fetch

ChatGPT

Tool: fetch (Factiva) Purpose Retrieves full article content using article IDs returned by search. What the tool does Fetches complete article text and full metadata for analysis and citation. Establishes the source identity (id, title, url) that should be referenced when citing information retrieved via fetch. When to use Use fetch to: Analyze article content in detail Extract figures, statistics, and reported findings Support data-backed summaries with inline citations Input requirements One valid article ID returned by the search tool Constraints Do not fetch the same article ID more than once in a session. If content has already been retrieved, reuse it rather than calling fetch again.

search

ChatGPT

Purpose Discovers relevant articles from Factiva to be used as citable sources for answering user questions, without retrieving full content What the tool does Searches Factiva for articles relevant to the user's query. Returns a ranked list of results with stable identifiers and source metadata only. Establishes the source identity (id, title, url) that should be referenced when citing information retrieved via fetch. When to use Use search to: Find the most relevant articles for a user's request. Build a shortlist of sources to cite in the final response. Collect article ids to retrieve content via fetch for the selected results. Narrow scope before requesting full text to reduce unnecessary retrieval. daysRange (optional query parameter) Limits results to a specific publication time window. Allowed values: LastDay, Last2Days, LastWeek, Last2Weeks, LastMonth, Last2Months, Last3Months (default), Last6Months, LastYear, Last2Years, Last5Years, AllDates. When to use Select the most appropriate value based on the user’s temporal intent. Use shorter ranges for breaking news or recent developments. Use medium ranges for ongoing trends or recent analysis. Use longer ranges for historical or background research. Default to Last3Months when no clear time constraint is implied. The LLM should select the most appropriate value based on the user’s temporal intent. Use shorter ranges for breaking news or recent developments, medium ranges for ongoing trends, and longer ranges for historical or background analysis. Default to Last3Months if no clear time intent is present. Time handling behavior When a user specifies a time range (e.g., “for the last 5 years”), the LLM must map that constraint to the daysRange query parameter. The original query text must not be expanded, rewritten, or enriched with inferred years, date ranges, or additional time-related keywords. Do not transform: “NVIDIA trends for the last 5 years” Into: “NVIDIA trends over the last 5 years revenue growth AI chips data center gaming stock performance strategic developments 2021 2022 2023 2024 2025” Instead: search(query="NVIDIA trends", daysRange="Last5Years") Time constraints belong exclusively in daysRange, not in the query string. Returned fields Each result includes: id — unique Factiva article identifier (stable handle for follow-on fetch) title — article headline url — canonical source link (for user-facing attribution/citations) Constraints / Notes This tool does not return the full article body text. The id/title/url returned by search should be treated as the citation anchors for any information drawn from the corresponding fetched article content. Keep results lightweight; return only metadata necessary for selection and attribution. Example usage pattern search → get a list of candidate sources (results[] with id/title/url) fetch → retrieve full text for the selected id(s) Respond → synthesize an answer and cite the relevant source url(s)/titles associated with those id(s)

Dow Jones Factiva

Overview

Tools

fetch

search