\r \r

Cloudsways SmartSearch Skill\r

\r Search the web and extract intelligent fragments or full-text content directly into the LLM context.\r \r

Authentication\r

\r [cite_start]Authentication is handled via the URL endpoint and an Access Key[cite: 3, 5]. You must set the following environment variables before using the script:\r \r

export CLOUDSWAYS_BASE_PATH="your-base-path"\r
export CLOUDSWAYS_ENDPOINT="your-endpoint"\r
export CLOUDSWAYS_AK="your-access-key"\r
```\r
---\r
# Quick Start\r
\r
## Using the Script\r
\r
If you have configured the script locally, you can call it by passing a JSON object:\r
\r
```bash\r
./scripts/search.sh '\x3Cjson>'\r
```\r
\r
**Examples:**\r
\r
```bash\r
# Basic search\r
./scripts/search.sh '{"q": "python async patterns"}'\r
\r
# Search with time filter and pagination\r
./scripts/search.sh '{"q": "Apple earnings", "freshness": "Week", "count": 20}'\r
\r
# Deep research (extracts full content and dynamic key fragments)\r
./scripts/search.sh '{"q": "Agentic AI architecture", "enableContent": true, "mainText": true}'\r
```\r
\r
---\r
\r
## Basic Search (cURL)\r
\r
```bash\r
curl -i --location --request GET 'https://{BasePath}/search/{Endpoint}/smart?q=latest+developments+in+quantum+computing&count=5' \\r
  --header 'Authorization: Bearer {AK}' \\r
  --header 'pragma: no-cache'\r
```\r
\r
## Advanced Search with Full Content (cURL)\r
\r
```bash\r
curl -i --location --request GET 'https://{BasePath}/search/{Endpoint}/smart?q=machine+learning+best+practices&count=10&enableContent=true&mainText=true' \\r
  --header 'Authorization: Bearer {AK}' \\r
  --header 'pragma: no-cache'\r
```\r
\r
---\r
\r
## API Reference\r
\r
### Endpoint\r
\r
```\r
GET https://{BasePath}/search/{Endpoint}/smart\r
```\r
\r
### Headers\r
\r
| Header | Type | Value | Description |\r
|---|---|---|---|\r
| `Authorization` | String | `Bearer {AK}` | Input value uses your assigned AccessKey. |\r
| `pragma` | String | `no-cache` | Entering `no-cache` returns results without caching, making each request independent; omitting this caches results for the same query term for 10 minutes. |\r
\r
### Request Parameters\r
\r
| Parameter | Required | Type | Default | Description |\r
|---|---|---|---|---|\r
| `q` | Y | String | - | User search query term; cannot be empty. |\r
| `count` | N | Short | 10 | Number of search results included. Enum values: `10`, `20`, `30`, `40`, `50`. |\r
| `freshness` | N | String | null | Filter search results by time periods: `Day` (Past 24 hours), `Week`, `Month`. |\r
| `offset` | N | Short | 0 | Zero-based offset indicating the number of search results to skip. Use with `count` to paginate. |\r
| `enableContent` | N | bool | false | Controls whether full text (content) extraction is enabled: `true` or `false`. |\r
| `contentType` | N | String | TEXT | Full text return format. Optional values: `HTML`, `MARKDOWN`, `TEXT`. |\r
| `contentTimeout` | N | Float | 3.0 | Full text read timeout. Maximum supported is 10 seconds. For low latency, set to `0.1`. |\r
| `mainText` | N | bool | false | Used to return dynamic summary key fragments. Takes effect only when `enableContent` is true. |\r
\r
### Response Format\r
\r
```json\r
{\r
  "queryContext": {\r
    "originalQuery": "open ai recent news"\r
  },\r
  "webPages": {\r
    "value": [\r
      {\r
        "name": "OpenAI just launched ChatGPT-5.1 Pro...",\r
        "url": "https://www.example.com/page",\r
        "datePublished": "2025-07-14T01:15:00.0000000Z",\r
        "snippet": "ChatGPT is looking to compete with Gemini...",\r
        "mainText": "Now GPT-5.1 Pro takes a lot of these improvements and makes them smarter...",\r
        "content": "Full extracted body text goes here...",\r
        "score": 0.8521444\r
      }\r
    ]\r
  }\r
}\r
```\r
\r
---\r
\r
## Content  Strategy\r
\r
Understanding which text field to request is critical for token optimization:\r
\r
| Field | Latency | Relevance | Description |\r
|---|---|---|---|\r
| `snippet` | Lowest | Good | A short text summary describing the webpage content. Best for general queries and fast browsing. |\r
| `mainText` | Medium | Highest | Dynamic summary fragments most relevant to the query term extracted from the body text. Smarter than the snippet. |\r
| `content` | Higher | High | Webpage full text information. Best for deep research, but consumes significant context tokens. |\r
\r
**Tips:**\r
\r
- For standard answers, rely on the default `snippet`.\r
- When you need precise, detailed answers without reading the whole page, use `enableContent=true` and `mainText=true`.\r
- Always set `pragma: no-cache` if you are querying real-time news or stock updates.\r
\r
---\r
\r
## MCP Client Integration (Alternative)\r
\r
For environments supporting the Model Context Protocol (MCP), configure your MCP Client using `streamable_http`:\r
\r
```json\r
{\r
  "mcpServers": {\r
    "smartsearch": {\r
      "transport": "streamable_http",\r
      "url": "https://searchmcp.cloudsway.net/search/{YOUR_SMARTSEARCH_ENDPOINT}/smart",\r
      "headers": {\r
        "Authorization": "Bearer {YOUR_SMARTSEARCH_AK}"\r
      }\r
    }\r
  }\r
}\r