cancel_talk_to_data
ChatGPTCancel a running agent execution. Marks the run as cancelled. The run will stop processing and no result will be available. Can only cancel runs that are in "starting" or "running" status. Already completed, failed, or cancelled runs cannot be cancelled. Returns: confirmation of cancellation with final status
cancel_talk_to_data
ChatGPTCancel a running agent execution. Marks the run as cancelled. The run will stop processing and no result will be available. Can only cancel runs that are in "starting" or "running" status. Already completed, failed, or cancelled runs cannot be cancelled. Returns: confirmation of cancellation with final status
cancel_talk_to_data
ChatGPTCancel a running agent execution. Marks the run as cancelled. The run will stop processing and no result will be available. Can only cancel runs that are in "starting" or "running" status. Already completed, failed, or cancelled runs cannot be cancelled. Returns: confirmation of cancellation with final status
check_talk_to_data
ChatGPTRead-only: Check the status of a running agent execution. Returns current status, elapsed time, and progress information including: - status_message: Current operation being performed - thinking: Agent's working notes as it processes Poll every 10-15 seconds. Runs can take up to 15 minutes. Possible statuses: - "starting": Run is initializing - "running": Agent is processing (check status_message and thinking for progress) - "completed": Run finished successfully; use get_talk_to_data_result to retrieve the response - "failed": Run encountered an error (error field contains details) - "cancelled": Run was cancelled Returns: tracking_id, status, elapsed_seconds, and progress information. This tool only reads data and does not modify anything.
check_talk_to_data
ChatGPTRead-only: Check the status of a running agent execution. Only for runs started with start_talk_to_data (tracking_id). Not for QBR jobs use qbr_get_result with job_id from qbr_start. Returns current status, elapsed time, and progress information including: - status_message: Current operation being performed - thinking: Agent's working notes as it processes Poll every 10-15 seconds. Runs can take up to 15 minutes. Possible statuses: - "starting": Run is initializing - "running": Agent is processing (check status_message and thinking for progress) - "completed": Run finished successfully; use get_talk_to_data_result to retrieve the response - "failed": Run encountered an error (error field contains details) - "cancelled": Run was cancelled Returns: tracking_id, status, elapsed_seconds, and progress information. This tool only reads data and does not modify anything.
check_talk_to_data
ChatGPTRead-only: Check the status of a running agent execution. Only for runs started with start_talk_to_data (tracking_id). Not for QBR jobs use qbr_get_result with job_id from qbr_start. Returns current status, elapsed time, and progress information including: - status_message: Current operation being performed - thinking: Agent's working notes as it processes Poll every 10-15 seconds. Runs can take up to 15 minutes. Possible statuses: - "starting": Run is initializing - "running": Agent is processing (check status_message and thinking for progress) - "completed": Run finished successfully; use get_talk_to_data_result to retrieve the response - "failed": Run encountered an error (error field contains details) - "cancelled": Run was cancelled Returns: tracking_id, status, elapsed_seconds, and progress information. This tool only reads data and does not modify anything.
get_client
ChatGPTRead-only. Finds clients by fuzzy name search or by client ID. Use name to search (e.g. 'ClientA') or id to fetch a specific client. Provide exactly one of 'id' or 'name'. limit (optional, default 10, max 50) applies when searching by name. No data is modified.
get_client
ChatGPTRead-only. Finds clients by fuzzy name search or by client ID. Use name to search (e.g. 'ClientA') or id to fetch a specific client. Provide exactly one of 'id' or 'name'. limit (optional, default 10, max 50) applies when searching by name. No data is modified.
get_client
ChatGPTRead-only. Finds clients by fuzzy name search or by client ID. Use name to search (e.g. 'ClientA') or id to fetch a specific client. Provide exactly one of 'id' or 'name'. limit (optional, default 10, max 50) applies when searching by name. No data is modified.
get_current_user_account
ChatGPTRead-only. Returns the current user's account username and email address. No data is modified.
get_current_user_account
ChatGPTRead-only. Returns the current user's account username and email address. No data is modified.
get_current_user_account
ChatGPTRead-only. Returns the current user's account username and email address. No data is modified.
get_product
ChatGPTRead-only. Returns products by ID or name based on the provided filters. Optional filter: salesforce_products=true limits results to items with an Internal ID starting with '01t'. No data is modified.
