Skip to main content

Media & Web

Give agents eyes and ears on the outside world: search the live web, scrape pages, fetch files and images into the conversation, source stock media, and turn audio and video into text.

None of these types require a connector. The live search and scrape tools are metered per company; outputs that produce files land in Autessa-managed S3 storage and come back as short-lived presigned or tiny URLs.

Extract Audio TOOLEXTRACT_AUDIO​

Downloads the media at videoFileUrl (presigned URL, https://, or s3:// URI), transcodes its audio track to WAV, and uploads the result to Autessa storage under your company. The defaults — 16 kHz mono — match what speech-to-text expects, making this the natural first step before Transcribe. Currently the output is always WAV regardless of the requested outputFormat.

The result contains {"s3Uri": ..., "presignedUrl": ..., "bytes": ..., "outputFormat": ...}; the presigned URL is valid for 1 hour. Nothing is written to AutessaDB — chain an AutessaDB update if you want the file recorded on a row.

Example​

Tool config
{
  "toolType": "EXTRACT_AUDIO",
  "videoFileUrl": "@@{VIDEO_FILE_URL}",
  "outputFormat": "WAV",
  "sampleRateHz": 16000,
  "channels": 1
}

Configuration fields

  • Name
    videoFileUrl
    Type
    string
    Description
    Presigned URL / s3:// URI / https URL of the source media. Required.
  • Name
    outputFormat
    Type
    string
    Description
    "WAV" or "MP3". Defaults to WAV.
  • Name
    sampleRateHz
    Type
    integer
    Description
    Sample rate in Hz. Defaults to 16000 — matches STT expectations.
  • Name
    channels
    Type
    integer
    Description
    Channel count. Defaults to 1 (mono).

Transcribe TOOLTRANSCRIBE​

Runs Autessa's long-form speech-to-text pipeline over the audio at audioFileUrl and returns the transcript inline — no files are written anywhere. An optional languageHint (ISO code such as en or es) can steer recognition.

The result is {"transcriptText": ..., "chars": ...} plus an isSilent flag when the audio contained no speech.

Example​

Tool config
{
  "toolType": "TRANSCRIBE",
  "audioFileUrl": "@@{AUDIO_FILE_URL}",
  "languageHint": "en"
}

Configuration fields

  • Name
    audioFileUrl
    Type
    string
    Description
    s3:// URI of a file in the caller's workspace, or an http(s) link; any common audio/video format. Required.
  • Name
    languageHint
    Type
    string · optional
    Description
    Optional ISO language hint (e.g., "en", "es").

Stock Image Generator TOOLSTOCK_IMAGE_GENERATOR​

Searches Unsplash and Pexels in parallel for stock photos matching query and returns direct image URLs from both providers. numberOfResults is capped at 5 per provider. With returnTinyUrl enabled, each URL is wrapped in a short Autessa link (valid 1 hour) that is easier to embed in emails or AutessaDB rows.

The result is {"unsplash_images": [...], "pexels_images": [...]} — arrays of URLs. A provider that errors contributes an empty array rather than failing the call.

Example​

Tool config
{
  "toolType": "STOCK_IMAGE_GENERATOR",
  "query": "@@{IMAGE_SEARCH_QUERY}",
  "numberOfResults": 5,
  "returnTinyUrl": true
}

Configuration fields

  • Name
    query
    Type
    string
    Description
    Required string value.
  • Name
    numberOfResults
    Type
    integer
    Description
    Required integer value.
  • Name
    returnTinyUrl
    Type
    boolean
    Description
    Required boolean value.

Stock Video Generator TOOLSTOCK_VIDEO_GENERATOR​

Searches Pexels for stock video clips matching query, preferring HD renditions when available. numberOfResults is capped at 5, and returnTinyUrl works the same as for stock images (1-hour short links).

The result is {"pexels_videos": [...]} — an array of direct video URLs.

