bulk_upsert_reviewed_queries
ChatGPTStage 1 to 500 SQL reviewed queries on a draft in one batch without executing them. Questions matching existing reviewed queries use the existing replacement semantics.
Data & Analytics
Overview
WisdomAI is the enterprise data harness inside ChatGPT, for the complex analytical questions where the answer has to be deterministic, traceable, and consistent. Connectors provide access. Catalogs and semantic layers provide definitions. The WisdomAI harness applies both at runtime by selecting the right context for each question, enforcing row- and column-level security for the person asking, optimizing execution for the underlying platform, verifying results, and repairing failed queries. Business context strengthens continuously as WisdomAI learns from documents, query logs, catalogs, semantic layers, dbt, and BI metadata, proactively identifies where it lacked sufficient context, and routes proposed changes through expert review and versioning. Corrections become governed organizational knowledge rather than one-time fixes buried in individual conversations. WisdomAI works across data platforms, sources, and deployment environments, including fully on-premises for regulated data that cannot leave your environment. The harness, the context, and query execution stay behind your firewall; only the answer crosses. Every answer is grounded in governed definitions, enterprise data, and visible sources making it trustworthy enough to run your business on. Use WisdomAI in ChatGPT to: - Ask questions of governed business data: "How did Q3 net revenue track against plan in EMEA?" - Follow up in the same line of analysis: "Break that down by segment and show the last 4 quarters" - Investigate an open-ended question end to end: "We missed plan in EMEA last quarter. What drove it, which segments and accounts moved most, and what should we watch going into Q4?" - Trace an answer to its definitions and sources: "Which definition of net revenue was used, and what data did it come from?" - See what's available to you: "What data do I have access to, and which metrics are defined for finance?"
Tools
cancel_chat
ChatGPTRequests cancellation of the response currently being generated for a conversation. If no response is running, no action is taken. The conversationId must be one returned by WisdomAI.
create_domain_draft
ChatGPTCreates a domain-building draft and returns its draftId. Domain-building writes target a draft; changes stay on the draft and do not affect the published domain until the draft is published.
delete_domain_derived_column
ChatGPTDelete a derived column from a draft. Metrics are not affected.
delete_domain_knowledge
ChatGPTDelete a domain knowledge item from a draft.
delete_domain_metric
ChatGPTDelete a domain metric from a draft.
delete_domain_parameterized_table
ChatGPTDelete a parameterized table from a draft.
delete_domain_relationship
ChatGPTDelete the recorded direct relationship between two tables from a draft.
delete_domain_synonym_set
ChatGPTDelete a domain synonym set from a draft.
delete_reviewed_query
ChatGPTDelete a reviewed query from a draft.
delete_table_rls_settings
ChatGPTClear the optional row-level security setting from a table on a draft.
discard_domain_draft
ChatGPTDiscards an explicitly identified domain draft and all of its staged changes.
discard_domain_draft_deltas
ChatGPTDiscards selected staged changes from an explicitly identified draft. Delta indices refer to the current draft and shift after changes are discarded.
get_chart_data
ChatGPTReturns a page of data for a table, chart, or metric card artifact from a completed answer. Rows preserve the chart or table’s grouping and filters; they are not raw source records.
get_chat_result
ChatGPTChecks a conversation for updates. Returns recent progress while it is running; when it is no longer running, returns the completed answer, a clarification request, or a failure without starting another analytical turn. Every result carries the conversationId it belongs to, and completed answers carry a responseMessageId identifying the turn. status is one of: "running" — the turn has not finished; get_chat_result returns the result for this conversationId once it does. "needs_clarification" — WisdomAI needs more information to answer; the result carries the question, the available choices, and allows_free_text. The answer is a message on the same conversationId. "needs_authorization" — authorization required in WisdomAI. Any answer content is partial. "completed" — the result carries the analytical response and its supporting artifacts. "cancelled" — the response was cancelled before it completed. "failed" — the result carries error.code and error.message describing the failure.
get_domain_derived_column
ChatGPTGet one derived column by table ID and name. Metrics are not included.
get_domain_draft
ChatGPTRead a domain draft, including staged deltas and validation errors.
get_domain_entity
ChatGPTGet one domain entity by ID.
get_domain_knowledge
ChatGPTGet one domain knowledge item by ID.
get_domain_metric
ChatGPTGet one domain metric by name.
get_domain_parameterized_table
ChatGPTGet one parameterized table by ID, including its Jinja template, input parameters, and inferred columns.
get_domain_synonym_set
ChatGPTGet one domain synonym set by ID.
get_domain_table_descriptions
ChatGPTGet table and column descriptions for a domain table.
get_reviewed_query
ChatGPTGet one reviewed query by ID, including SQL and metadata.