get_product
ChatGPTRead-only. Returns products by ID or name based on the provided filters. Optional filter: salesforce_products=true limits results to items with an Internal ID starting with '01t'. No data is modified.
get_product
ChatGPTRead-only. Returns products by ID or name based on the provided filters. Optional filter: salesforce_products=true limits results to items with an Internal ID starting with '01t'. No data is modified.
get_talk_to_data_result
ChatGPTRead-only: Retrieve results or terminal failure details for an agent run. When check_talk_to_data shows status "completed", returns the agent response (text, tables). When status is "failed" or "cancelled", returns JSON with status, error, and message explaining what happened (same details as check_talk_to_data) so the client can handle it without guessing. If the run is still "starting" or "running", this returns an error — poll check_talk_to_data first. Use thread_id and message_id from successful runs in subsequent start_talk_to_data calls. This tool only reads data and does not modify anything.
get_talk_to_data_result
ChatGPTRead-only: Retrieve results or terminal failure details for an agent run. Not for QBR jobs; use qbr_get_result with job_id from qbr_start. When check_talk_to_data shows status "completed", returns the agent response (text, tables). When status is "failed" or "cancelled", returns JSON with status, error, and message explaining what happened (same details as check_talk_to_data) so the client can handle it without guessing. If the run is still "starting" or "running", this returns an error — poll check_talk_to_data first. Use thread_id and message_id from successful runs in subsequent start_talk_to_data calls. This tool only reads data and does not modify anything.
get_talk_to_data_result
ChatGPTRead-only: Retrieve results or terminal failure details for an agent run. Not for QBR jobs; use qbr_get_result with job_id from qbr_start. When check_talk_to_data shows status "completed", returns the agent response (text, tables). When status is "failed" or "cancelled", returns JSON with status, error, and message explaining what happened (same details as check_talk_to_data) so the client can handle it without guessing. If the run is still "starting" or "running", this returns an error — poll check_talk_to_data first. Use thread_id and message_id from successful runs in subsequent start_talk_to_data calls. This tool only reads data and does not modify anything.
list_clients
ChatGPTRead-only. List of clients for the current user with fullname and client_id.
list_clients
ChatGPTRead-only. List of clients for the current user with fullname and client_id.
list_clients
ChatGPTRead-only. List of clients for the current user with fullname and client_id.
list_products
ChatGPTRead-only. Returns a product catalog for a specific client. Includes product IDs, names, prices, and status information. No data is modified.
list_products
ChatGPTRead-only. Returns a product catalog for a specific client. Includes product IDs, names, prices, and status information. No data is modified.
list_products
ChatGPTRead-only. Returns a product catalog for a specific client. Includes product IDs, names, prices, and status information. No data is modified.
qbr_detail
ChatGPTGet raw data for a specific QBR section. Returns full agent result data including complete chart data and tables. Use this when you need more detail than qbr_get_result provides (which limits charts to 12 points, tables to 10 rows).
qbr_detail
ChatGPTGet raw data for a specific QBR section. Returns full agent result data including complete chart data and tables. Use this when you need more detail than qbr_get_result provides (which limits charts to 12 points, tables to 10 rows).
qbr_detail
ChatGPTGet raw data for a specific QBR section. Returns full agent result data including complete chart data and tables. Use this when you need more detail than qbr_get_result provides (which limits charts to 12 points, tables to 10 rows).
qbr_get_result
ChatGPTGet QBR job status or results. For pending/running jobs: returns progress info with elapsed time and batch status. For completed jobs: returns slide-ready data for all sections with charts (12 points max), tables (10 rows max), metrics, and insights. Use qbr_detail for full raw data of a specific section.
qbr_get_result
ChatGPTqbr_start jobs: status or slide-ready results. ~20–25 min—don't poll every few seconds (e.g. 3s); slow state, wastes context/load. Say ~20–25 min after start; coffee/other work; recheck on ask or 3–5 min max if monitoring. Pending/running: job_id, client_id, state, started_at, elapsed_in_minutes, progress, batches_completed, last_logs. Completed: all sections—charts ≤12 pts, tables ≤10 rows, metrics, insights. Per-section raw: qbr_detail. Not check_talk_to_data (tracking_id Talk-to-Data only).
