llm-context — скилл для Claude Code: что делает и как установить

---
name: llm-context
description: USE FOR RAG/LLM grounding. Returns pre-extracted web content (text, tables, code) optimized for LLMs. GET + POST. Adjust max_tokens/count based on complexity. Supports Goggles, local/POI. For AI answers use answers. Recommended for anyone building AI/agentic applications.
---

# LLM Context

> **Requires API Key**: Get one at https://api.search.brave.com
>
> **Plan**: Included in the **Search** plan. See https://api-dashboard.search.brave.com/app/subscriptions/subscribe

Brave LLM Context API delivers pre-extracted, relevance-ranked web content optimized for grounding LLM responses in real-time search results. Unlike traditional web search APIs that return links and snippets, LLM Context extracts the actual page content—text chunks, tables, code blocks, and structured data—so your LLM or AI agent can reason over it directly.

## LLM Context vs AI Grounding

| Feature | LLM Context (this) | AI Grounding (`answers`) |
|--|--|--|
| Output | Raw extracted content for YOUR LLM | End-to-end AI answers with citations |
| Interface | REST API (GET/POST) | OpenAI-compatible `/chat/completions` |
| Searches | Single search per request | Multi-search (iterative research) |
| Speed | Fast (<1s) | Slower |
| Plan | Search | Answers |
| Endpoint | `/res/v1/llm/context` | `/res/v1/chat/completions` |
| Best for | AI agents, RAG pipelines, tool calls | Chat interfaces, research mode |

## Endpoint

```http
GET  https://api.search.brave.com/res/v1/llm/context
POST https://api.search.brave.com/res/v1/llm/context
```

**Authentication**: `X-Subscription-Token: <API_KEY>` header

**Optional Headers**:
- `Accept-Encoding: gzip` — Enable gzip compression

## Quick Start

### GET Request
```bash
curl -s "https://api.search.brave.com/res/v1/llm/context?q=tallest+mountains+in+the+world" \
  -H "Accept: application/json" \
  -H "X-Subscription-Token: ${BRAVE_SEARCH_API_KEY}"
```

### POST Request (JSON body)
```bash
curl -s --compressed -X POST "https://api.search.brave.com/res/v1/llm/context" \
  -H "Accept: application/json" \
  -H "Accept-Encoding: gzip" \
  -H "X-Subscription-Token: ${BRAVE_SEARCH_API_KEY}" \
  -H "Content-Type: application/json" \
  -d '{"q": "tallest mountains in the world"}'
```

### With Goggles (Inline)
```bash
curl -s "https://api.search.brave.com/res/v1/llm/context" \
  -H "Accept: application/json" \
  -H "X-Subscription-Token: ${BRAVE_SEARCH_API_KEY}" \
  -G \
  --data-urlencode "q=rust programming" \
  --data-urlencode 'goggles=$discard
$site=docs.rs
$site=rust-lang.org'
```

## Parameters

### Query Parameters

| Parameter | Type | Required | Default | Description |
|--|--|--|--|--|
| `q` | string | **Yes** | - | Search query (1-400 chars, max 50 words) |
| `country` | string | No | `US` | Search country (2-letter country code or `ALL`) |
| `search_lang` | string | No | `en` | Language preference (2+ char language code) |
| `count` | int | No | `20` | Max search results to consider (1-50) |
| `spellcheck` | bool | No | `true` | Whether to spellcheck the query before searching |
| `freshness` | string | No | `""` | Filters search results by page age. The age of a page is determined by the most relevant date reported by the content, such as its published or last modified date. Supported values: `pd` (24h or less), `pw` (7 days or less), `pm` (31 days or less), `py` (365 days or less), or a custom date range `YYYY-MM-DDtoYYYY-MM-DD` (e.g. `2022-04-01to2022-07-30`). |

### Context Size Parameters

| Parameter | Type | Required | Default | Description |
|--|--|--|--|--|
| `maximum_number_of_urls` | int | No | `20` | Max URLs in response (1-50) |
| `maximum_number_of_tokens` | int | No | `8192` | Approximate max tokens in context (1024-32768) |
| `maximum_number_of_snippets` | int | No | `50` | Max snippets across all URLs (1-256) |
| `maximum_number_of_tokens_per_url` | int | No | `4096` | Max tokens per individual URL (512-8192) |
| `maximum_number_of_snippets_per_url` | int | No | `50` | Max snippets per individual URL (1-100) |


