> ## Documentation Index > Fetch the complete documentation index at: https://docs.nimbleway.com/llms.txt > Use this file to discover all available pages before exploring further. # Dataset Building > One prompt in, a verified dataset out, every cell cited **Dataset Building** agents create structured datasets from data scattered across the web. Describe what you're looking for, define the row shape, and the agent discovers, verifies, and structures the results: no scraper to build, no site list to maintain, no deduplication pipeline. [Every cell arrives cited](/nimble-sdk/web-search-agents/trust). ## Example Request The `business-discovery` template is tuned for local-business datasets: verified addresses, phone numbers, and ratings: ```python Python theme={"system"} agent = nimble.agents.create(template="business-discovery") run = nimble.agents.runs.create( agent.id, input=( "Find all spas and salons in New York City, real addresses, " "phone numbers, and ratings from actual review sites." ), ) ``` ```typescript TypeScript theme={"system"} const agent = await nimble.agents.create({ template: "business-discovery" }); const run = await nimble.agents.runs.create(agent.id, { input: "Find all spas and salons in New York City, real addresses, " + "phone numbers, and ratings from actual review sites.", }); ``` ```go Go theme={"system"} agent, err := client.Agents.New(ctx, nimble.AgentNewParams{ Template: param.NewOpt("business-discovery"), }) run, err := client.Agents.Runs.New(ctx, agent.ID, nimble.AgentRunNewParams{ Input: "Find all spas and salons in New York City, real addresses, " + "phone numbers, and ratings from actual review sites.", }) ``` ```bash CLI theme={"system"} AGENT_ID=$(nimble agents create --template business-discovery | jq -r .id) nimble agents:runs create \ --agent-id "$AGENT_ID" \ --input "Find all spas and salons in New York City, real addresses, phone numbers, and ratings from actual review sites." ``` ```bash cURL theme={"system"} curl -X POST "https://sdk.nimbleway.com/v2/agents/$AGENT_ID/runs" \ --header 'Authorization: Bearer ' \ --header 'Content-Type: application/json' \ --data '{ "input": "Find all spas and salons in New York City, real addresses, phone numbers, and ratings from actual review sites." }' ``` ## Example Response The result is `type: "json"`: an array of rows matching the template's schema, each with a cited source: ```json theme={"system"} { "type": "json", "content": [ { "business_name": "Haven Spa", "address": "150 Mercer St, New York, NY 10012", "phone_number": "+1 212-555-0181", "website": "https://havenspanyc.com", "rating": "4.6", "review_count": "412", "category": "Day spa", "source_url": "https://www.yelp.com/biz/haven-spa-new-york" } ] } ``` Each value carries a per-path claim in the trust report: ```json theme={"system"} { "path": "$[0].phone_number", "confidence": "high", "reasoning": "Supported by 2 independent sources", "citations": [ { "url": "https://www.yelp.com/biz/haven-spa-new-york", "excerpts": ["(212) 555-0181"] }, { "url": "https://havenspanyc.com/contact", "excerpts": ["Call us: (212) 555-0181"] } ] } ``` The dataset gets a deterministic overall grade, the lower of its **cell fill rate** and its **per-claim trust ratio**, stated in the run's trust reasoning: *"High confidence - cells fill rate is 96% and trust per claim ratio is 91%"*. A verifiably empty cell (a business with no website) comes back as a cited `null` graded `high`: [verified absence](/nimble-sdk/web-search-agents/trust#trust-for-structured-outputs). ## Custom output schema Define your own row shape on the agent or per run. Filters in the input are treated as hard constraints. The `company-discovery` template's first goal is *"Treat user filters (location, vertical, size, funding) as hard constraints, never dilute to inflate volume."* Schemas must follow the hard limits below. Invalid schemas return **422** at request time. ```python Python theme={"system"} run = nimble.agents.runs.create( agent.id, input=( "Find SF tech companies that have adopted AI agents and have active " "engineering teams, potential CRM leads, explaining why each qualifies." ), output_schema={ "type": "object", "properties": { "leads": { "type": "array", "items": { "type": "object", "properties": { "company": {"type": "string"}, "why_qualified": {"type": "string"}, "engineering_signal": {"type": "string"}, }, }, } }, }, ) ``` ```typescript TypeScript theme={"system"} const run = await nimble.agents.runs.create(agent.id, { input: "Find SF tech companies that have adopted AI agents and have active " + "engineering teams, potential CRM leads, explaining why each qualifies.", output_schema: { type: "object", properties: { leads: { type: "array", items: { type: "object", properties: { company: { type: "string" }, why_qualified: { type: "string" }, engineering_signal: { type: "string" }, }, }, }, }, }, }); ``` ```go Go theme={"system"} run, err := client.Agents.Runs.New(ctx, agent.ID, nimble.AgentRunNewParams{ Input: "Find SF tech companies that have adopted AI agents and have active " + "engineering teams, potential CRM leads, explaining why each qualifies.", OutputSchema: map[string]any{ "type": "object", "properties": map[string]any{ "leads": map[string]any{ "type": "array", "items": map[string]any{ "type": "object", "properties": map[string]any{ "company": map[string]any{"type": "string"}, "why_qualified": map[string]any{"type": "string"}, "engineering_signal": map[string]any{"type": "string"}, }, }, }, }, }, }) ``` ```bash CLI theme={"system"} nimble agents:runs create \ --agent-id "$AGENT_ID" \ --input "Find SF tech companies that have adopted AI agents and have active engineering teams, potential CRM leads, explaining why each qualifies." \ --output-schema '{type: object, properties: {leads: {type: array, items: {type: object, properties: {company: {type: string}, why_qualified: {type: string}, engineering_signal: {type: string}}}}}}' ``` ```bash cURL theme={"system"} curl -X POST "https://sdk.nimbleway.com/v2/agents/$AGENT_ID/runs" \ --header 'Authorization: Bearer ' \ --header 'Content-Type: application/json' \ --data '{ "input": "Find SF tech companies that have adopted AI agents and have active engineering teams, potential CRM leads, explaining why each qualifies.", "output_schema": { "type": "object", "properties": { "leads": { "type": "array", "items": { "type": "object", "properties": { "company": {"type": "string"}, "why_qualified": {"type": "string"}, "engineering_signal": {"type": "string"} } } } } } }' ``` ## Output schema limits ### Allowed roots The root must be one of: * An **object** with non-empty `properties` * An **array of objects** (`items.type` must be `"object"` with `properties`) Omit `output_schema` (or pass `{}`) to let the generator infer the shape from the prompt. ### Supported examples **Object root with nullable fields:** ```json theme={"system"} { "type": "object", "properties": { "company_name": { "type": "string" }, "website": { "type": ["string", "null"] } } } ``` **Array root (table shape):** ```json theme={"system"} { "type": "array", "items": { "type": "object", "properties": { "company_name": { "type": "string" }, "hq_city": { "type": ["string", "null"] } } } } ``` **Nested array of objects:** ```json theme={"system"} { "type": "object", "properties": { "key_contacts": { "type": "array", "items": { "type": "object", "properties": { "name": { "type": "string" }, "email": { "type": ["string", "null"] } } } } } } ``` ### Rejected patterns The API returns **422** when a schema violates these hard limits: | Limit | Rule | | - | - | | Root shape | No scalar roots, empty objects, or root `anyOf` | | Array root | Must be an array of objects, not primitives | | Nullable fields | Use `{"type": ["string", "null"]}`, not standalone `{"type": "null"}` | | Nesting depth | At most **5** levels (objects and arrays count) | | Property count | At most **100** properties across all levels | | Enum size | At most **500** enum values total; enums with more than 250 values are also capped at 7,500 combined characters | | Unsupported keywords | `format`, `pattern`, `minLength`, `maxLength`, `minimum`, `maximum`, `multipleOf`, `minItems`, `maxItems`, `uniqueItems`, `minProperties`, `maxProperties`, `patternProperties`, `propertyNames`, `contains`, `minContains`, `maxContains`, `unevaluatedItems`, `unevaluatedProperties` | Move validation constraints into field `description` text instead of JSON Schema keywords. **Unsupported examples:** ```json theme={"system"} { "type": "array", "items": { "type": "string" } } ``` ```json theme={"system"} { "type": "object", "properties": { "email": { "type": "string", "format": "email" } } } ``` ```json theme={"system"} { "type": "object", "properties": { "x": { "type": "null" } } } ``` ```json theme={"system"} { "type": "object", "properties": { "name": { "type": "string" } }, "anyOf": [{ "required": ["name"] }] } ``` ### Still accepted These are **not** hard-failed: * `required` arrays on objects * `additionalProperties: false` * Nested `anyOf` / `oneOf` for nullability (not at the root) * Enums within the size limits above * Nested arrays of objects or primitives under an object root ### Field maps are invalid Each field must be a JSON Schema subschema, not a bare type string: ```json theme={"system"} { "properties": { "city": { "type": "string" } } } ``` Not: ```json theme={"system"} { "city": "string" } ``` ### Agent generation behavior When creating or updating a Web Search Agent, the design path can infer and sanitize a schema from your goals and template. The generator strips unsupported keywords and fixes common mistakes (such as standalone `null`) before validation. When you pass an explicit `output_schema` on create, update, or run, it is validated immediately. Fix the schema and retry on **422**. ### Common validation errors | Error fragment | Cause | | - | - | | `must not use "anyOf"` | Root-level `anyOf` | | `array of objects` | Root array whose `items` is not an object | | `standalone "null" type` | Property typed as `"null"` alone | | `unsupported keyword(s)` | A rejected keyword such as `format` or `pattern` | | `nesting depth` | More than 5 nested object/array levels | | `maximum of 100 properties` | Too many properties across the schema | | `non-empty "properties"` | Object root with missing or empty `properties` | | `declares nothing to write` | Field map syntax instead of proper subschemas | Example **422** response: ```json theme={"system"} { "detail": [ { "type": "value_error", "loc": ["body", "output_schema"], "msg": "Value error, output_schema uses unsupported keyword(s): format. Remove them (or move constraints into field descriptions).", "input": { "type": "object", "properties": { "email": { "type": "string", "format": "email" } } } } ] } ``` ## Run it on a schedule A dataset run produces a snapshot. To turn it into a recurring feed, point a [Job](/nimble-sdk/agentic/jobs) at the same agent with a list of inputs (one per market, category, or account), and Nimble runs it on a schedule and delivers results to S3, Databricks, or your warehouse. ## Templates | Template | Built for | | - | - | | `business-discovery` | Local businesses with verified contact details and ratings | | `company-discovery` | Companies matching hard filters: location, vertical, size, funding | | `gtm-lead-discovery` | Qualified sales leads with the reason each qualifies | | `open-positions-search` | Live job openings across companies and markets | | `social-media-monitor` | Structured social activity around brands and topics | Research, Enrichment, and Dataset Building side by side This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.