Skip to main content
Collections are org-scoped saved source sets. Bundle the datasets, domains, and URLs you reach for repeatedly, then reference them by name with the collection:<name> prefix in included_sources. Create once, reuse across every search in your organization.
Collections are in beta. Create and manage them at platform.valyu.ai/user/collections.
Collections vs presets: presets (academic, finance, …) are Valyu-curated and available to everyone. Collections are bundles you define and save for your org.

​
Create a collection

  1. Go to Collections
  2. Click Create Collection
  3. Add sources (Valyu datasets, domains, or URLs) - or use Generate with AI to suggest sources from a description of your research needs
  4. Save with a memorable name

​
Use a collection

Reference it with the collection: prefix. You can mix collections with individual sources in the same request: 
from valyu import Valyu

valyu = Valyu()
response = valyu.search(
    query="biotech company clinical trial results",
    included_sources=[
        "collection:medical-research",  # your saved collection
        "github.com/topics/artificial-intelligence",  # additional domain
        "valyu/valyu-patents",  # additional dataset
    ],
)

import { Valyu } from "valyu-js";

const valyu = new Valyu();
const response = await valyu.search("biotech company clinical trial results", {
  includedSources: [
    "collection:medical-research",
    "github.com/topics/artificial-intelligence",
    "valyu/valyu-patents",
  ],
});

curl -X POST https://api.valyu.ai/v1/search \
  -H "x-api-key: your-valyu-api-key" \
  -H "Content-Type: application/json" \
  -d '{
    "query": "biotech company clinical trial results",
    "included_sources": [
      "collection:medical-research",
      "github.com/topics/artificial-intelligence",
      "valyu/valyu-patents"
    ]
  }'

​
Visibility and permissions

  • All organization members can view and use any collection in their org.
  • Only the creator can edit or delete their collection.
  • Collections are scoped to your organization and not visible to other orgs.

​
Errors

If a collection doesn’t exist or isn’t accessible, the API returns a warning and continues with any other valid sources: 
{
  "success": true,
  "warnings": ["Collection 'unknown-collection' not found"],
  "results": [...]
}
If only an invalid collection was specified, you’ll get empty results with the warning.
Keep collections small and focused (medical-research, finance-sec-filings, tech-docs) rather than one catch-all, and use clear names so teammates know what each is for.

Source filtering

Filter by individual domains and datasets

Data sources

Every dataset id and preset