fix(agents-api): Fix failing tests

Signed-off-by: Diwank Singh Tomer <[email protected]>

Author: creatorrr

agents-api/.pytest-runtimes

@@ -242,10 +242,11 @@ async def test_doc_with_embedding(pg_dsn, test_developer, test_doc):
     """Create a test document with embeddings."""
     pool = await create_db_pool(dsn=pg_dsn)
     embedding_with_confidence_0 = make_vector_with_similarity(d=0.0)
-    make_vector_with_similarity(d=0.5)
-    make_vector_with_similarity(d=-0.5)
+    embedding_with_confidence_0_5 = make_vector_with_similarity(d=0.5)
+    embedding_with_confidence_neg_0_5 = make_vector_with_similarity(d=-0.5)
     embedding_with_confidence_1_neg = make_vector_with_similarity(d=-1.0)
 
+    # Insert embedding with all 1.0s (similarity = 1.0)
     await pool.execute(
         """
         INSERT INTO docs_embeddings_store (developer_id, doc_id, index, chunk_seq, chunk, embedding)
@@ -269,7 +270,7 @@ async def test_doc_with_embedding(pg_dsn, test_developer, test_doc):
         f"[{', '.join([str(x) for x in embedding_with_confidence_0])}]",
     )
 
-    # Insert embedding with confidence -1
+    # Insert embedding with confidence 0.5
     await pool.execute(
         """
         INSERT INTO docs_embeddings_store (developer_id, doc_id, index, chunk_seq, chunk, embedding)
@@ -278,6 +279,30 @@ async def test_doc_with_embedding(pg_dsn, test_developer, test_doc):
         test_developer.id,
         test_doc.id,
         "Test content 2",
+        f"[{', '.join([str(x) for x in embedding_with_confidence_0_5])}]",
+    )
+    
+    # Insert embedding with confidence -0.5
+    await pool.execute(
+        """
+        INSERT INTO docs_embeddings_store (developer_id, doc_id, index, chunk_seq, chunk, embedding)
+        VALUES ($1, $2, 3, 3, $3, $4)
+        """,
+        test_developer.id,
+        test_doc.id,
+        "Test content 3",
+        f"[{', '.join([str(x) for x in embedding_with_confidence_neg_0_5])}]",
+    )
+    
+    # Insert embedding with confidence -1
+    await pool.execute(
+        """
+        INSERT INTO docs_embeddings_store (developer_id, doc_id, index, chunk_seq, chunk, embedding)
+        VALUES ($1, $2, 4, 4, $3, $4)
+        """,
+        test_developer.id,
+        test_doc.id,
+        "Test content 4",
         f"[{', '.join([str(x) for x in embedding_with_confidence_1_neg])}]",
     )
 
@@ -552,16 +577,26 @@ async def s3_client():
     with get_localstack() as localstack:
         s3_endpoint = localstack.get_url()
 
+        from botocore.config import Config
+        
         session = get_session()
         s3 = await session.create_client(
             "s3",
             endpoint_url=s3_endpoint,
             aws_access_key_id=localstack.env["AWS_ACCESS_KEY_ID"],
             aws_secret_access_key=localstack.env["AWS_SECRET_ACCESS_KEY"],
+            config=Config(s3={'addressing_style': 'path'})
         ).__aenter__()
 
         app.state.s3_client = s3
 
+        # Create the bucket if it doesn't exist
+        from agents_api.env import blob_store_bucket
+        try:
+            await s3.head_bucket(Bucket=blob_store_bucket)
+        except Exception:
+            await s3.create_bucket(Bucket=blob_store_bucket)
+        
         try:
             yield s3
         finally:

agents-api/tests/conftest.py

@@ -660,9 +660,10 @@ async def test_query_search_docs_by_embedding_with_different_confidence_levels(
     confidence_tests = [
         (0.99, 0),  # Very high similarity threshold - should find no results
         (0.7, 1),  # High similarity - should find 1 result (the embedding with all 1.0s)
-        (0.3, 2),  # Medium similarity - should find 2 results (including 0.3-0.7 embedding)
-        (-0.8, 3),  # Low similarity - should find 3 results (including -0.8 to 0.8 embedding)
-        (-1.0, 4),  # Lowest similarity - should find all 4 results (including alternating -1/1)
+        (0.3, 2),  # Medium similarity - should find 2 results (including 0.5 similarity embedding)
+        (-0.3, 3),  # Low similarity - should find 3 results (including 0 similarity embedding)
+        (-0.8, 4),  # Lower similarity - should find 4 results (including -0.5 similarity)
+        (-1.0, 5),  # Lowest similarity - should find all 5 results (including -1 similarity)
     ]
 
     for confidence, expected_min_results in confidence_tests:
@@ -686,7 +687,11 @@ async def test_query_search_docs_by_embedding_with_different_confidence_levels(
 
         if results:
             # Verify that all returned results meet the confidence threshold
+            # Distance uses cosine distance (0=identical, 2=opposite)
+            # The SQL converts confidence to search_threshold = 1.0 - confidence
+            # and filters results where distance <= search_threshold
+            search_threshold = 1.0 - confidence
             for result in results:
-                assert result.distance >= confidence, (
-                    f"Result distance {result.distance} is below confidence threshold {confidence}"
+                assert result.distance <= search_threshold, (
+                    f"Result distance {result.distance} exceeds search threshold {search_threshold} (confidence={confidence})"
                 )

