ask_pillar_analyst
ChatGPTAsk a deep analytical question about your bank and get a grounded answer from the TBPI pillar analyst (the same engine behind the web chat). Best for 'why' and 'what should we do about' questions that need reasoning over the bank's data and peer context. Slower than the data tools (runs a full analysis).
check_upload_status
ChatGPTStatus of your bank's recent data-room uploads: ready (with DR code, row count, SHA-256 checksums), failed (with the exact validation reasons to relay), or pending. Call after the user says they've uploaded, or when they ask about their data.
compare_to_peers
ChatGPTYour bank's peer standing for a pillar's key metrics (or all pillars): performance percentile, positional label, and peer median for each. Higher percentile always means better. Use for 'how do we compare to peers?' questions.
get_bank_overview
ChatGPTOverview of your bank: identity, latest data snapshot (code and quarter), and the most recent TBPI pillar scores if generated. Start here for any conversation about the bank's performance.
get_metric_history
ChatGPTQuarterly history for one or more of your bank's financial metrics (e.g. 'Net Interest Margin (NIM)', 'Efficiency Ratio', 'Return on Assets (ROA)'), with the current peer standing. Percentiles are PERFORMANCE percentiles: higher always means better. Some metrics carry a standing word instead of a percentile (banded/threshold/context metrics).
get_pillar_scores
ChatGPTYour bank's five TBPI pillar scores (Revenue, Cost & Efficiency, Risk, Capital, Operations) from the latest completed assessment. Scale: 1=Strong, 2=Satisfactory, 3=Fair, 4=Marginal, 5=Unsatisfactory.
get_report
ChatGPTFetch the full markdown content of one of your bank's reports by its id (from list_reports).
get_snapshot_provenance
ChatGPTProvenance and audit trail for your bank's current data snapshot: snapshot code, quarter, SHA-256 checksums of each source CSV, and when/by whom it was created. Use to answer 'where does this data come from?' and auditability questions.
get_uploaded_data
ChatGPTThe normalized rows of a READY data-room upload (latest, or by DR code) so you can analyze the bank's own uploaded data alongside its TBPI metrics. Data is deterministic-validated and checksummed; cite values verbatim.
list_my_banks
ChatGPTList the bank(s) your account is assigned to. Use when unsure which institution your other tool calls will operate on.
list_reports
ChatGPTList the TBPI reports and documents generated for your bank (pillar deep dives, executive assessment, CAMELS, uploads). Returns each report's id, title, type, and pillar.
request_upload
ChatGPTStart a data-room upload for your bank. Returns a secure, single-use upload link (valid 30 minutes) plus the exact template the file must follow. Flow: call this when the user wants to upload data → give them the link → they upload in their browser → when they say they're done, call check_upload_status to confirm ingestion. Available template: custom_metrics (columns: metric_name, quarter, value — any metric time series).
