account_details
ChatGPTWho is connected: the user's name and email, the organizations they belong to with their role in each, and the team of each organization (member names and roles). Call it first to learn the organization ids. Teammates are listed by name and role, never by email. Owners and Admins also get a user_id for each teammate, which update_ticket_settings needs to assign a ticket to them; other roles get no ids. Data Requesters do not get a team list, as in the Clema app.
create_ticket
ChatGPTRaise a data request (ticket) with the IR team in Clema, as the connected user. Give a short title and the request in plain words, and, as the Clema app does, the purpose (the specific decision or deliverable it supports), the output format and the target audience; ask the user only for what is missing, in as few questions as possible. Before calling it, show the user the title and summary and get their OK. Every role can raise a ticket. The same request raised twice returns the existing ticket instead of a duplicate. It does not set a priority or assignee: use update_ticket_settings.
data_dictionary_search
ChatGPTFind variables in the IPEDS or College Scorecard data dictionary by name, title or description. Always do this before writing SQL so you use real column names and understand coded values.
data_query
ChatGPTRun a read-only SQL SELECT against the IPEDS or College Scorecard public dataset and get the rows. Look up variables with data_dictionary_search and tables with data_tables first, resolve schools with institution_search, and cite the tables and variables you used in your answer. Results are paged: when hasNextPage is true, ask for the next page or tighten the query with filters and aggregation instead of pulling many rows.
data_tables
ChatGPTList the tables of the IPEDS or College Scorecard dataset with row counts, or the columns of one table. IPEDS has many tables named per survey year (e.g. staff_new_hires_2013 ... _2025) and v_ views that combine years: prefer a v_ view when one fits, and check its columns here before querying. Use data_dictionary_search first to find which variables exist, then this to see where they live.
institution_search
ChatGPTLook up an institution by name or alias and get its unitid. Use it as unitid in IPEDS tables and as institution_id in College Scorecard tables (same number; a few institutions exist in only one dataset). Several institutions can match: confirm the right one with the user when it is ambiguous.
ticket_details
ChatGPTOne ticket in full: status, priority, due date, requester, assignee, feedback, the intake (what was requested, why, output format, audience), the activity timeline (who assigned, changed priority, closed or reopened, and when). It does not include the conversation messages. People are returned by name, never by email.
tickets_list
ChatGPTList the organization's tickets (data requests), newest activity first, with status, priority, assignee and requester. Filter by assignment (your inbox or unassigned), status and priority. Data Requesters cannot use tickets in Clema, so this refuses them.
update_ticket_settings
ChatGPTChange a ticket's status (close or reopen), priority, or assignee, following the same rules as the Clema app: owners and admins can assign any owner, admin or member and can unassign; members can only assign a ticket to themselves; Data Requesters cannot use tickets. Reopening a ticket assigns it to whoever reopens it, as in the app. Every field is optional but at least one is required. Nothing is changed if any part of the request is not allowed.