agents-api/tests/test_docs_queries.py

@@ -235,7 +235,11 @@ async def test_middleware_cant_create_session_when_cost_limit_is_reached(make_re
     pool = await create_db_pool(dsn=pg_dsn)
     developer_id = uuid7()
     email = f'test-{developer_id}@example.com'
-    await create_developer(email=email, active=True, tags=[], settings={}, developer_id=developer_id, connection_pool=pool)
+    try:
+        await create_developer(email=email, active=True, tags=[], settings={}, developer_id=developer_id, connection_pool=pool)
+    finally:
+        await pool.close()
+    
     mock_user_cost_data = {'active': True, 'cost': float(free_tier_cost_limit) + 1.0, 'developer_id': developer_id, 'tags': []}
     with patch('agents_api.web.get_usage_cost', new=AsyncMock(return_value=mock_user_cost_data)):
         response = make_request(method='POST', url='/sessions', json={'agent_id': str(test_agent.id)}, headers={'X-Developer-Id': str(developer_id)})
@@ -248,12 +252,16 @@ async def test_middleware_cant_delete_session_when_cost_limit_is_reached(make_re
     pool = await create_db_pool(dsn=pg_dsn)
     developer_id = uuid7()
     email = f'test-{developer_id}@example.com'
-    await create_developer(email=email, active=True, tags=[], settings={}, developer_id=developer_id, connection_pool=pool)
+    try:
+        await create_developer(email=email, active=True, tags=[], settings={}, developer_id=developer_id, connection_pool=pool)
+    finally:
+        await pool.close()
+    
     mock_responses = [{'active': True, 'cost': float(free_tier_cost_limit) - 0.5, 'developer_id': developer_id, 'tags': []}, {'active': True, 'cost': float(free_tier_cost_limit) + 1.0, 'developer_id': developer_id, 'tags': []}]
     mock_get_usage_cost = AsyncMock()
     mock_get_usage_cost.side_effect = mock_responses
     with patch('agents_api.web.get_usage_cost', new=mock_get_usage_cost):
-        session_response = make_request(method='POST', url='/sessions', json={'agent': str(test_agent.id)}, headers={'X-Developer-Id': str(developer_id)})
+        session_response = make_request(method='POST', url='/sessions', json={'agent_id': str(test_agent.id)}, headers={'X-Developer-Id': str(developer_id)})
         assert session_response.status_code == status.HTTP_201_CREATED
         session_id = session_response.json()['id']
         delete_response = make_request(method='DELETE', url=f'/sessions/{session_id}', headers={'X-Developer-Id': str(developer_id)})

agents-api/tests/test_middleware.py

@@ -0,0 +1,61 @@
+# Python 3.13 Migration Plan for agents-api
+
+## Assessment Summary
+Switching to Python 3.13 is **mostly feasible** with some caveats. Most critical dependencies support Python 3.13, but there are a few blockers.
+
+## Key Findings
+
+### ✅ Dependencies that support Python 3.13:
+- **temporalio**: Full support (dropped Python 3.8, added 3.13)
+- **uvloop 0.21.0**: Full support with cp313 wheels
+- **numpy 2.0+**: Full support (requires 2.0 or higher)
+- **FastAPI, Pydantic, and most other dependencies**: Compatible
+
+### ❌ Blockers:
+- **pytype**: Only supports Python 3.8-3.12, no 3.13 support yet
+  - This is used for type checking in the development workflow
+
+## Migration Plan
+
+### 1. Update Python version constraints:
+- `agents-api/pyproject.toml`: Change `requires-python = ">=3.12,<3.13"` to `requires-python = ">=3.12,<3.14"`
+- `agents-api/.python-version`: Change from `3.12` to `3.13`
+- `agents-api/Dockerfile`: Change `FROM python:3.12-slim` to `FROM python:3.13-slim`
+- `agents-api/Dockerfile.worker`: Update similarly
+
+### 2. Handle pytype incompatibility:
+- **Option A**: Replace pytype with pyright (already in dev dependencies) for type checking
+- **Option B**: Keep pytype but run it with Python 3.12 while running the service with 3.13
+- **Option C**: Wait for pytype to add Python 3.13 support
+
+### 3. Update other services for consistency:
+- `integrations-service` uses same Python constraints (`>=3.12,<3.13`)
+- `cli` service uses `>=3.11,<3.13`
+- Both would need similar updates
+
+### 4. Update CI/CD:
+- GitHub Actions workflows use `uv python install` which respects `.python-version`
+- Docker builds will automatically use new Python version
+- No manual changes needed for workflows
+
+### 5. Testing Plan:
+- Run full test suite with Python 3.13
+- Check for any deprecation warnings or compatibility issues
+- Test Docker builds and deployments
+- Verify all integration tests pass
+
+## Recommendation
+The migration is mostly trivial except for the pytype issue. I recommend proceeding with **Option A** (replacing pytype with pyright) since pyright is already in your dev dependencies and supports Python 3.13.
+
+## Implementation Steps
+1. Replace pytype with pyright in poe tasks
+2. Update all Python version references
+3. Run `uv sync` to update dependencies
+4. Run full test suite
+5. Update Docker images
+6. Test in staging environment
+
+## Estimated Effort
+- Low complexity: Most changes are version string updates
+- Main effort: Replacing pytype with pyright configuration
+- Timeline: 1-2 hours of work + testing time