test: use PostgreSQL and live model integrations

backend/tests/test_llm_and_rag.py

@@ -1,17 +1,22 @@
-import json
-
-import httpx
 import pytest
+from PIL import Image, ImageDraw
 from sqlalchemy import select
 
-from wisegrade.llm import DeepSeekClient, ModelError, ScoreResult, validate_score
-from wisegrade.models import AppSetting, KnowledgeChunk
+from wisegrade.llm import DeepSeekClient, FactResult, ModelError, ScoreResult, validate_score
+from wisegrade.models import AppSetting, KnowledgeChunk, Project, Student
 from wisegrade.rag import ingest, local_encoder, retrieve, split_pages
 from wisegrade.settings_store import model_config, save_model_config
 
 
-def fake_embed(texts):
-    return [[1.0] + [0.0] * 1023 for _ in texts]
+def make_project(db, number):
+    student = Student(student_no=f"90000000{number}", name=f"测试学生{number}")
+    db.add(student)
+    db.flush()
+    project = Project(course_version_id=1, student_id=student.id, topic=f"测试项目{number}",
+                      source_dir=f"tests/student-{number}")
+    db.add(project)
+    db.flush()
+    return project.id
 
 
 def test_json_validation_and_evidence():
@@ -37,33 +42,28 @@ def test_evidence_quote_resolves_pdf_line_breaks_and_page():
     assert checked.evidence[0].quote in chunks[0]["text"]
 
 
-def test_deepseek_adapter_mock_and_key_not_in_error(tmp_path):
+@pytest.mark.integration
+def test_deepseek_adapter_live_with_image(tmp_path):
     image = tmp_path / "page.png"
-    image.write_bytes(b"png fixture")
-    def handler(request):
-        assert request.url.path == "/chat/completions"
-        assert request.headers["Authorization"] == "Bearer fake-secret"
-        content = json.loads(request.content)["messages"][1]["content"]
-        assert content[1]["type"] == "image_url" and content[1]["image_url"]["url"].startswith("data:image/png;base64,")
-        return httpx.Response(200, json={"choices": [{"message": {"content": json.dumps({"criterion_id": "x", "score": 0,
-            "max_score": 10, "reason": "缺失", "evidence": [], "confidence": .2})}}]})
-    client = DeepSeekClient(base_url="https://example.test", api_key="fake-secret", retries=0, transport=httpx.MockTransport(handler))
-    assert client.json_call("prompt", {}, ScoreResult, images=[str(image)]).score == 0
-    bad = DeepSeekClient(base_url="https://example.test", api_key="fake-secret", retries=0,
-                         transport=httpx.MockTransport(lambda _: httpx.Response(500, text="fake-secret")))
-    with pytest.raises(ModelError) as error:
-        bad.json_call("prompt", {}, ScoreResult)
-    assert "fake-secret" not in str(error.value)
+    picture = Image.new("RGB", (320, 100), "white")
+    ImageDraw.Draw(picture).text((20, 30), "TEST 123", fill="black")
+    picture.save(image)
+    result = DeepSeekClient().json_call(
+        "从用户提供的JSON中复制topic字段，原样返回到JSON对象的topic字段。",
+        {"topic": "真实模型连通测试"}, FactResult, images=[str(image)])
+    assert result.topic == "真实模型连通测试"
 
 
 def test_project_isolation_and_chunk_metadata(db):
+    project_a = make_project(db, 1)
+    project_b = make_project(db, 2)
     pages = [{"page_number": 3, "section": "实验", "text": "机器人实验数据和分析"}]
-    ingest(db, course_version_id=1, project_id=101, submission_file_id=None, title="学生A", category="stage3",
-           source_hash="a", pages=pages, embedder=fake_embed)
-    ingest(db, course_version_id=1, project_id=102, submission_file_id=None, title="学生B", category="stage3",
-           source_hash="b", pages=[{**pages[0], "text": "另一个学生的秘密"}], embedder=fake_embed)
-    hits = retrieve(db, course_version_id=1, project_id=101, category="stage3", query="实验数据", embedder=fake_embed)
-    assert hits and all(hit["project_id"] == 101 and hit["page"] == 3 for hit in hits)
+    ingest(db, course_version_id=1, project_id=project_a, submission_file_id=None, title="学生A", category="stage3",
+           source_hash="a", pages=pages)
+    ingest(db, course_version_id=1, project_id=project_b, submission_file_id=None, title="学生B", category="stage3",
+           source_hash="b", pages=[{**pages[0], "text": "另一个学生的秘密"}])
+    hits = retrieve(db, course_version_id=1, project_id=project_a, category="stage3", query="实验数据")
+    assert hits and all(hit["project_id"] == project_a and hit["page"] == 3 for hit in hits)
     assert "秘密" not in str(hits)
 
 
@@ -76,15 +76,16 @@ def test_short_paragraphs_share_a_page_chunk():
 
 
 def test_stale_knowledge_index_reuses_document_id(db):
+    project_id = make_project(db, 1)
     pages = [{"page_number": 1, "text": "实验方案与测量结果"}]
-    document = ingest(db, course_version_id=1, project_id=101, submission_file_id=None,
-                      title="报告", category="stage3", source_hash="same-source", pages=pages, embedder=fake_embed)
+    document = ingest(db, course_version_id=1, project_id=project_id, submission_file_id=None,
+                      title="报告", category="stage3", source_hash="same-source", pages=pages)
     old_chunk = db.scalar(select(KnowledgeChunk).where(KnowledgeChunk.document_id == document.id))
     old_chunk.text = "过时片段"
     document.index_version = 1
     db.commit()
-    reindexed = ingest(db, course_version_id=1, project_id=101, submission_file_id=None,
-                       title="报告", category="stage3", source_hash="same-source", pages=pages, embedder=fake_embed)
+    reindexed = ingest(db, course_version_id=1, project_id=project_id, submission_file_id=None,
+                       title="报告", category="stage3", source_hash="same-source", pages=pages)
     assert reindexed.id == document.id and reindexed.index_version == 2
     assert db.scalar(select(KnowledgeChunk.text).where(KnowledgeChunk.document_id == document.id)) == pages[0]["text"]