generate_content
ChatGPTGenerate content for a record based on the type of content requested. Available conversation types are a cultivation plan (donor communication strategy), conversation starter,Phone call starter, a birthday email, emails about giving milestones, emails about donation, an event attendance email, a new employee congratulation email, pledge reminder email, event invitation email, gift thank you note.
get_constituent_details
ChatGPTReturns details about the constituent record such as name, email, phone, and address. Returns a response with default values if the constituent is not found.
get_event_details
ChatGPTReturns details about the event record
get_gift_details
ChatGPTReturns details about the gift such as gift amount and payment method. Returns an empty response if the gift is not found.
read_thread
ChatGPTReturns the prior conversation of one Blackbaud Unified Chat thread (identified by threadId from search_threads) as a readable transcript, so you can continue it. The transcript is prior turns between the user and a different assistant; tool activity is summarized and the internal tools are not callable here. Continue the conversation from where it left off.
search_data_fundraising_renxt_ask
ChatGPTQueries Raiser's Edge NXT fundraising data across constituents and gifts using one SQL query. ALWAYS use this tool for cross-record fundraising questions about who did or did not give, including: 'Which constituents have not donated to the Capital Campaign?', 'Who gave to Fund A but not Fund B?', 'Which donors have zero gifts this year?', and other inclusion, exclusion, or set-difference lists. Also use it for fundraising counts, totals, averages, percentages, rankings, grouped metrics, campaign progress, and filtered multi-record results. Prefer this tool over RENXT constituent lists, gift lists, record lookups, details, summaries, and paginated tools when answering requires comparing or calculating across gifts and constituents. Pass the user's natural-language question directly; do not resolve campaign, fund, appeal, or constituent IDs first. Use record tools only for a known individual record or a purpose-built workflow that directly provides the requested result. Use search_data_fundraising_renxt_research instead when the fundraising question requires several queries, iterative investigation, trend explanation, cohort analysis, or root-cause analysis. For a cross-product request, answer the fundraising portion here and combine it with the appropriate product tool for the remaining portion. Reads only; never writes.
search_data_fundraising_renxt_research
ChatGPTRaiser's Edge NXT (RENXT) multi-step analytics over the organization's fundraising data in the Blackbaud lakehouse. ALWAYS use this tool for an open-ended fundraising investigation that needs several SQL queries, comparisons, or iterative exploration within the fundraising analysis. Trigger on requests to compare periods or segments, explain why a fundraising metric changed, identify trends or patterns, analyze cohorts or retention, evaluate campaign, fund, appeal, or donor performance, find drivers or anomalies, or give an analytical overview, even when the user does not say research, search, data, analytics, lakehouse, query, or report. For campaign questions, use this tool when the user needs multi-query investigation or explanation; use search_data_fundraising_renxt_ask for a single campaign goal, amount raised or committed, percent achieved, remaining or variance to goal, or progress-against-target calculation. A compound or cross-product request alone does not require research: use the ask tool for its fundraising analytics portion and combine it with the appropriate product tool for other portions, such as FENXT budget or expense data. Prefer this over generic RENXT record lookup, detail, list, summary, or paginated tools for organization-wide, portfolio-wide, cross-record, or aggregate analysis. A purpose-built tool that directly returns the exact requested measure or performs the requested workflow may still take precedence. This tool builds a research plan, runs several SQL queries, iterates on the results, and returns a written report plus reusable structured result rows. Results are never truncated, so state how many rows you want for lists or rankings, such as 'top 10'; an unbounded list can return thousands of rows. Pass conversationId to continue a prior investigation. Reads only; never writes.
search_for_constituent_by_name
ChatGPTReturns the constituent details that matches the name provided. Returns an empty response if no constituents are not found.
search_for_event_by_name
ChatGPTReturns the event details that match the name provided
search_threads
ChatGPTFinds the current user's prior Blackbaud Unified Chat threads so you can resume one. Pass a query to match thread titles and summaries (e.g. "gala budget"); omit it for the most recent threads. To resume one specific conversation — e.g. "continue my last conversation" — omit the query and pass limit=1 to get just the most recent; for a specific topic, pass the query (with limit=1 for just the best match). When exactly one thread is returned it is the intended one: call read_thread with its threadId and continue it, without asking the user to choose. When several are returned they are genuinely ambiguous — show them and let the user pick. Each result has a title, a short summary, when it was last active, and, when available, a relevance score.
