: elaboration-only sanity check against a real Mathlib-backed Lean project (must run outside the Cowork sandbox, which has no Lean toolchain and can't run macOS binaries anyway). - build_library.py: assemble solved lean-bench results into a reusable .lean file, reusing each winning proof verbatim. - llm.py: standalone OpenAI-compatible chat client + .env loader, kept independent of lean_agent so this pipeline works without it installed. - Source PDFs and raw/layout text dumps are gitignored (copyrighted material); only code and the final short Lean statement datasets are tracked.
62 lines
2.5 KiB
Python
62 lines
2.5 KiB
Python
import json
|
|
import tempfile
|
|
import unittest
|
|
from pathlib import Path
|
|
|
|
from textbook.build_library import build
|
|
|
|
|
|
def write_jsonl(path, records):
|
|
path.write_text('\n'.join(json.dumps(r) for r in records) + '\n', encoding='utf-8')
|
|
|
|
|
|
class BuildLibraryTests(unittest.TestCase):
|
|
def test_build_includes_only_solved_theorems_and_hoists_shared_imports(self):
|
|
with tempfile.TemporaryDirectory() as directory:
|
|
directory = Path(directory)
|
|
dataset = directory / 'dataset.jsonl'
|
|
results = directory / 'results.jsonl'
|
|
write_jsonl(dataset, [
|
|
{"id": "solved-one", "statement": "True", "prelude": "import Mathlib",
|
|
"source_label": "Theorem 1", "natural_language": "trivial"},
|
|
{"id": "unsolved-one", "statement": "False", "prelude": "import Mathlib",
|
|
"source_label": "Theorem 2", "natural_language": "not actually true"},
|
|
])
|
|
write_jsonl(results, [
|
|
{"id": "solved-one", "success": True,
|
|
"attempts": [{"proof": "trivial", "success": True, "diagnostics": ""}]},
|
|
{"id": "unsolved-one", "success": False, "attempts": []},
|
|
])
|
|
|
|
text, solved, skipped = build(dataset, results)
|
|
|
|
self.assertEqual(solved, ['solved-one'])
|
|
self.assertEqual(skipped, ['unsolved-one'])
|
|
self.assertIn('import Mathlib', text)
|
|
self.assertEqual(text.count('import Mathlib'), 1) # hoisted once, not per-theorem
|
|
self.assertIn('theorem solved_one : True := by', text)
|
|
self.assertNotIn('unsolved_one', text)
|
|
|
|
def test_build_keeps_non_import_prelude_attached_per_theorem(self):
|
|
with tempfile.TemporaryDirectory() as directory:
|
|
directory = Path(directory)
|
|
dataset = directory / 'dataset.jsonl'
|
|
results = directory / 'results.jsonl'
|
|
write_jsonl(dataset, [
|
|
{"id": "custom-def", "statement": "myPred 1", "prelude": "def myPred (n : Nat) := True",
|
|
"source_label": "Theorem 3", "natural_language": ""},
|
|
])
|
|
write_jsonl(results, [
|
|
{"id": "custom-def", "success": True,
|
|
"attempts": [{"proof": "trivial", "success": True, "diagnostics": ""}]},
|
|
])
|
|
|
|
text, solved, skipped = build(dataset, results)
|
|
|
|
self.assertIn('def myPred (n : Nat) := True', text)
|
|
self.assertEqual(solved, ['custom-def'])
|
|
|
|
|
|
if __name__ == '__main__':
|
|
unittest.main()
|