# Brave LLM Context - Core.Today AI API

> Web search for AI agents from Brave's independent index - returns pre-extracted, relevance-ranked snippets per source, sized to a token budget, ready to drop into an LLM prompt. 12 credits per request.

- **Provider**: Brave
- **Model ID**: brave/llm-context
- **Category**: Web Search
- **Credits**: 12 per request
- **Speed**: Fast
- **Quality**: High

## Features
- Snippets pre-extracted per source - no scraping step
- Token budget controls (total and per URL)
- Brave's independent web index
- Freshness filters and Goggles re-ranking

## Use Cases
- RAG context for chatbots and agents
- Citation-backed answers
- Research agents with a fixed token budget

## API Endpoint

Base URL: https://api.core.today/v1
Create Prediction: POST /predictions
Get Status: GET /predictions/{job_id}

## Authentication
Header: X-API-Key: YOUR_API_KEY

## Input Parameters

### Required
- **q**: string - The search query (max 600 characters, 75 words).

### Optional
- **country**: string (default: US) - Country the results come from (ALL for worldwide).
- **search_lang**: string (default: en) - Search language, e.g. ko, en, ja, en-gb, zh-hans.
- **count**: integer - Search results considered for context selection (1-50, default 20).
- **maximum_number_of_urls**: integer - Max distinct URLs in the context (1-50, default 20).
- **maximum_number_of_tokens**: integer - Approximate max tokens in the context (1024-32768, default 8192).
- **maximum_number_of_snippets**: integer - Max text snippets in the context (1-256, default 50).
- **maximum_number_of_tokens_per_url**: integer - Max tokens per URL (512-8192, default 4096).
- **maximum_number_of_snippets_per_url**: integer - Max snippets per URL (1-100, default 50).
- **context_threshold_mode**: string - Relevance threshold for including content - strict keeps less, lenient keeps more.
  Options: disabled, strict, lenient, balanced
- **freshness**: string - Page age filter - pd (24h), pw (7 days), pm (31 days), py (365 days) or YYYY-MM-DDtoYYYY-MM-DD.
- **safesearch**: string - Adult content filtering.
  Options: off, moderate, strict
- **spellcheck**: boolean (default: true) - Spell-check the query and search with the correction.
- **goggles**: array - Up to 3 Goggles (URL or inline definition) to re-rank results.
- **enable_source_metadata**: boolean (default: false) - Add site name, favicon, thumbnail and description to sources.

## Examples

### Basic request
A typical request with the most common options.
```json
{
  "model": "brave/llm-context",
  "input": {
    "q": "latest AI video generation models",
    "country": "US",
    "search_lang": "en",
    "maximum_number_of_tokens": 4096
  }
}
```

## Response Format
```json
{
  "job_id": "abc123",
  "status": "pending | processing | completed | failed",
  "result": "URL or data (when completed)"
}
```

## Usage Flow
1. POST /predictions with model and input -> receive job_id
2. GET /predictions/{job_id} -> poll until status is completed or failed
3. Result contains output URL(s)

## Tips
- The result is the provider's JSON as-is (grounding, sources)
- Links in the result point to the original pages - nothing is copied to your storage
- Pass grounding.generic[].snippets straight into your prompt; tune maximum_number_of_tokens to your context window

## Documentation
https://api-dashboard.search.brave.com/api-reference/ai/llm_context/get