Data enrichment

Lasso Enrich takes partial product records—as little as a product name or SKU—and researches the target fields in your schema. A field remains unchanged when the worker cannot determine a value. With web search enabled, returned researched fields include basis metadata.

How it works

Pass in what you have — Provide one or more items with whatever data you already know (name, SKU, brand, etc.).
Define the target schema — Tell Lasso what fields you want filled in (or use the default product schema).
Lasso researches each product — The AI uses web tools when web_search is enabled and structures the information it finds.
Receive the available values — Successful researched fields include citations, a reasoning summary, and a confidence label. Fields the worker cannot determine stay unchanged.

The basis

When web search is enabled, each target field receives a basis entry that contains:

citations — The sources used (URL, title, excerpt)
reasoning — How the AI determined the value
confidence — high, medium, or low

This transparency is critical for product data quality. You can verify values, filter by confidence, or show sources to your users.

Schema options

Same as Search — use schema_id, inline columns, or omit both for the default product schema.

Sync vs async

Sync (default): For up to 5 items, Lasso processes and returns results in the same request.
Async: Pass webhook_url for 6 to 50 items or whenever you want webhook delivery. Lasso returns 202 immediately.

Do not send more than 5 items without webhook_url. Although the current handler returns 202, it exposes no public route for retrieving that job’s result.

Example: Search then Enrich

A common pattern is to search for products first, then enrich the results with additional detail.

// 1. Search for products
const search = await client.search({
  query: "best wireless mice for gaming 2025",
  columns: [
    { key: "name", label: "Name", type: "text" },
    { key: "brand", label: "Brand", type: "text" },
  ],
  max_results: 5,
});

// 2. Enrich with full detail
const enriched = await client.enrich({
  items: search.results.map(r => ({ data: r.data })),
  columns: [
    { key: "name", label: "Name", type: "text" },
    { key: "brand", label: "Brand", type: "text" },
    { key: "price", label: "Price", type: "number" },
    { key: "dpi", label: "Max DPI", type: "number" },
    { key: "weight", label: "Weight (g)", type: "number" },
    { key: "features", label: "Features", type: "tags" },
    { key: "description", label: "Description", type: "text" },
  ],
});

for (const item of enriched.items) {
  console.log(`${item.data.name} — ${item.data.weight} g, ${item.data.dpi} DPI`);
}

# 1. Search for products
search = client.search(
    query="best wireless mice for gaming 2025",
    columns=[
        {"key": "name", "label": "Name", "type": "text"},
        {"key": "brand", "label": "Brand", "type": "text"},
    ],
    max_results=5,
)

# 2. Enrich with full detail
enriched = client.enrich(
    items=[{"data": r["data"]} for r in search["results"]],
    columns=[
        {"key": "name", "label": "Name", "type": "text"},
        {"key": "brand", "label": "Brand", "type": "text"},
        {"key": "price", "label": "Price", "type": "number"},
        {"key": "dpi", "label": "Max DPI", "type": "number"},
        {"key": "weight", "label": "Weight (g)", "type": "number"},
        {"key": "features", "label": "Features", "type": "tags"},
        {"key": "description", "label": "Description", "type": "text"},
    ],
)

for item in enriched["items"]:
    print(f"{item['data']['name']} — {item['data']['weight']} g, {item['data']['dpi']} DPI")

Credits

Credit cost depends on the thinking level you choose:

Thinking	Credits/item	Best for
`hard`	4	Complex or ambiguous products
`medium`	2	Balanced speed and quality (default)
`low`	1	Fast, straightforward lookups

The service controls the model behind each thinking level. The legacy model input is ignored; treat the thinking level and credit price as the public contract.

Web search

By default, Lasso uses web search and returns basis metadata for target fields. With web_search: false, the worker uses model knowledge and omits basis; verify those uncited values independently.

Next steps

Enrich API reference — Full parameter and response documentation.
Search — Discover products to enrich.
Glossary — Control terminology during enrichment with use_glossary: true.

Getting started

TypeScript

Python

Learn

Data enrichment

How it works

The basis

Schema options

Sync vs async

Example: Search then Enrich

Credits

Web search

Next steps

​How it works

​The basis

​Schema options

​Sync vs async

​Example: Search then Enrich

​Credits

​Web search

​Next steps

How it works

The basis

Schema options

Sync vs async

Example: Search then Enrich

Credits

Web search

Next steps