### Filtering & Local Parameters

| Parameter | Type | Required | Default | Description |
|--|--|--|--|--|
| `context_threshold_mode` | string | No | `null` | Relevance threshold for including content (`strict`/`balanced`/`lenient`/`disabled`) |
| `safesearch` | string | No | `null` | Adult content filter (`off`/`moderate`/`strict`); not set means no filtering, except local recall which stays `strict` |
| `enable_local` | bool | No | `null` | Local recall control (`true`/`false`/`null`, see below) |
| `goggles` | string/list | No | `null` | Goggle URL or inline definition for custom re-ranking |
| `enable_source_metadata` | bool | No | `false` | Adds `site_name`, `favicon`, `thumbnail` and `description` to each `sources[url]` entry |

## Context Size Guidelines

| Task Type | count | max_tokens | Example |
|--|--|--|--|
| Simple factual | 5 | 2048 | "What year was Python created?" |
| Standard queries | 20 | 8192 | "Best practices for React hooks" |
| Complex research | 50 | 16384 | "Compare AI frameworks for production" |

Larger context windows provide more information but increase latency and cost (of your inference). Start with defaults and adjust.

## Threshold Modes

| Mode | Behavior |
|--|--|
| `null` (not set) | **Default** — resolves to `lenient` on the current API version |
| `strict` | Higher threshold — fewer but more relevant results |
| `balanced` | Good balance between coverage and relevance |
| `lenient` | Lower threshold — more results, may include less relevant content |
| `disabled` | No threshold filtering — return all extracted content |

## Local Recall

The `enable_local` parameter controls location-aware recall:

| Value | Behavior |
|--|--|
| `null` (not set) | **Auto-detect** — local recall enabled when any location header is provided |
| `true` | **Force local** — always use local recall, even without location headers |
| `false` | **Force standard** — always use standard web ranking, even with location headers |

For most use cases, omit `enable_local` and let the API auto-detect from location headers.

## Location Headers

| Header | Type | Description |
|--|--|--|
| `X-Loc-Lat` | float | Latitude (-90.0 to 90.0) |
| `X-Loc-Long` | float | Longitude (-180.0 to 180.0) |
| `X-Loc-City` | string | City name |
| `X-Loc-State` | string | State/region code (ISO 3166-2) |
| `X-Loc-State-Name` | string | State/region name |
| `X-Loc-Country` | string | 2-letter country code |
| `X-Loc-Postal-Code` | string | Postal code |

> **Priority**: `X-Loc-Lat` + `X-Loc-Long` take precedence. When provided, text-based headers (City, State, Country, Postal-Code) are not used for location resolution. Provide text-based headers **only** when you don't have coordinates.

### Example: With Coordinates
```bash
curl -s "https://api.search.brave.com/res/v1/llm/context" \
  -H "Accept: application/json" \
  -H "X-Subscription-Token: ${BRAVE_SEARCH_API_KEY}" \
  -H "X-Loc-Lat: 37.7749" \
  -H "X-Loc-Long: -122.4194" \
  -G \
  --data-urlencode "q=best coffee shops near me"
```

### Example: With Place Name
```bash
curl -s "https://api.search.brave.com/res/v1/llm/context" \
  -H "Accept: application/json" \
  -H "X-Subscription-Token: ${BRAVE_SEARCH_API_KEY}" \
  -H "X-Loc-City: San Francisco" \
  -H "X-Loc-State: CA" \
  -H "X-Loc-Country: US" \
  -G \
  --data-urlencode "q=best coffee shops near me"
```

## Goggles (Custom Ranking) — Unique to Brave

Goggles let you **control which sources ground your LLM** — essential for RAG quality.

| Use Case | Goggle Rules |
|--|--|
| Official docs only | `$discard\n$site=docs.python.org` |
| Exclude user content | `$discard,site=reddit.com\n$discard,site=stackoverflow.com` |
| Academic sources | `$discard\n$site=arxiv.org\n$site=scholar.google.com` |
| No paywalls | `$discard,site=medium.com` |

| Method | Example |
|--|--|
| **Hosted** | `--data-urlencode "goggles=https://raw.githubusercontent.com/brave/goggles-quickstart/main/goggles/1k_short.goggle"` |
| **Inline** | `--data-urlencode 'goggles=$discard\n$site=example.com'` |

> **Hosted** goggles must be on GitHub/GitLab, include `! name:`, `! description:`, `! author:` headers, and be registered at https://search.brave.com/goggles/create. **Inline** rules need no registration.

