Refactor GenAI backend and improve testability

Author: marvin-heinrich

.github/workflows/ci-tests.yml

@@ -254,66 +254,12 @@ jobs:
           microservices/document-service/build/reports/tests/test/
           microservices/document-service/build/test-results/test/
           
-  # Job 5: GenAI Service Tests (Kotlin/Spring Boot)
-  genai-service-tests:
-    name: GenAI Service Tests
-    runs-on: ubuntu-latest
-    timeout-minutes: 15
-    
-    services:
-      postgres:
-        image: postgres:15
-        env:
-          POSTGRES_DB: genai_test
-          POSTGRES_USER: test
-          POSTGRES_PASSWORD: test
-        ports:
-          - 5432:5432
-        options: >-
-          --health-cmd "pg_isready -U test -d genai_test"
-          --health-interval 10s
-          --health-timeout 5s
-          --health-retries 5
-          
-    steps:
-    - name: Checkout code
-      uses: actions/checkout@v4
-      
-    - name: Setup Java
-      uses: actions/setup-java@v4
-      with:
-        java-version: ${{ env.JAVA_VERSION }}
-        distribution: 'temurin'
-        cache: 'gradle'
-        
-    - name: Make gradlew executable
-      run: chmod +x microservices/genai-service/gradlew
-      
-    - name: Run tests
-      env:
-        SPRING_PROFILES_ACTIVE: test
-        SPRING_DATASOURCE_URL: jdbc:postgresql://localhost:5432/genai_test
-        SPRING_DATASOURCE_USERNAME: test
-        SPRING_DATASOURCE_PASSWORD: test
-      run: |
-        cd microservices/genai-service
-        ./gradlew test --info --stacktrace
-        
-    - name: Archive test results
-      uses: actions/upload-artifact@v4
-      if: always()
-      with:
-        name: genai-service-test-results
-        path: |
-          microservices/genai-service/build/reports/tests/test/
-          microservices/genai-service/build/test-results/test/
-          
-  # Job 6: Integration Tests
+  # Job 5: Integration Tests
   integration-tests:
     name: Integration Tests
     runs-on: ubuntu-latest
     timeout-minutes: 30
-    needs: [client-tests, genai-tests, auth-service-tests, document-service-tests, genai-service-tests]
+    needs: [client-tests, genai-tests, auth-service-tests, document-service-tests]
 
     services:
       postgres:
@@ -384,9 +330,6 @@ jobs:
         chmod +x gradlew
         nohup ./gradlew bootRun &
         
-        cd ../genai-service
-        chmod +x gradlew
-        nohup ./gradlew bootRun &
         
         # Wait for services to start
         sleep 30
@@ -408,9 +351,8 @@ jobs:
         curl -f http://localhost:8081/health || echo "GenAI service not ready"
         curl -f http://localhost:8086/actuator/health || echo "Auth service not ready"
         curl -f http://localhost:8084/actuator/health || echo "Document service not ready"
-        curl -f http://localhost:8085/actuator/health || echo "GenAI service not ready"
         
-  # Job 7: Security Tests
+  # Job 6: Security Tests
   security-tests:
     name: Security Tests
     runs-on: ubuntu-latest
@@ -441,7 +383,7 @@ jobs:
         base: main
         head: HEAD
 
-  # Job 8: Performance Tests
+  # Job 7: Performance Tests
   performance-tests:
     name: Performance Tests
     runs-on: ubuntu-latest
@@ -492,11 +434,11 @@ jobs:
         # Run load test (will fail without running services, but validates config)
         k6 run --vus 1 --duration 10s load-test.js || true
         
-  # Job 9: Test Summary
+  # Job 8: Test Summary
   test-summary:
     name: Test Summary
     runs-on: ubuntu-latest
-    needs: [client-tests, genai-tests, auth-service-tests, document-service-tests, genai-service-tests, integration-tests, security-tests]
+    needs: [client-tests, genai-tests, auth-service-tests, document-service-tests, integration-tests, security-tests]
     if: always()
 
     steps:
@@ -535,11 +477,6 @@ jobs:
           echo "❌ Document Service Tests: FAILED" >> test-summary.md
         fi
         
-        if [ "${{ needs.genai-service-tests.result }}" = "success" ]; then
-          echo "✅ GenAI Service Tests: PASSED" >> test-summary.md
-        else
-          echo "❌ GenAI Service Tests: FAILED" >> test-summary.md
-        fi
         
         if [ "${{ needs.integration-tests.result }}" = "success" ]; then
           echo "✅ Integration Tests: PASSED" >> test-summary.md
@@ -581,7 +518,7 @@ jobs:
             body: summary
           });
           
-  # Job 10: Cleanup
+  # Job 9: Cleanup
   cleanup:
     name: Cleanup
     runs-on: ubuntu-latest

genAi/chains.py

@@ -12,15 +12,18 @@
 from langchain.output_parsers import PydanticOutputParser
 from langchain.chains.combine_documents.reduce import ReduceDocumentsChain
 
+
 class FlashcardChain:
     """
     Custom chain for generating flashcards from a summary.
     """
+
     def __init__(self, llm):
         llm = llm.with_structured_output(FlashcardResponse)
 
-        self.map_chain = ChatPromptTemplate.from_template(
-            """
+        self.map_chain = (
+            ChatPromptTemplate.from_template(
+                """
             Generate 3 flashcards from the following content.
 
