Skip to main content
Use AI models to research companies, extract insights, and generate personalized content. 
const summary = await canvas.ai.research({
  prompt: "Summarize this company in one sentence",
  data: company,
});

​
Parameters

​
prompt
string
required
Instructions for what you want the AI to research or generate.
​
data
Record<string, unknown>
required
Context data for the AI to analyze.
​
model
string
default:"sonar-pro"
AI model to use:
  • sonar-pro - Web search (default)
  • sonar-deep-research - Thorough research
  • gemini-2.5-pro - Google search
  • gpt-4o - Fast, no search
  • o4-mini-deep-research - Deep research
​
resultType
string
default:"Text"
Output format: Text, Number, Option, StructuredOutput
​
resultOptions
string[]
For Option type: valid choices to pick from.
​
resultSchema
object
For StructuredOutput type: define fields with types.

​
Result Types

​
Text

Free-form response. Use for summaries, research, content. 
const signals = await canvas.ai.research({
  prompt: "Find recent news: funding, hiring, or product launches",
  data: { name: "Stripe", website: "stripe.com" },
  model: "sonar-deep-research",
});
// "Stripe announced $6.5B funding in March 2023..."

​
Number

Numeric value. Use for scoring, estimating, counting. 
const score = await canvas.ai.research({
  prompt: `Score 1-100 for fit with our ICP:
    - B2B SaaS company
    - 50-500 employees  
    - Has a sales team`,
  data: company,
  resultType: "Number",
});
// 85

​
Option

Pick from predefined choices. Use for categorization, qualification. 
const tier = await canvas.ai.research({
  prompt: `Qualify this lead:
    HOT = Decision maker, clear need
    WARM = Right profile, unclear intent
    COLD = Poor fit`,
  data: lead,
  resultType: "Option",
  resultOptions: ["HOT", "WARM", "COLD"],
});
// "HOT"

​
StructuredOutput

Extract multiple typed fields. Use for complex analysis. 
const analysis = await canvas.ai.research({
  prompt: "Analyze for sales outreach",
  data: lead,
  resultType: "StructuredOutput",
  resultSchema: {
    fields: [
      { name: "qualified", type: "boolean" },
      { name: "score", type: "number" },
      { name: "pain_points", type: "string" },
      { name: "opener", type: "string" },
    ],
  },
});
// { qualified: true, score: 85, pain_points: "Manual lead routing", opener: "Saw you just expanded to 15 reps..." }

​
Full Example

Combine result types in a workflow: 
// 1. Score leads
const score = await canvas.ai.research({
  prompt: "Score 1-100 based on company size, industry fit, and title seniority",
  data: lead,
  resultType: "Number",
});

// 2. Qualify high scorers
if (score >= 70) {
  const analysis = await canvas.ai.research({
    prompt: "Generate personalized outreach",
    data: lead,
    model: "sonar-pro",
    resultType: "StructuredOutput",
    resultSchema: {
      fields: [
        { name: "subject", type: "string" },
        { name: "opener", type: "string" },
        { name: "cta", type: "string" },
      ],
    },
  });
}