examples: 5 hand-curated fixtures + parametrized regression suite

tests/test_examples_fixtures.py

@@ -0,0 +1,52 @@
+"""Regression suite for LLM extraction — drives the extractor against
+hand-curated fixtures and asserts the parsed JSON matches expectations.
+
+Each fixture is `{post: RawPost, expected: dict}`. The test does NOT
+hit a live LLM — it mocks the response to return the *expected* JSON,
+exercising the parser, validator, and currency-handling paths."""
+from __future__ import annotations
+
+import json
+from pathlib import Path
+
+import httpx
+import pytest
+import respx
+
+from fire_planner.examples.llm_extract import extract_with_qwen
+from fire_planner.examples.models import RawPost
+
+LLAMA_URL = "http://llama-cpp.llama-cpp.svc.cluster.local:8000/v1/chat/completions"
+
+FIXTURE_DIR = Path(__file__).parent / "fixtures" / "reddit"
+
+
+def _fixtures() -> list[Path]:
+    return sorted(FIXTURE_DIR.glob("example_*.json"))
+
+
+@respx.mock
+@pytest.mark.asyncio
+@pytest.mark.parametrize("fixture_path", _fixtures(), ids=lambda p: p.stem)
+async def test_extractor_matches_fixture(fixture_path: Path) -> None:
+    data = json.loads(fixture_path.read_text())
+    post = RawPost.model_validate(data["post"])
+    expected = data["expected"]
+    expected_with_conf = {**expected, "confidence": 0.9}
+
+    respx.post(LLAMA_URL).respond(
+        200,
+        json={"choices": [{"message": {"content": json.dumps(expected_with_conf)}}]},
+    )
+
+    async with httpx.AsyncClient() as client:
+        out = await extract_with_qwen(post, llama_url=LLAMA_URL, client=client)
+
+    assert out is not None
+    for k, v in expected.items():
+        actual = getattr(out, k)
+        if hasattr(actual, "__float__"):
+            assert float(actual) == float(v), f"{fixture_path.stem}: {k}"
+        else:
+            # Pydantic StrEnum compares equal to its string value
+            assert actual == v or str(actual) == v, f"{fixture_path.stem}: {k}"