> ## Documentation Index
> Fetch the complete documentation index at: https://composio-preview.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Octen

> Search Octen's real-time web index, extract web content, and create text or multimodal embeddings.

![Octen logo](https://mintcdn.com/composio-preview/gqjBKmmJCINafaiK/images/toolkits/octen.svg?fit=max\&auto=format\&n=gqjBKmmJCINafaiK\&q=85\&s=d7acf388727426d7a780145595bc7aa0)

* **Category:** artificial intelligence
* **Auth:** API\_KEY
* **Composio-managed OAuth available?** N/A
* **Tools:** 9
* **Triggers:** 0
* **Slug:** `OCTEN`
* **Version:** 00000000\_00

## Tools

> [accordion] Broad Search
>
> **Slug:** `OCTEN_BROAD_SEARCH`
>
> Decompose one question into up to 30 sub-queries and search Octen's live web index concurrently. Returns each non-deduplicated result group and the native usage envelope. Cost fans out: one invocation may bill multiple search calls plus full-content overage. This action makes one request and does not automatically retry ambiguous failures.
>
> #### Input Parameters
>
> **query** `string` *required*
>
> Question Octen should decompose into multiple web-search queries.
>
> **max\_queries** `integer`
>
> Maximum generated sub-queries (1-30). Each generated query can become a separately metered search call, so larger values multiply cost.
>
> **search\_options** `object`
>
> Complete Web Search options applied independently to every generated sub-query. Counts and full-content charges fan out across all groups.
>
> #### Output
>
> **data** `string` *required*
>
> Data from the action execution
>
> **error** `string`
>
> Error if any occurred during the execution of the action
>
> **successful** `boolean` *required*
>
> Whether or not the action execution was successful or not

> [accordion] Create Multimodal Embeddings
>
> **Slug:** `OCTEN_CREATE_MULTIMODAL_EMBEDDINGS`
>
> Create independent or fused embeddings from up to 20 text, image, and URL-video elements. This is one metered, non-idempotent computation: standard-model text and media cost \$0.05/\$0.12 per 1M tokens, while large-model text and media cost \$0.10/\$0.25 per 1M tokens. The instruction is billed as text. The action sends one request and does not retry ambiguous failures.
>
> #### Input Parameters
>
> **fps** `number`
>
> Video sampling rate greater than 0 and at most 1 frame per second; Octen caps sampling at 64 frames.
>
> **input** `object` *required*
>
> Bounded text, image, and video content to embed.
>
> **model** `string ("octen-vl-embedding" | "octen-vl-embedding-large")` *required*
>
> Embedding model. octen-vl-embedding supports up to 2,048 dimensions and costs \$0.05 per 1M text tokens and \$0.12 per 1M image/video tokens; octen-vl-embedding-large supports up to 4,096 dimensions and costs \$0.10 and \$0.25 per 1M tokens respectively.
>
> **instruct** `string`
>
> Embedding instruction, billed as text tokens. Its characters and all text content together must not exceed 10,000.
>
> **dimension** `integer`
>
> Output dimensions. Omit for the model maximum; standard supports at most 2,048 and large supports at most 4,096.
>
> **enable\_fusion** `boolean`
>
> Return one fused vector for all elements when true; otherwise return one vector per content element.
>
> #### Output
>
> **data** `string` *required*
>
> Data from the action execution
>
> **error** `string`
>
> Error if any occurred during the execution of the action
>
> **successful** `boolean` *required*
>
> Whether or not the action execution was successful or not

