tlg
36eac66abc
Root cause chain (evidence: 25+ failed traces on nproj/3022414): straight quote in page text ends the guided-decoding JSON string early -> grammar deadlock -> whitespace loop until context/token limit -> chat_json fails 3x (~35 min) -> IMAP conn idles out -> move_to_trash raises -> mail never trashed -> reprocessed forever. - ingest/mailer: MailBox.ensure_alive() reconnect before finalize; move_to_trash failure falls back to flag_checked, run continues - stages: page_text() feeds only div.content minus skill-tag cloud to the LLM and maps quote chars to apostrophes (the actual degeneration trigger, live-verified: extract now OK in 21 s) - llm: requirements maxItems 40; max_tokens 8192 caps degenerate runs at ~100 s instead of ~10 min Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
75 lines
2.8 KiB
Python
75 lines
2.8 KiB
Python
import json
|
|
from unittest import mock
|
|
|
|
import pytest
|
|
|
|
from projektmatch import llm
|
|
|
|
|
|
def fake_post(payloads):
|
|
"""Return a mock for requests.post yielding chat completions."""
|
|
responses = []
|
|
for p in payloads:
|
|
r = mock.Mock()
|
|
r.status_code = 200
|
|
r.json.return_value = {"choices": [{"message": {"content": p}}]}
|
|
responses.append(r)
|
|
return mock.Mock(side_effect=responses)
|
|
|
|
|
|
def test_chat_json_strips_think_and_parses():
|
|
content = "<think>lange Kette</think>{\"a\": 1}"
|
|
with mock.patch("projektmatch.llm.requests.post", fake_post([content])) as p:
|
|
out = llm.chat_json("http://v/v1", "m", [{"role": "user", "content": "x"}],
|
|
{"type": "object"})
|
|
assert out == {"a": 1}
|
|
body = p.call_args.kwargs["json"]
|
|
assert body["max_tokens"] == 8192
|
|
assert body["chat_template_kwargs"] == {"enable_thinking": False}
|
|
assert body["response_format"]["json_schema"]["schema"] == {"type": "object"}
|
|
|
|
|
|
def test_extract_schema_caps_requirements_list():
|
|
reqs = llm.EXTRACT_SCHEMA["properties"]["requirements"]
|
|
assert reqs["maxItems"] == 40
|
|
|
|
|
|
def test_chat_json_strips_markdown_fences():
|
|
content = "```json\n{\"a\": 2}\n```"
|
|
with mock.patch("projektmatch.llm.requests.post", fake_post([content])):
|
|
out = llm.chat_json("http://v/v1", "m", [], {"type": "object"})
|
|
assert out == {"a": 2}
|
|
|
|
|
|
def test_chat_json_retries_on_empty_content():
|
|
good = json.dumps({"a": 1})
|
|
with mock.patch("projektmatch.llm.time.sleep", lambda s: None), \
|
|
mock.patch("projektmatch.llm.requests.post", fake_post(["", good])):
|
|
assert llm.chat_json("http://v/v1", "m", [], {"type": "object"}) == {"a": 1}
|
|
|
|
|
|
def test_chat_json_raises_after_exhausted_retries():
|
|
with mock.patch("projektmatch.llm.time.sleep", lambda s: None), \
|
|
mock.patch("projektmatch.llm.requests.post", fake_post(["", "", ""])):
|
|
with pytest.raises(llm.LlmError, match="kein JSON"):
|
|
llm.chat_json("http://v/v1", "m", [], {"type": "object"})
|
|
|
|
|
|
def test_match_cv_retries_on_missing_nr_then_raises():
|
|
reqs = [{"nr": 1, "text": "Python"}, {"nr": 2, "text": "K8s"}]
|
|
partial = json.dumps({"ratings": [
|
|
{"nr": 1, "rating": "yes", "reason": "ok"}]})
|
|
with mock.patch("projektmatch.llm.requests.post",
|
|
fake_post([partial, partial])):
|
|
with pytest.raises(llm.LlmError, match="unvollständig"):
|
|
llm.match_cv("http://v/v1", "m", "CV", reqs)
|
|
|
|
|
|
def test_match_cv_ok():
|
|
reqs = [{"nr": 1, "text": "Python"}]
|
|
full = json.dumps({"ratings": [
|
|
{"nr": 1, "rating": "unknown", "reason": "Teilevidenz"}]})
|
|
with mock.patch("projektmatch.llm.requests.post", fake_post([full])):
|
|
out = llm.match_cv("http://v/v1", "m", "CV", reqs)
|
|
assert out[0]["rating"] == "unknown"
|