# Context Pack

> Token-budgeted context for an LLM prompt, gathered from the web.

Source: http://localhost:3000/docs/endpoints/context

## Context Pack: `POST /v1/context/`

A token-budgeted context pack for coding and RAG agents. Cost: 2 credits. Available on the free demo.

```bash
curl -X POST https://api.hushsearch.io/v1/context/ \
  -H "X-API-Key: hs_live_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "query": "FastAPI dependency injection with async generators",
  "tokens_num": 2000
}'
```

Returns a single `context` string sized to a token budget, built from the most relevant passages across several sources, with the sources listed separately. Useful for dropping straight into a prompt.

## Parameters

| Field | Type | Default | Notes |
|---|---|---|---|
| `query` | string | — | **Required.** 1–2,000 characters. |
| `tokens_num` | integer or `"dynamic"` | `"dynamic"` | Approximate token budget, 50–100,000. |
| `num_results` | integer | `12` | Sources to consider, 1–40. |
| `category` | string | — | Soft hint such as `code`, `docs`, `github`, `financial_report`. |

## Response

```json
{
  "query": "how does pgvector HNSW indexing work",
  "context": "[1] pgvector adds HNSW indexes …",
  "tokens_num": 2048,
  "sources": [{ "index": 1, "title": "…", "url": "https://…" }],
  "credits_used": 2,
  "credits_remaining": 498,
  "response_time_ms": 2410
}
```