             Each flashcard is an object with:
@@ -31,13 +34,18 @@ def __init__(self, llm):
             Content:
             {doc}
             """
-        ) | llm
+            )
+            | llm
+        )
+
+        self.reduce_chain = (
+            ChatPromptTemplate.from_template(
+                "Given the below list of flashcards, clean and deduplicate them. Return a final list of flashcards.\n\n"
+                "Flashcards:\n{flashcards}\n"
+            )
+            | llm
+        )
 
-        self.reduce_chain = ChatPromptTemplate.from_template(
-            "Given the below list of flashcards, clean and deduplicate them. Return a final list of flashcards.\n\n"
-            "Flashcards:\n{flashcards}\n"
-        ) | llm
-        
     async def invoke(self, documents: List[Document]):
         """
         Generate flashcards from the provided documents.
@@ -49,28 +57,32 @@ async def invoke(self, documents: List[Document]):
             FlashcardResponse: The generated flashcards in structured format.
         """
         contents = [doc.page_content for doc in documents]
+
         async def process(content: str):
             return (await self.map_chain.ainvoke({"doc": content})).flashcards
-        
+
         # Map in parallel
         results = await asyncio.gather(*[process(c) for c in contents])
         all_flashcards = [fc for group in results for fc in group]
 
         # Reduce
         joined = "\n".join(str(fc) for fc in all_flashcards)
         reduced = await self.reduce_chain.ainvoke({"flashcards": joined})
-        
+
         return reduced
-    
+
+
 class QuizChain:
     """
     Custom chain for generating a quiz from a summary.
     """
+
     def __init__(self, llm):
         llm = llm.with_structured_output(QuizResponse)
 
-        self.map_chain = ChatPromptTemplate.from_template(
-            """
+        self.map_chain = (
+            ChatPromptTemplate.from_template(
+                """
             Generate a mini (3-4 question) Quiz (MCQ and Short Answer) from the content below.
 
             Each question should include:
@@ -83,15 +95,20 @@ def __init__(self, llm):
             Content:
             {doc}
             """
-        ) | llm
-
-        self.reduce_chain = ChatPromptTemplate.from_template(
-            "Given the below list of quiz questions, consolidate, clean, and deduplicate them into a single whole quiz.\n"
-            "Ensure a realistic distribution of question types and difficulty levels.\n"
-            "Return a quiz with the selected questions.\n\n"
-            "Questions:\n{questions}\n"
-        ) | llm
-        
+            )
+            | llm
+        )
+
+        self.reduce_chain = (
+            ChatPromptTemplate.from_template(
+                "Given the below list of quiz questions, consolidate, clean, and deduplicate them into a single whole quiz.\n"
+                "Ensure a realistic distribution of question types and difficulty levels.\n"
+                "Return a quiz with the selected questions.\n\n"
+                "Questions:\n{questions}\n"
+            )
+            | llm
+        )
+
     async def invoke(self, documents: List[Document]):
         """
         Generate a quiz from the provided documents.
@@ -103,16 +120,17 @@ async def invoke(self, documents: List[Document]):
             QuizResponse: The generated quiz in structured format.
         """
         contents = [doc.page_content for doc in documents]
+
         async def process(content: str):
             mini_quiz: QuizResponse = await self.map_chain.ainvoke({"doc": content})
             return mini_quiz.questions
-        
+
         # Map in parallel
         results = await asyncio.gather(*[process(c) for c in contents])
         all_questions = [q for group in results for q in group]
 
         # Reduce
         joined = "\n".join(str(q) for q in all_questions)
         reduced: QuizResponse = await self.reduce_chain.ainvoke({"questions": joined})
-        
-        return reduced 
+
+        return reduced

genAi/conftest.py

@@ -0,0 +1,32 @@
+import pytest
+import os
+from unittest.mock import Mock, patch
+
+
+@pytest.fixture(autouse=True)
+def mock_openai_environment():
+    """Mock OpenAI environment variables and LLM initialization"""
+    with patch.dict(
+        os.environ,
+        {
+            "OPEN_WEBUI_API_KEY_CHAT": "test-key-chat",
+            "OPEN_WEBUI_API_KEY_GEN": "test-key-gen",
+            "OPENAI_API_KEY": "test-key",
+        },
+    ):
+        with patch("llm.ChatOpenAI") as mock_chat_openai:
+            # Create mock LLM instances
+            mock_llm = Mock()
+            mock_chat_openai.return_value = mock_llm
+
+            yield mock_llm
+
+
+@pytest.fixture(autouse=True)
+def clear_llm_instances():
+    """Clear LLM instances before each test"""
+    from main import llm_instances
+
+    llm_instances.clear()
+    yield
+    llm_instances.clear()

genAi/helpers.py

@@ -15,26 +15,27 @@ def save_document(document_name: str, document_base64: str) -> str:
 
     # Decode the base64 document
     document_bytes = base64.b64decode(document_base64)
-    
+
     # Create a unique file name
     file_path = os.path.join(documents_dir, document_name)
 
     # Save the document to a file
     with open(file_path, "wb") as file:
         file.write(document_bytes)
-    
+
     return file_path
 
+
 def delete_document(document_path: str):
     """
     Delete the document file.
     """
-    if '/example/' in document_path:
+    if "/example/" in document_path:
         # Skip deletion for example documents
         return
-    
+
     if os.path.exists(document_path):
         os.remove(document_path)
         logging.info(f"Document {document_path} deleted successfully.")
     else:
-        logging.warning(f"Document {document_path} does not exist.")
+        logging.warning(f"Document {document_path} does not exist.")