> [accordion] Create Text Embeddings
>
> **Slug:** `OCTEN_CREATE_TEXT_EMBEDDINGS`
>
> Create indexed text embeddings for up to 1,000 inputs with Octen in one metered request. Billing per 1 million input tokens is \$0.01 for 0.6b, \$0.04 for 4b, or \$0.07 for 8b; query mode also bills its prepended retrieval instruction. Replaying an ambiguous request can incur the token charge again, so do not retry it automatically.
>
> #### Input Parameters
>
> **input** `array` *required*
>
> Texts to embed (at most 1,000; each at most 10,000 characters). The complete serialized request body must not exceed 2 MB. Input order maps to each returned result index.
>
> **model** `string ("octen-embedding-0.6b" | "octen-embedding-4b" | "octen-embedding-8b")`
>
> Embedding model. Per 1 million input tokens, Octen documents prices of \$0.01 for 0.6b, \$0.04 for 4b, and \$0.07 for 8b. Maximum dimensions are 1,024, 2,560, and 4,096 respectively.
>
> **dimension** `integer`
>
> Output dimensions. Omit for the selected model's maximum; cannot exceed 1,024 for 0.6b, 2,560 for 4b, or 4,096 for 8b.
>
> **input\_type** `string ("query" | "document")`
>
> Use query for retrieval queries or document for corpus text. Query mode prepends a retrieval instruction that Octen includes in input\_tokens; null or omission adds no special prompt.
>
> #### Output
>
> **data** `string` *required*
>
> Data from the action execution
>
> **error** `string`
>
> Error if any occurred during the execution of the action
>
> **successful** `boolean` *required*
>
> Whether or not the action execution was successful or not

> [accordion] Extract URL Content
>
> **Slug:** `OCTEN_EXTRACT_URL_CONTENT`
>
> Extract content and optional media or links from up to 20 URLs in one metered request. Every URL's success or failed status is returned; only successful URLs are billed by resolved mode. Do not retry an ambiguous whole batch, because successful URLs may be billed again; inspect item errors and retry only known failed URLs when safe.
>
> #### Input Parameters
>
> **mode** `string ("standard" | "advanced" | "auto")`
>
> Extraction mode. Advanced costs more; auto may resolve and bill each successful URL as standard or advanced.
>
> **urls** `array` *required*
>
> URLs to extract in one batch (at most 20, each at most 2,048 characters).
>
> **query** `string`
>
> Relevance query. When supplied, Octen returns highlights instead of full\_content.
>
> **format** `string ("markdown" | "text")`
>
> Output format for extracted content or highlights.
>
> **timeout** `integer`
>
> Per-URL extraction timeout in seconds (1-60), distinct from the action's network timeout.
>
> **include\_audio** `boolean`
>
> Include extracted audio metadata when available.
>
> **include\_links** `object`
>
> Controls extracted links and their per-item limit.
>
> **include\_images** `boolean`
>
> Include extracted image metadata when available.
>
> **include\_videos** `boolean`
>
> Include extracted video metadata when available.
>
> **max\_age\_seconds** `integer`
>
> Maximum acceptable cached-content age in seconds (300-31,536,000).
>
> #### Output
>
> **data** `string` *required*
>
> Data from the action execution
>
> **error** `string`
>
> Error if any occurred during the execution of the action
>
> **successful** `boolean` *required*
>
> Whether or not the action execution was successful or not

> [accordion] Search Businesses
>
> **Slug:** `OCTEN_SEARCH_BUSINESSES`
>
> Search Octen's web index once for business pages and optional company or person cards. Domain, time, and language filters affect ranked results only. Calls cost \$5 per 1,000; full content after the first 10 results costs \$0.50 per 1,000 results. This action does not automatically retry ambiguous failures.
>
> #### Input Parameters
>
> **count** `integer`
>
> Maximum ranked results returned (1-100).
>
> **query** `string` *required*
>
> Business search query; supports site:\<domain> and -site:\<domain> operators for ranked results.
>
> **end\_time** `string`
>
> ISO 8601 end time for ranked results only; overrides time\_range.
>
> **entities** `object`
>
> Company and person entity-card settings.
>
> **language** `array`
>
> ISO 639-1 languages used to restrict ranked results only; entity cards are not language-filtered.
>
> **highlight** `object`
>
> Highlight extraction settings for ranked results.
>
> **start\_time** `string`
>
> ISO 8601 start time for ranked results only; overrides time\_range.
>
> **time\_basis** `string ("auto" | "published" | "crawled")`
>
> Timestamp used to time-filter ranked results only.
>
> **time\_range** `string ("day" | "week" | "month" | "year" | "d" | "w" | "m" | "y")`
>
> Relative window for ranked results only. Explicit start\_time or end\_time takes precedence.
>
> **full\_content** `object`
>
> Full-page content settings for ranked results. Content beyond the first 10 results costs \$0.50 per 1,000 results.
>
> **exclude\_domains** `array`
>
> Domains to exclude from ranked results only. Exclusions take precedence; entity cards are not domain-filtered.
>
> **include\_domains** `array`
>
> Domains to include in ranked results only; entity cards are not domain-filtered.
>
> #### Output
>
> **data** `string` *required*
>
> Data from the action execution
>
> **error** `string`
>
> Error if any occurred during the execution of the action
>
> **successful** `boolean` *required*
>
> Whether or not the action execution was successful or not

