Handle cases where the LLM produces a valid JSON array

src/neo4j_graphrag/experimental/components/entity_relation_extractor.py

@@ -225,7 +225,7 @@ async def extract_for_chunk(
                 logger.debug(f"Invalid JSON: {llm_result.content}")
             result = {"nodes": [], "relationships": []}
         try:
-            chunk_graph = Neo4jGraph(**result)
+            chunk_graph = Neo4jGraph.model_validate(result)
         except ValidationError as e:
             if self.on_error == OnError.RAISE:
                 raise LLMGenerationError("LLM response has improper format") from e

tests/unit/experimental/components/test_entity_relation_extractor.py

@@ -178,6 +178,24 @@ async def test_extractor_llm_invalid_json() -> None:
         await extractor.run(chunks=chunks)
 
 
+@pytest.mark.asyncio
+async def test_extractor_llm_invalid_json_is_a_list() -> None:
+    """Test what happens when the returned JSON is a valid JSON list,
+     but it does not match the expected Pydantic model"""
+    llm = MagicMock(spec=LLMInterface)
+    llm.ainvoke.return_value = LLMResponse(
+        # missing "label" for entity
+        content='[{"nodes": [{"id": 0, "entity_type": "Person", "properties": {}}], "relationships": []}]'
+    )
+
+    extractor = LLMEntityRelationExtractor(
+        llm=llm,
+    )
+    chunks = TextChunks(chunks=[TextChunk(text="some text", index=0)])
+    with pytest.raises(LLMGenerationError):
+        await extractor.run(chunks=chunks)
+
+
 @pytest.mark.asyncio
 async def test_extractor_llm_badly_formatted_json_gets_fixed() -> None:
     llm = MagicMock(spec=LLMInterface)