contextual-retrieval

# Contextual Retrieval
Prepend situational context to chunks before embedding to preserve document-level meaning.

## The Problem

Traditional chunking loses context:
```
Original document: "ACME Q3 2024 Earnings Report..."
Chunk: "Revenue increased 15% compared to the previous quarter."

Query: "What was ACME's Q3 2024 revenue growth?"
Result: Chunk doesn't mention "ACME" or "Q3 2024" - retrieval fails
```

## The Solution

**Contextual Retrieval** prepends a brief context to each chunk:
```
Contextualized chunk:
"This chunk is from ACME Corp's Q3 2024 earnings report, specifically
the revenue section. Revenue increased 15% compared to the previous quarter."
```

## Implementation

### Context Generation
```python
import anthropic

client = anthropic.Anthropic()

CONTEXT_PROMPT = """
<document>
{document}
</document>

Here is the chunk we want to situate within the document:
<chunk>
{chunk}
</chunk>

Please give a short, succinct context (1-2 sentences) to situate this chunk
within the overall document. Focus on information that would help retrieval.
Answer only with the context, nothing else.
"""

def generate_context(document: str, chunk: str) -> str:
    """Generate context for a single chunk."""
    response = client.messages.create(
        model="claude-sonnet-4-20250514",
        max_tokens=150,
        messages=[{
            "role": "user",
            "content": CONTEXT_PROMPT.format(document=document, chunk=chunk)
        }]
    )
    return response.content[0].text

def contextualize_chunk(document: str, chunk: str) -> str:
    """Prepend context to chunk."""
    context = generate_context(document, chunk)
    return f"{context}\n\n{chunk}"
```

### Batch Processing with Caching
```python
from anthropic import Anthropic

client = Anthropic()

def contextualize_chunks_cached(document: str, chunks: list[str]) -> list[str]:
    """
    Use prompt caching to efficiently process many chunks from same document.
    Document is cached, only chunk changes per r