> [accordion] Search Images
>
> **Slug:** `OCTEN_SEARCH_IMAGES`
>
> Search Octen's web image index from exactly one text query or reference image. This paid action costs \$2 per 1,000 image-search calls. It sends one billable request with no automatic retries; do not replay an ambiguous failure because the search may be charged again.
>
> #### Input Parameters
>
> **count** `integer`
>
> Number of image results to return (1-10).
>
> **inputs** `array` *required*
>
> Exactly one text query or one URL/base64 image input.
>
> **safesearch** `string ("off" | "strict")`
>
> Explicit-content filtering: strict removes adult content; off disables filtering.
>
> **exclude\_domains** `array`
>
> Domains to exclude, matched against each result's source\_page URL.
>
> **include\_domains** `array`
>
> Domains to include, matched against each result's source\_page URL.
>
> #### Output
>
> **data** `string` *required*
>
> Data from the action execution
>
> **error** `string`
>
> Error if any occurred during the execution of the action
>
> **successful** `boolean` *required*
>
> Whether or not the action execution was successful or not

> [accordion] Search News
>
> **Slug:** `OCTEN_SEARCH_NEWS`
>
> Search live news once for ranked articles and optional event subjects. News Search costs \$3 per 1,000 calls; full content beyond the first 10 results per search costs \$0.50 per 1,000 results. The native usage envelope reports text and subject search counts. This metered action makes one request and does not automatically retry ambiguous failures.
>
> #### Input Parameters
>
> **count** `integer`
>
> Maximum flat news results returned (1-100); does not limit subjects.
>
> **query** `string` *required*
>
> News search query. Supports site:\<domain> and -site:\<domain> operators; those domain filters affect results only, not subjects.
>
> **end\_time** `string`
>
> ISO 8601 upper time bound for articles and subjects; explicit bounds take precedence over time\_range.
>
> **language** `array`
>
> ISO 639-1 language codes restricting both results and subjects.
>
> **subjects** `object`
>
> Subject grouping settings. Omission uses Octen's defaults: enabled, up to 2 subjects, and up to 5 sub-events per subject.
>
> **highlight** `object`
>
> Highlight settings; when omitted, highlighting defaults on at 512 tokens.
>
> **start\_time** `string`
>
> ISO 8601 lower time bound for articles and subjects; explicit bounds take precedence over time\_range.
>
> **time\_basis** `string ("auto" | "published" | "crawled")`
>
> Timestamp used to filter articles and subjects. Published uses time\_published; crawled uses time\_last\_crawled.
>
> **time\_range** `string ("day" | "week" | "month" | "year" | "d" | "w" | "m" | "y")`
>
> Relative window for articles and subjects. Explicit start\_time or end\_time takes precedence when supplied.
>
> **full\_content** `object`
>
> Full-content settings; omitted defaults off. When enabled, content after the first 10 results per search costs \$0.50 per 1,000 results.
>
> **exclude\_domains** `array`
>
> Domains to exclude from flat results only (at most 1,200). Exclusions take precedence over inclusions and do not affect subjects.
>
> **include\_domains** `array`
>
> Domains to include in flat results only (at most 1,200); subject grouping does not support domain filters.
>
> #### Output
>
> **data** `string` *required*
>
> Data from the action execution
>
> **error** `string`
>
> Error if any occurred during the execution of the action
>
> **successful** `boolean` *required*
>
> Whether or not the action execution was successful or not

