Document Ingestion Pipeline

Build a scalable document ingestion pipeline for our RAG system.

## Document Sources
{{document_sources}}

## Document Types
{{document_types}}

## Volume Expectations
{{volume_expectations}}

Implement a complete pipeline:

```python
class IngestionPipeline:
    async def ingest(self, source: DocumentSource) -> IngestionResult:
        """
        1. Document extraction (PDF, DOCX, HTML, etc.)
        2. Content cleaning and normalization
        3. Chunking with metadata preservation
        4. Embedding generation (batched)
        5. Vector store upsert
        6. Index refresh
        """
        pass
```

Include:
- Parallel processing architecture
- Error handling and retry logic
- Progress tracking and resumability
- Deduplication strategy
- Incremental update support