​
Column Types

FindAll uses two types of columns:
  • Constraint: Boolean criteria that entities must satisfy (e.g., “is an AI company”)
  • Enrichment: Attributes to extract for matching entities (e.g., “CEO name”)

​
Constraint Results

Constraint results can have the following values:
  • yes: Entity clearly meets the criteria
  • not_sure: Insufficient information
  • no: Entity does not meet criteria
  • skipped: Constraint not evaluated
Each constraint result also includes:
  • confidence: Level of confidence in the result (high, medium, low)
  • reasoning: Detailed explanation of the decision
  • citations: URLs supporting the reasoning

​
Entity Scoring

Each result includes:
  • score: Score to rank the results, higher means better match for constraints
  • pages_read: Number of web pages analyzed for this entity
  • pages_considered: Total pages evaluated for this entity

​
Response Metadata

Poll responses include additional metadata:
  • pages_read: Total web pages analyzed across all entities
  • pages_considered: Total web pages evaluated across all entities
  • billing_metrics: Cost tracking information
    • cost_mode: Processor used (base or pro)
    • rows_processed: Number of entities processed
    • enrichment_cells: Number of enrichment fields populated
  • are_enrichments_active: Whether enrichments are still running
  • filters: Array of constraint definitions with status
  • enrichments: Array of enrichment definitions with status
Important: When polling, check both is_active AND are_enrichments_active to determine if the run is truly complete. A run is finished only when both fields are false.

​
Processing Steps

FindAll runs follow these steps:
  1. understanding_question: Parse and understand the natural language query
  2. gen_candidates_step: Generate potential entity candidates using Parallel Index
  3. filter_candidates_step: Process candidates using constraints and enrichments
Each step has a status field: running, done, or failed.