Helicone Query Language (HQL)
AI model · Language model (LLM) · Text → Text by Helicone
Helicone Query Language (HQL) is a SQL-based interface for querying Helicone analytics data with row-level security and built-in limits. It is designed for developers and data teams who need to perform deep analysis on LLM request logs, including cost metrics, token usage, and custom properties. Users can execute queries directly within the Helicone dashboard or programmatically via a REST API to build custom reports or export data. The language supports standard SQL operations like filtering by timestamp, grouping by model, and calculating error rates. HQL is available to selected workspaces,
- Maximum rows per query
- 300,000
- Query timeout
- 30 seconds
- API rate limit
- 100 queries/min
- CSV download limit
- 10 downloads/min
Overview
Helicone Query Language (HQL) is a SQL-based interface for querying Helicone analytics data with row-level security and built-in limits. It is designed for developers and data teams who need to perform deep analysis on LLM request logs, including cost metrics, token usage, and custom properties. Users can execute queries directly within the Helicone dashboard or programmatically via a REST API to build custom reports or export data. The language supports standard SQL operations like filtering by timestamp, grouping by model, and calculating error rates. HQL is available to selected workspaces, providing a powerful tool for understanding unit economics and application performance.
Query Capabilities
- SQL-based Querying: The language allows users to query request logs, response bodies, status codes, and custom properties using standard SQL syntax.
- Analytics Queries: Users can analyze active users by day, error rates over specific time windows, and top costly requests.
- Dashboard and API Access: Queries can be executed directly in the Helicone dashboard editor or programmatically via a REST API.
- Data Export: Results can be downloaded as CSV files through a dedicated API endpoint.
- Saved Queries: Queries can be saved, revisited, and shared within an organization.
Limits and Specifications
- Row Limit: Queries are limited to a maximum of 300,000 rows per execution.
- Timeout Limit: Each query has a maximum timeout of 30 seconds.
- Rate Limits: The API supports up to 100 queries per minute and 10 CSV downloads per minute.
- Precision Storage: Cost values are stored multiplied by one billion for precision, requiring division by 1,000,000,000 for USD values.
Specifications
- Maximum rows per query
- 300,000
- Query timeout
- 30 seconds
- API rate limit
- 100 queries/min
- CSV download limit
- 10 downloads/min
Added by AntSpace from public sources (docs.helicone.ai).