> [accordion] Search Videos
>
> **Slug:** `OCTEN_SEARCH_VIDEOS`
>
> INVITE-ONLY BETA: search the web for videos only after confirming the Octen account has Video Search entitlement. Each call costs \$5 per 1,000 calls. The action sends one metered request and does not retry automatically after an ambiguous response.
>
> #### Input Parameters
>
> **count** `integer`
>
> Number of video results to return (1-10). Invite-only beta: an entitled Octen account is required. Video Search costs \$5 per 1,000 calls, and this action does not retry automatically.
>
> **inputs** `array` *required*
>
> Required array containing at most one text input. Invite-only beta: an entitled Octen account is required. Video Search costs \$5 per 1,000 calls, and this action does not retry automatically.
>
> **end\_time** `string`
>
> ISO 8601 publish-time upper bound; takes precedence over time\_range. Invite-only beta: an entitled Octen account is required. Video Search costs \$5 per 1,000 calls, and this action does not retry automatically.
>
> **safesearch** `string ("off" | "strict")`
>
> Explicit-content filtering: strict drops all adult content; off disables filtering. Invite-only beta: an entitled Octen account is required. Video Search costs \$5 per 1,000 calls, and this action does not retry automatically.
>
> **start\_time** `string`
>
> ISO 8601 publish-time lower bound; takes precedence over time\_range. Invite-only beta: an entitled Octen account is required. Video Search costs \$5 per 1,000 calls, and this action does not retry automatically.
>
> **time\_range** `string ("day" | "week" | "month" | "year" | "d" | "w" | "m" | "y")`
>
> Relative publish-time window. start\_time or end\_time takes precedence when supplied. Invite-only beta: an entitled Octen account is required. Video Search costs \$5 per 1,000 calls, and this action does not retry automatically.
>
> #### Output
>
> **data** `string` *required*
>
> Data from the action execution
>
> **error** `string`
>
> Error if any occurred during the execution of the action
>
> **successful** `boolean` *required*
>
> Whether or not the action execution was successful or not

> [accordion] Search Web
>
> **Slug:** `OCTEN_SEARCH_WEB`
>
> Search Octen's live web index once for ranked pages, with bounded filters, highlights, optional full content, and usage metadata. This metered call has no pagination; do not retry automatically after an ambiguous response.
>
> #### Input Parameters
>
> **count** `integer`
>
> Maximum results returned for each search query (1-100).
>
> **query** `string` *required*
>
> Search query for Octen's live web index; supports site:\<domain> and -site:\<domain> operators.
>
> **format** `string ("markdown" | "text")`
>
> Formatting used for returned highlights.
>
> **end\_time** `string`
>
> Inclusive ISO 8601 upper bound; takes precedence over time\_range.
>
> **language** `array`
>
> ISO 639-1 language codes used to restrict results.
>
> **highlight** `object`
>
> Highlight extraction settings.
>
> **safesearch** `string ("off" | "strict")`
>
> Safe-search filtering mode.
>
> **start\_time** `string`
>
> Inclusive ISO 8601 lower bound; takes precedence over time\_range.
>
> **time\_basis** `string ("auto" | "published" | "crawled")`
>
> Timestamp Octen uses for time filtering.
>
> **time\_range** `string ("day" | "week" | "month" | "year" | "d" | "w" | "m" | "y")`
>
> Relative time window. Explicit time bounds take precedence.
>
> **exclude\_text** `array`
>
> Strings that must not appear in result page text (at most 5).
>
> **full\_content** `object`
>
> Optional full-page content settings.
>
> **include\_text** `array`
>
> Strings that must appear in result page text (at most 5).
>
> **include\_images** `boolean`
>
> Include available image metadata in each result.
>
> **exclude\_domains** `array`
>
> Domains to exclude. Exclusions take precedence over inclusions.
>
> **include\_domains** `array`
>
> Domains to include. Exclusions take precedence over inclusions.
>
> #### Output
>
> **data** `string` *required*
>
> Data from the action execution
>
> **error** `string`
>
> Error if any occurred during the execution of the action
>
> **successful** `boolean` *required*
>
> Whether or not the action execution was successful or not
