claude-bot
598f7fd662
Replaces the previous tree entirely. The pipeline is four separable pieces, each
testable on its own:
1. leanbench/llm.py provider-neutral OpenAI-compatible connector, stdlib only,
with retries, JSON coercion and a scripted client so the
rest of the code is testable without a network.
2. leanbench/detect.py per-page detection of formalizable mathematics. Case 1 is
mathematics asserted as fact (definition/theorem/lemma/...,
deliberately not sub-classified); case 2 is a problem left
open to the reader. Hits come back as spans -- start page
and index, end page and index -- so several per page is
normal. A hit cut off by the page end makes the detector
re-read that page together with the next, repeatedly, so an
item spread over three pages is found whole.
Offsets come from verbatim quotes resolved by
leanbench/anchors.py (exact -> whitespace-insensitive ->
fuzzy), because models quote well and count badly.
3. leanbench/formalize.py + leanbench/verify.py
each span becomes one Lean 4 declaration, judged by the real
compiler: full elaboration for case 1, statement legality
only for case 2, which has no proof to check. Lean's own
diagnostics feed back for a repair attempt. "verified" (Lean
accepted it) and "proved" (accepted with no sorry) are kept
strictly apart.
4. leanbench/db.py indexed SQLite plus JSONL export. Every item is indexed by
the types of its inputs, or by its own type when it is not a
proof, and records the source book (sha256) and the page it
starts on -- so `drop --source H --pages A-B` followed by a
re-ingest redoes exactly those pages and nothing else.
Infrastructure:
- lean/ Lake project pinned to Mathlib.
- docker/ image carrying Lean 4 and a prebuilt Mathlib. The Mathlib
revision and the toolchain are resolved against each other at
build time rather than both hardcoded, so the build does not rot.
One Docker layer per Lean step, each long one under `timeout`:
the gigabyte cache fetch earns a layer that a later failure never
invalidates, and a stalled step fails with a message instead of
sitting silent for an hour. There is deliberately no `lake build`
of our own library -- verification runs `lake env lean` against
Mathlib's oleans and needs nothing else, and building the library
was pure cost. docker/smoke.py runs the whole pipeline inside the
image against the real compiler, with the LLM scripted.
- .gitea/workflows ci (lint + tests in a venv, since the runner image's Python is
Debian-managed and refuses installs into itself), docker (build,
verify, optionally push), ingest (submit a book by hand). The
docker build runs with --progress=plain, because BuildKit's
default renderer collapses a long step into one rewritten line
and a stall then looks exactly like progress.
The docker workflow is not a throwaway build: it leaves the image, the layer cache,
an idle `lean-bench` container and the `lean-bench-data` / `lean-bench-books`
volumes in place on the runner, and never prunes, so the environment is there to be
used afterwards. Pushing to the registry is optional and cannot fail the build --
Gitea's container registry does not accept the built-in Actions token, so that needs
a REGISTRY_TOKEN secret with write:package.
58 tests; they need neither network nor Lean.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ci5T3gWDhhmqDec3e1XtYZ
134 lines
4.8 KiB
Python
134 lines
4.8 KiB
Python
from __future__ import annotations
|
|
|
|
from leanbench.detect import KIND_PROBLEM, KIND_STATEMENT, detect_document
|
|
from leanbench.document import Document
|
|
from leanbench.llm import ScriptedClient
|
|
|
|
from .conftest import detection_reply
|
|
|
|
|
|
def test_single_page_detection_reports_page_and_index(two_page_book: Document) -> None:
|
|
client = ScriptedClient(
|
|
[
|
|
detection_reply(
|
|
[
|
|
{
|
|
"kind": "statement",
|
|
"label": "Definition 1.6.1",
|
|
"summary": "divisibility",
|
|
"start_quote": "Definition 1.6.1. Let a and b be integers.",
|
|
"end_quote": "b = a k.",
|
|
"continues_past_end": False,
|
|
}
|
|
]
|
|
),
|
|
detection_reply([]),
|
|
]
|
|
)
|
|
report = detect_document(client, two_page_book, max_window_pages=3)
|
|
assert len(report.detections) == 1
|
|
found = report.detections[0]
|
|
assert found.kind == KIND_STATEMENT
|
|
assert found.start_page == 1 and found.end_page == 1
|
|
page = two_page_book.page(1).text
|
|
assert page[found.start_index : found.end_index].startswith("Definition 1.6.1")
|
|
assert page[found.start_index : found.end_index].endswith("b = a k.")