get_table_rls_settings
ChatGPTGet row-level security settings for one table. Unset tables return configured false.
list_domain_derived_columns
ChatGPTList derived columns from the published domain or draft-materialized view. Metrics are not included.
list_domain_entities
ChatGPTList domain entities from the published domain or draft-materialized view.
list_domain_knowledge
ChatGPTList domain knowledge previews from the published domain or draft-materialized view. Previews omit the writable full content, which is available from get_domain_knowledge.
list_domain_metrics
ChatGPTList domain metrics from the published domain or draft-materialized view.
list_domain_parameterized_tables
ChatGPTList parameterized tables -- tables backed by a Jinja SQL template, such as a stored procedure call -- from the published domain or draft-materialized view. Previews omit the template and parameter details, which are available from get_domain_parameterized_table.
list_domain_relationships
ChatGPTList direct table relationships from the published domain or draft-materialized view.
list_domain_synonym_sets
ChatGPTList synonym sets from the published domain or draft-materialized view.
list_domain_tables
ChatGPTList domain tables from the published domain or draft-materialized view.
list_domains_v2
ChatGPTLists the domains and domain groups available to the user, grouped by organization
list_reviewed_queries
ChatGPTList reviewed query examples from the published domain or draft-materialized view.
list_table_rls_settings
ChatGPTList tables that have row-level security settings in the published domain or draft view.
new_chat
ChatGPTWisdomAI interprets the question using business knowledge and context that may not be available to the caller, and uses that understanding to carry out the analysis and return a complete answer. Starts a new conversation for a new user objective. send_message is preferred if an existing conversation on the same domain covers the user's topic. Every result carries the conversationId it belongs to, and completed answers carry a responseMessageId identifying the turn. status is one of: "running" — the turn has not finished; get_chat_result returns the result for this conversationId once it does. "needs_clarification" — WisdomAI needs more information to answer; the result carries the question, the available choices, and allows_free_text. The answer is a message on the same conversationId. "needs_authorization" — authorization required in WisdomAI. Any answer content is partial. "completed" — the result carries the analytical response and its supporting artifacts. "cancelled" — the response was cancelled before it completed. "failed" — the result carries error.code and error.message describing the failure.
publish_domain_draft
ChatGPTPublishes an explicitly identified domain draft. Validation failures are returned with the draft left open so they can be fixed and republished.
review_last_conversation_query
ChatGPTStages the query behind the latest assistant answer in a conversation onto a domain draft as a reusable reviewed-query example. Reviewed queries are curated domain examples that inform future answers. Staging one records the query; it does not analyze, validate or critique it.
send_message
ChatGPTContinue an existing conversation with the next message. Useful for follow-ups, related questions, refinements, and clarifications. Asking topic-related questions in the same conversation allows for sharing analytical context. Each message starts another analytical turn. Use new_chat if the question requires a different domain. The WisdomAI conversation retains its latest active chat mode. Every result carries the conversationId it belongs to, and completed answers carry a responseMessageId identifying the turn. status is one of: "running" — the turn has not finished; get_chat_result returns the result for this conversationId once it does. "needs_clarification" — WisdomAI needs more information to answer; the result carries the question, the available choices, and allows_free_text. The answer is a message on the same conversationId. "needs_authorization" — authorization required in WisdomAI. Any answer content is partial. "completed" — the result carries the analytical response and its supporting artifacts. "cancelled" — the response was cancelled before it completed. "failed" — the result carries error.code and error.message describing the failure.
trigger_domain_reindex
ChatGPTTriggers the existing domain data refresh and indexing path for the published domain. If an equivalent crawl is already active, its run ID is returned.
update_table_column_descriptions
ChatGPTStage table and column description updates on a draft.
upsert_domain_derived_column
ChatGPTCreate or update a SQL derived column on a draft. Metrics are managed by the metric tools.
upsert_domain_knowledge
ChatGPTCreate or update a domain knowledge item on a draft.
upsert_domain_metric
ChatGPTCreate or update a domain metric on a draft.
upsert_domain_parameterized_table
ChatGPTCreate or update a parameterized table on a draft. Staging runs the template once with each parameter's sample value to infer the table's columns, so the sample values must return rows.
upsert_domain_relationship
ChatGPTAdd or atomically replace the direct relationship between two tables on a draft.
upsert_domain_synonym_set
ChatGPTCreate or fully replace a domain synonym set on a draft.
upsert_table_rls_settings
ChatGPTCreate or replace row-level security settings for a table on a draft.
Capabilities
Writes
Example Prompts
Click any prompt to copy it.
Links
App Stats
49
Tools
3
Prompts
ChatGPT
Platforms
Category
Data ExplorationWorks with
ChatGPT
Data refreshed daily