Search Endpoints

A Search Endpoint defines how SearchTweak talks to your search API.

What You Configure

• Name: human-readable label.
• Description: optional business context.
• URL: full API endpoint URL.
• Method: GET, POST, or PUT.
• Mapper Code: rules to extract documents from response payload.
• Custom Headers: optional key-value headers.
• Advanced: Multi-threading:
  • Auto: parallel snapshot jobs (snapshots-auto).
  • Single: serialized snapshot jobs (snapshots-single).

Practical Setup Checklist

1. Start from a real API call that already works outside SearchTweak.
2. Add minimal required mapper fields (id, name).
3. Validate with at least one query that returns results.
4. Add optional fields (image, brand, etc.) only after base mapping works.
5. Keep authentication headers stable and secret-safe.

Endpoint Behavior Notes

• Endpoint activation controls whether it can be used for new models.
• Existing models/evaluations are not automatically deleted when endpoint is deactivated.
• Deleting an endpoint may be blocked by unfinished evaluations linked through models.
• Endpoints list supports segmented status filter: All | Active | Archived.

Common Mistakes

• Missing id or name in mapper output.
• Wrong method/body/header combination (GET with body assumptions).
• Mapping path does not match real response shape.
• Over-aggressive parallelism against strict upstream rate limits.

When to Use Single Queue Mode

Use Single when upstream API has strict rate limits, unstable latency, or anti-burst behavior.

Next Step

After endpoint is stable, continue with Search Models.