|
|
|
|
|
|
def test_item_running_off_the_page_triggers_a_second_page(two_page_book: Document) -> None:
|
|
truncated = detection_reply(
|
|
[
|
|
{
|
|
"kind": "statement",
|
|
"label": "Theorem 1.6.3",
|
|
"summary": "gcd exists",
|
|
"start_quote": "Theorem 1.6.3. Let a and b be nonzero integers.",
|
|
"end_quote": "the set of common",
|
|
"continues_past_end": True,
|
|
}
|
|
]
|
|
)
|
|
whole = detection_reply(
|
|
[
|
|
{
|
|
"kind": "statement",
|
|
"label": "Theorem 1.6.3",
|
|
"summary": "gcd exists",
|
|
"start_quote": "Theorem 1.6.3. Let a and b be nonzero integers.",
|
|
"end_quote": "the greatest common divisor.",
|
|
"continues_past_end": False,
|
|
},
|
|
{
|
|
"kind": "problem",
|
|
"label": "Exercise 1.6.5",
|
|
"summary": "odd + odd is even",
|
|
"start_quote": "Exercise 1.6.5. Prove that the sum of two odd",
|
|
"end_quote": "odd integers is even.",
|
|
"continues_past_end": False,
|
|
},
|
|
]
|
|
)
|
|
client = ScriptedClient([truncated, whole])
|
|
report = detect_document(client, two_page_book, max_window_pages=3)
|
|
|
|
assert len(client.calls) == 2, "the page should be re-read with its successor"
|
|
theorem = report.detections[0]
|
|
assert theorem.start_page == 1
|
|
assert theorem.end_page == 2, "the span must cross the page boundary"
|
|
assert theorem.spans_pages == 2
|
|
assert not theorem.truncated
|
|
assert two_page_book.page(2).text[: theorem.end_index].endswith("greatest common divisor.")
|
|
|
|
exercise = report.detections[1]
|
|
assert exercise.kind == KIND_PROBLEM
|
|
assert exercise.start_page == 2
|
|
|
|
|
|
def test_window_stops_growing_at_the_limit_and_says_so(two_page_book: Document) -> None:
|
|
always_truncated = detection_reply(
|
|
[
|
|
{
|
|
"kind": "statement",
|
|
"label": "",
|
|
"summary": "",
|
|
"start_quote": "Definition 1.6.1. Let a and b be integers.",
|
|
"end_quote": "b = a k.",
|
|
"continues_past_end": True,
|
|
}
|
|
]
|
|
)
|
|
client = ScriptedClient([always_truncated] * 4)
|
|
report = detect_document(client, two_page_book, max_window_pages=2)
|
|
assert len(client.calls) == 2
|
|
assert report.detections[0].truncated is True
|
|
|
|
|
|
def test_unresolvable_quotes_are_recorded_not_invented(two_page_book: Document) -> None:
|
|
client = ScriptedClient(
|
|
[
|
|
detection_reply(
|
|
[
|
|
{
|
|
"kind": "statement",
|
|
"label": "Hallucination 9",
|
|
"summary": "not in the book",
|
|
"start_quote": "Theorem 9.9.9 about topological quantum widgets",
|
|
"end_quote": "which is absurd.",
|
|
"continues_past_end": False,
|
|
}
|
|
]
|
|
),
|
|
detection_reply([]),
|
|
]
|
|
)
|
|
report = detect_document(client, two_page_book)
|
|
assert report.detections == []
|
|
assert len(report.unresolved) == 1
|
|
|
|
|
|
def test_a_failing_page_does_not_kill_the_run(two_page_book: Document) -> None:
|
|
client = ScriptedClient(["not json at all", detection_reply([])])
|
|
report = detect_document(client, two_page_book)
|
|
assert report.pages_scanned == 2
|
|
assert any("detector failed" in u.get("reason", "") for u in report.unresolved)
|