**Syntax**: Rules start with `$` + comma-separated options. **Actions** (pick one): `discard`, `boost[=N]`, `downrank[=N]` — N is an integer 1–10. **Site filter**: `site=DOMAIN`. Example: `$site=example.com,boost=3`. Separate rules with `\n` (`%0A`).

**Allow list**: `$discard\n$site=docs.python.org\n$site=developer.mozilla.org` — **Block list**: `$discard,site=pinterest.com\n$discard,site=quora.com`

**Resources**: [Discover](https://search.brave.com/goggles/discover) · [Syntax](https://search.brave.com/help/goggles) · [Quickstart](https://github.com/brave/goggles-quickstart)

## Response Format

### Standard Response

```json
{
  "grounding": {
    "generic": [
      {
        "url": "https://example.com/page",
        "title": "Page Title",
        "snippets": [
          "Relevant text chunk extracted from the page...",
          "Another relevant passage from the same page..."
        ]
      }
    ],
    "map": []
  },
  "sources": {
    "https://example.com/page": {
      "title": "Page Title",
      "hostname": "example.com",
      "age": ["Wednesday, January 15, 2025", "2025-01-15", "392 days ago", "2025-01-15T13:45:02Z"]
    }
  }
}
```

### Local Response (with `enable_local`)

```json
{
  "grounding": {
    "generic": [...],
    "poi": {
      "name": "Business Name",
      "url": "https://business.com",
      "title": "Title of business.com website",
      "snippets": ["Business details and information..."]
    },
    "map": [
      {
        "name": "Place Name",
        "url": "https://place.com",
        "title": "Title of place.com website",
        "snippets": ["Place information and details..."]
      }
    ]
  },
  "sources": {
    "https://business.com": {
      "title": "Business Name",
      "hostname": "business.com",
      "age": []
    }
  }
}
```

### Response Fields

| Field | Type | Description |
|--|--|--|
| `grounding` | object | Container for all grounding content by type |
| `grounding.generic` | array | Array of URL objects with extracted content (main grounding data) |
| `grounding.generic[].url` | string | Source URL |
| `grounding.generic[].title` | string | Page title |
| `grounding.generic[].snippets` | array | Extracted smart chunks relevant to the query |
| `grounding.poi` | object/null | Point of interest data (only with local recall) |
| `grounding.poi.name` | string/null | Point of interest name |
| `grounding.poi.url` | string/null | POI source URL |
| `grounding.poi.title` | string/null | POI page title |
| `grounding.poi.snippets` | array/null | POI text snippets |
| `grounding.map` | array | Map/place results (only with local recall) |
| `grounding.map[].name` | string/null | Place name |
| `grounding.map[].url` | string/null | Place source URL |
| `grounding.map[].title` | string/null | Place page title |
| `grounding.map[].snippets` | array/null | Place text snippets |
| `sources` | object | Metadata for all referenced URLs, keyed by URL |
| `sources[url].title` | string | Page title |
| `sources[url].hostname` | string | Source hostname |
| `sources[url].age` | array | The page's date in four fixed positions: full date, `YYYY-MM-DD`, relative age, ISO 8601 timestamp. Empty when the page has no known date |
| `sources[url].description` | string | The page's own description, independent of the query. Requires `enable_source_metadata` |
| `sources[url].site_name` | string | Site name. Requires `enable_source_metadata` |
| `sources[url].favicon` | string | Favicon URL. Requires `enable_source_metadata` |
| `sources[url].thumbnail` | object/null | Page thumbnail (`src`, `original`). Requires `enable_source_metadata` |
| `sources[url].snippet` | string? | Best snippet for the page |

**Note**: Snippets may contain plain text OR JSON-serialized structured data (tables, schemas, code blocks). LLMs handle this mixed format well.

## Use Cases

- **AI Agents**: Give your agent a web search tool that returns ready-to-use content in a single call
- **RAG Pipelines**: Ground LLM responses in fresh, relevant web content
- **AI Assistants & Chatbots**: Provide factual answers backed by real sources
- **Question Answering**: Retrieve focused context for specific queries
- **Fact Checking**: Verify claims against current web content
- **Content Research**: Gather source material on any topic with one API call

## Best Practices

- **Token budget**: Start with defaults (`maximum_number_of_tokens=8192`, `count=20`). Reduce for simple lookups, increase for complex research.
- **Source quality**: Use Goggles to restrict to trusted sources. Set `context_threshold_mode=strict` when precision > recall.
- **Performance**: Use smallest `count` and `maximum_number_of_tokens` that meet your needs. For local queries, provide location headers.