Example​

Tool config
{
  "toolType": "STOCK_VIDEO_GENERATOR",
  "query": "@@{VIDEO_SEARCH_QUERY}",
  "numberOfResults": 3,
  "returnTinyUrl": false
}

Configuration fields

  • Name
    query
    Type
    string
    Description
    Required string value.
  • Name
    numberOfResults
    Type
    integer
    Description
    Required integer value.
  • Name
    returnTinyUrl
    Type
    boolean
    Description
    Required boolean value.

Live Search Engine TOOLLIVE_SEARCH_ENGINE​

Runs a live web search through Autessa's managed browser service and returns the raw search results for the agent to read. numResults defaults to 5. Searches count against your company's browser-search limit; when the limit is hit the call retries briefly, then returns a rate-limit error with the remaining quota.

The result is {"statusCode": ..., "body": ..., "headers": {}} where body carries the search results as returned by the search backend.

Example​

Tool config
{
  "toolType": "LIVE_SEARCH_ENGINE",
  "query": "@@{SEARCH_QUERY}",
  "numResults": 5
}

Configuration fields

  • Name
    query
    Type
    string
    Description
    Required string value.
  • Name
    numResults
    Type
    integer
    Description
    Required integer value.

Scrape URL TOOLSCRAPE_URL​

Fetches a public web page through the managed browser service and returns its content. Set waitForJs to true for JavaScript-heavy pages — the browser renders the page before extracting content, and JS scrapes are metered in a separate (more expensive) quota bucket than plain fetches. Every URL is validated before the request is sent; requests to private networks and internal infrastructure are rejected with a validation error.

The result is {"statusCode": ..., "body": ..., "headers": {}} with the scraped content in body.

Example​

Tool config
{
  "toolType": "SCRAPE_URL",
  "url": "@@{PAGE_URL}",
  "waitForJs": true
}

Configuration fields

  • Name
    url
    Type
    string
    Description
    Required string value.
  • Name
    waitForJs
    Type
    boolean
    Description
    Required boolean value.

Fetch and View TOOLFETCH_AND_VIEW​

Fetches an image or document from a URL and places it directly into the conversation as multimodal content, so the agent can actually look at it on its next turn — unlike Scrape URL, which returns page text. Supports https://, http://, and s3:// URIs (S3 objects are presigned for 1 hour), and contentType (image or file) is auto-detected from the extension when omitted.

The result confirms the fetch and carries an enrichment payload that the platform injects into the next model request as an image or file block: {"success": true, "message": ..., "enrichment": [...]}.

Example​

Tool config
{
  "toolType": "FETCH_AND_VIEW",
  "url": "@@{FILE_OR_IMAGE_URL}",
  "contentType": "image"
}

Configuration fields

  • Name
    url
    Type
    string
    Description
    The URL to fetch. Supports https://, http://, and s3:// URIs. External URLs must be from allowed domains.
  • Name
    contentType
    Type
    string · optional
    Description
    Optional content type hint: "image" or "file". If not provided, will be auto-detected from the URL extension.

Perplexity Search TOOLPERPLEXITY_SEARCH​

Queries the Perplexity search API, which returns search results with extracted page content — richer than plain result links. maxResults (default 10) controls how many results come back, maxTokens (default 25000) bounds the total content payload, and maxTokensPerPage (default 2048) bounds how much extracted content each page contributes.

The result is {"statusCode": ..., "body": ..., "headers": {...}} with Perplexity's response JSON in body.

Example​

Tool config
{
  "toolType": "PERPLEXITY_SEARCH",
  "query": "@@{RESEARCH_QUERY}",
  "maxResults": 10,
  "maxTokens": 25000,
  "maxTokensPerPage": 2048
}

Configuration fields

  • Name
    query
    Type
    string
    Description
    Required string value.
  • Name
    maxResults
    Type
    integer
    Description
    Required integer value.
  • Name
    maxTokens
    Type
    integer
    Description
    Required integer value.
  • Name
    maxTokensPerPage
    Type
    integer
    Description
    Required integer value.