Make an expert forecast for a news event.
Changed onMake an expert forecast for a news event.
This endpoint reaches into the news archive, looking back lookback days to extract the most relevant news articles, building a timeline of events, and then making an expert forecast.
This endpoint is more expensive than the search endpoint, it is calling gpt-4o or claude 3-5 on approx 15k tokens to build the forecast. This endpoint counts toward "deep" calls in the billing system.
It returns the forecast, the reasoning, and the sources.
Request
- Base URLhttps://api.asknews.app
- AuthAPI key in header Authorization
- TypeScriptA function with typed inputs. Its response is checked against the schema.
Query parameters
The requested forecast.
The number of days to look back for the forecast.
The total number of relevant articles to be extracted from the news archive and used for the forecast.
Method to use for the context search Currently only 'kw' is supported.
The model to use for the forecast.
The cutoff date for the forecast. String format is 'YYYY-MM-DD-HH:MM'. This is useful for backtesting forecasts.
Whether to use Reddit data for the forecast.enterprise customers only.
Additional context to use for the forecast.
Whether to run a live web search and include results in the forecast. enterprise customers only.
The type of expert to use for the forecast.
Response
A fully structured Forecast, complete with research.
Changes
- info
added the optional property
//////to the response with the status
- info
- breaking
the response property
//became optional for the status - breaking
the response property
//became optional for the status - breaking
response property
sources/items/centroid_distancelist-of-types was widened by adding typesnullto media typeapplication/jsonof response200 - breaking
response property
sources/items/cluster_probabilitylist-of-types was widened by adding typesnullto media typeapplication/jsonof response200
- breaking