---
id: task-33
title: Add import and integration smoke tests
status: Open
assignee: []
created_date: '2025-12-04'
updated_date: '2025-12-04'
labels:
- testing
- reliability
- high
dependencies: []
priority: high
---
## Description
<!-- SECTION:DESCRIPTION:BEGIN -->
A missing dependency (`pymupdf4llm`) caused production failures despite 243 passing tests. The root cause: our test suite exercises the wrong code paths.
### Root Cause Analysis
**pytest.ini line 28 excludes critical tests by default:**
```
-m "not expensive and not integration and not accuracy and not marker"
```
When developers run `pytest`, they see all tests pass but the following critical paths are NEVER exercised:
1. **fast_extractor.py** - Zero test imports. This is the module that failed.
2. **services/document_processor.py** - Zero test imports. Orchestrates all processing.
3. **services/background_jobs.py** - Zero test imports. Manages async tasks.
4. **8 tools/*.py files** - Minimal coverage. MCP tool implementations.
5. **9 prompts/*.py files** - Zero test imports. LLM templates.
### The Lazy Import Problem
Five locations use `try: import X except ImportError` patterns that defer failures to runtime:
- `fast_extractor.py:60` - pymupdf4llm (THE BUG)
- `marker_extractor.py:68,105,113,857` - torch (optional, but masks issues)
These pass at import time because the try/except defers the actual import until function execution.
### Solution: Two New Test Types
**1. Import Verification Test (Fast, Always Runs)**
A test that imports every module and verifies all required dependencies are installed:
```python
# tests/test_imports.py
"""Verify all modules import successfully and dependencies are installed."""
import pytest
def test_all_modules_import():
"""Import every module to catch missing dependencies."""
# Core
from registry_review_mcp import server
from registry_review_mcp.config import settings
# Services
from registry_review_mcp.services import document_processor
from registry_review_mcp.services import background_jobs
# Extractors (including lazy dependencies)
from registry_review_mcp.extractors import fast_extractor
from registry_review_mcp.extractors import marker_extractor
from registry_review_mcp.extractors import llm_extractors
# All tools
from registry_review_mcp.tools import upload_tools
from registry_review_mcp.tools import session_tools
from registry_review_mcp.tools import document_tools
from registry_review_mcp.tools import evidence_tools
from registry_review_mcp.tools import mapping_tools
from registry_review_mcp.tools import report_tools
from registry_review_mcp.tools import validation_tools
from registry_review_mcp.tools import human_review_tools
def test_pymupdf4llm_available():
"""Verify pymupdf4llm is installed (the bug we caught)."""
import pymupdf4llm
assert hasattr(pymupdf4llm, 'to_markdown')
def test_marker_available():
"""Verify marker-pdf is installed."""
from marker.models import create_model_dict
from marker.converters.pdf import PdfConverter
```
**2. Integration Smoke Test (Fast, Runs on CI)**
A lightweight test that exercises critical paths without expensive API calls:
```python
# tests/test_smoke_integration.py
"""Smoke tests for critical integration paths."""
import pytest
from pathlib import Path
@pytest.mark.integration
class TestIntegrationSmoke:
"""Fast integration tests that verify critical paths work."""
@pytest.mark.asyncio
async def test_fast_extract_pdf(self, example_documents_path):
"""Verify fast_extract_pdf works with real PDF."""
from registry_review_mcp.extractors.fast_extractor import fast_extract_pdf
pdf = example_documents_path / "4997Botany22_Public_Project_Plan.pdf"
result = await fast_extract_pdf(str(pdf))
assert "markdown" in result
assert len(result["markdown"]) > 100
assert result["extraction_method"] == "pymupdf4llm"
@pytest.mark.asyncio
async def test_mcp_server_startup(self):
"""Verify MCP server can start without errors."""
from registry_review_mcp.server import mcp, start_review
# Verify tools are registered
assert mcp is not None
# Verify start_review exists and is callable
assert callable(start_review)
```
### Why This Matters
Without these tests:
- A developer removes a dependency from pyproject.toml
- They run `pytest` - all 243 tests pass
- They commit and push
- Production breaks
With these tests:
- Import test fails immediately
- Developer knows something is wrong
- Bug never reaches production
### Acceptance Criteria
- [ ] `test_imports.py` imports all source modules
- [ ] `test_imports.py` verifies critical dependencies (pymupdf4llm, marker, pdfplumber)
- [ ] `test_smoke_integration.py` exercises fast_extract_pdf with real PDF
- [ ] `test_smoke_integration.py` verifies MCP server can start
- [ ] Integration smoke tests run in CI (not excluded by default)
- [ ] Import tests run on every pytest invocation
<!-- SECTION:DESCRIPTION:END -->
