Exa Websets icon

Exa Websets

Create, manage, and query structured datasets from web sources using Exa Websets API

Actions28

Overview

The node "Exa Websets" enables users to enrich webset content using AI-powered analysis. Specifically, the Enrichment - Create operation allows you to perform various types of enrichment on a specified webset by providing an ID and configuring how the enrichment should be done.

This node is beneficial when you want to extract insights, classify content, summarize information, or apply custom AI-driven transformations to collections of web content (websets). For example, you could use it to:

  • Extract key facts and insights from articles in a webset.
  • Classify documents into categories automatically.
  • Summarize large volumes of text content for quick review.
  • Customize enrichment with specific prompts tailored to your needs.

Properties

Name Meaning
Webset ID The unique identifier of the webset you want to enrich. This is required to specify which dataset the enrichment applies to.
Enrichment Type The type of enrichment to perform. Options: Classification, Custom, Extract, Summarize.
Prompt A textual instruction or prompt guiding the enrichment process, e.g., "Extract key facts and insights from each article". Used mainly for custom or extract types to tailor the AI's output.
Configuration Collection of settings controlling the enrichment behavior:
• Batch Size: Number of items processed per batch (1-100).
• Include Original Content: Whether to include original data alongside enriched results.
• Include Source URL: Whether to add source URLs.
• Max Tokens: Maximum tokens in AI response (1-4000).
• Model: AI model to use (Claude-3-Haiku, Claude-3-Sonnet, GPT-3.5 Turbo, GPT-4).
• Output Format: Format of enriched output (CSV, JSON, Markdown, Text).
• System Message: Context message for the AI model.
• Temperature: Controls randomness in AI output (0-2).
Filter Items Filters to limit which items in the webset are enriched:
• Content Type: Only enrich items of this type (HTML, Text, PDF).
• Domain: Only enrich items from this domain.
• Min Score: Minimum relevance score threshold (0-1).
• Item IDs: Comma-separated list of specific item IDs to enrich.

Output

The node outputs JSON data containing the enriched content for the specified webset items. The structure depends on the chosen output format but generally includes:

  • Enriched data fields generated by the AI model according to the enrichment type and prompt.
  • Optionally, the original content and source URLs if configured.
  • Metadata about the enrichment process such as item identifiers and possibly scores.

If binary data output is supported (not explicitly shown in the provided code), it would represent files or attachments related to the enrichment results.

Dependencies

  • Requires an API key credential for authenticating with the Exa Websets API service.
  • Uses external AI models such as Claude and GPT variants via the Exa Websets API.
  • Node configuration must include the base URL https://api.exa.ai and appropriate headers for JSON communication.

Troubleshooting

  • Common issues:

    • Invalid or missing Webset ID will cause the operation to fail.
    • Providing an unsupported enrichment type or invalid prompt may result in errors or unexpected output.
    • Exceeding token limits or batch size constraints can cause API rejections.
    • Network or authentication failures with the Exa Websets API will prevent enrichment.

  • Error messages:

    • "Unknown resource": Occurs if the resource parameter is not set to a valid value like "enrichments".
    • API errors typically return messages indicating invalid parameters or authentication failure; ensure credentials and inputs are correct.
    • If the node throws errors related to batch processing, reduce batch size or check input data validity.

  • Resolutions:

    • Double-check all required parameters, especially Webset ID and enrichment type.
    • Validate prompt and configuration values against allowed ranges.
    • Ensure API credentials are correctly configured in n8n.
    • Use smaller batches and simpler prompts to isolate issues.

Links and References

Note: This summary is based solely on static analysis of the provided source code and property definitions without runtime execution.

Discussion