deepseek_fetch_many
Fetch up to 10 known URLs concurrently and return each page's title and Markdown content, with per-URL error handling and shared content limits.
Instructions
Fetch up to 10 known URLs concurrently and return each page's title and Markdown content.
Each URL goes through the same validation, cache, rate limiter and SSRF checks as deepseek_fetch. Empty entries are skipped and duplicates (same canonical URL) are fetched once. Results keep the input order; a failing URL yields an item with ok=false and an error instead of failing the batch. All fetches share one deadline (120 s by default); URLs still pending then are cancelled and reported with error code deadline_exceeded.
Each page gets min(max_chars_per_url, max_total_chars // count) characters and the total content never exceeds max_total_chars; a note says when any content was truncated. The envelope reports count (unique URLs), requests_used (upstream fetches attempted), cached_count (items served from the cache), failed_count (items with ok=false), partial (true when any item failed), results and notes. The top-level ok only says the batch ran; check partial or each item's ok for per-URL failures. Each result carries its own cached flag.
Args: urls: 1-10 absolute http(s) URLs, max 8192 characters each. max_chars_per_url: Content cap per page, applied after cleaning. max_total_chars: Total content budget across all pages. clean: Remove DeepSeek's spurious Markdown escaping (# at line start, , etc.). refresh: Bypass the response cache and fetch every page again.
Input Schema
| Name | Required | Description | Default |
|---|---|---|---|
| urls | Yes | ||
| clean | No | ||
| refresh | No | ||
| max_total_chars | No | ||
| max_chars_per_url | No |
Output Schema
| Name | Required | Description | Default |
|---|---|---|---|
No arguments | |||