qbr_get_result
ChatGPTqbr_start jobs: status or slide-ready results. ~20–25 min—don't poll every few seconds (e.g. 3s); slow state, wastes context/load. Say ~20–25 min after start; coffee/other work; recheck on ask or 3–5 min max if monitoring. Pending/running: job_id, client_id, state, started_at, elapsed_in_minutes, progress, batches_completed, last_logs. Completed: all sections—charts ≤12 pts, tables ≤10 rows, metrics, insights. Per-section raw: qbr_detail. Not check_talk_to_data (tracking_id Talk-to-Data only).
qbr_start
ChatGPTStart a QBR (Quarterly Business Review) generation job. This is a long-running operation (15-25 minutes) that runs in the background. Returns a job_id to check status and retrieve results with qbr_get_result.
qbr_start
ChatGPTStart a QBR (Quarterly Business Review): single quarter (year + quarter), inclusive multi-quarter range (from_year/from_quarter/to_year/to_quarter), or omit year/quarter → last complete quarter. Only one QBR job per user at a time. Quarters run sequentially in one job. Cached agent sections may skip Snowflake when DynamoDB results table is configured. Background job (often ~20+ min). Don't spam qbr_get_result. Returns job_id—use qbr_get_result(job_id). Not Talk-to-Data.
qbr_start
ChatGPTStart a QBR (Quarterly Business Review): single quarter (year + quarter), inclusive multi-quarter range (from_year/from_quarter/to_year/to_quarter), or omit year/quarter → last complete quarter. Only one QBR job per user at a time. Quarters run sequentially in one job. Cached agent sections may skip Snowflake when DynamoDB results table is configured. Background job (often ~20+ min). Don't spam qbr_get_result. Returns job_id—use qbr_get_result(job_id). Not Talk-to-Data.
start_talk_to_data
ChatGPTStart a read-only background agent run to answer a data question. Returns a tracking_id immediately so you can monitor progress. Runs can take up to 15 minutes. Getting client_id: - If client_id is not provided, call list_clients to retrieve the available client IDs - Then use one of those client IDs for this tool - Note: users may refer to client_id as 'account' or 'account ID', but always use the parameter name 'client_id' Starting a new conversation: - Provide client_id and message only - A new thread will be created automatically Continuing a conversation: - Provide thread_id from a previous run's result - Provide parent_message_id from the previous assistant response Workflow: 1. Call this tool to start the run 2. Use check_talk_to_data to monitor progress (poll every 10-15 seconds) 3. When status is "completed", use get_talk_to_data_result to retrieve the response Returns: tracking_id and initial status
start_talk_to_data
ChatGPTStart a read-only background agent run to answer a data question. Returns a tracking_id immediately so you can monitor progress. Runs can take up to 15 minutes. Not for QBR (Quarterly Business Review) reports: use qbr_start and qbr_get_result with job_id instead. Getting client_id: - If client_id is not provided, call list_clients to retrieve the available client IDs - Then use one of those client IDs for this tool - Note: users may refer to client_id as 'account' or 'account ID', but always use the parameter name 'client_id' Starting a new conversation: - Provide client_id and message only - A new thread will be created automatically Continuing a conversation: - Provide thread_id from a previous run's result - Provide parent_message_id from the previous assistant response Workflow: 1. Call this tool to start the run 2. Use check_talk_to_data to monitor progress (poll every 10-15 seconds) 3. When status is "completed", use get_talk_to_data_result to retrieve the response Returns: tracking_id and initial status
start_talk_to_data
ChatGPTStart a read-only background agent run to answer a data question. Returns a tracking_id immediately so you can monitor progress. Runs can take up to 15 minutes. Not for QBR (Quarterly Business Review) reports: use qbr_start and qbr_get_result with job_id instead. Getting client_id: - If client_id is not provided, call list_clients to retrieve the available client IDs - Then use one of those client IDs for this tool - Note: users may refer to client_id as 'account' or 'account ID', but always use the parameter name 'client_id' Starting a new conversation: - Provide client_id and message only - A new thread will be created automatically Continuing a conversation: - Provide thread_id from a previous run's result - Provide parent_message_id from the previous assistant response Workflow: 1. Call this tool to start the run 2. Use check_talk_to_data to monitor progress (poll every 10-15 seconds) 3. When status is "completed", use get_talk_to_data_result to retrieve the response Returns: tracking_id and initial status
