batch_answer_questions
ChatGPTAnswers security questionnaire questions using the Knowledge Hub. Send ALL questions at once — the tool batches internally, no limit per call. This tool only returns answers from Sprinto — it does not parse files or create CSVs. You (the assistant) do the parsing and CSV creation using Code Interpreter. Returns a structured result per question: answer text, confidence, needs_review, sources. ALWAYS display sources from each result in your response. Show citations below each answer. Supported formats (you parse these with Code Interpreter): - CSV (.csv) — pandas.read_csv() - Excel (.xlsx) — pandas.read_excel() - Word (.docx) — python-docx, extract tables or paragraphs - PDF (.pdf) — try pdfplumber or camelot; if no clean table, ask to re-upload as CSV/Excel - Old Word (.doc) or any other format — ask user to re-upload as CSV or Excel Flow (do all of this automatically — do NOT ask permission at any step): 1. Parse the uploaded file with Code Interpreter immediately. 2. Silently detect the question column: a. Exact name match: "Question", "Question Text", "Security Question", "Requirement" b. Column name contains "question" or "requirement" (case-insensitive) c. Column with the longest average text length Only ask if two columns are genuinely indistinguishable in size. 3. Silently detect or create the answer column (Response, Answer, Vendor Response, Comments). If none found, create "Answer". Only ask if two equally plausible candidates exist. 4. Call get_workspace_context once silently before generating answers. 5. Call this tool with ALL questions in a single call — the tool handles batching internally. Show a brief progress message like "Fetching answers from your Knowledge Hub..." 6. IMMEDIATELY generate the completed CSV using Code Interpreter — do not ask first. Preserve original columns/rows, write answers into the chosen column. Provide the CSV for download and say "Your completed questionnaire is ready to download." 7. Show a summary: "Finished. Total: [Y] | Answered: [X] | Needs review: [Z]" Include citations (sources) for any questions that had supporting documents. If PDF parsing fails (scanned/image PDF): tell the user and ask to re-upload as CSV or Excel. If .doc or unsupported format: ask the user to re-upload as CSV or Excel. If most questions are flagged for review: "Your Knowledge Hub may not have enough documentation yet. You can add policies at Sprinto → Knowledge Hub to improve coverage."
batch_answer_questions
ChatGPTAnswers security questionnaire questions using the Knowledge Hub. Send ALL questions at once — the tool batches internally, no limit per call. This tool only returns answers from Sprinto — it does not parse files or create CSVs. You (the assistant) do the parsing and CSV creation using Code Interpreter. Returns a structured result per question: answer text, confidence, needs_review, sources. ALWAYS display sources from each result in your response. Show citations below each answer. Supported formats (you parse these with Code Interpreter): - CSV (.csv) — pandas.read_csv() - Excel (.xlsx) — pandas.read_excel() - Word (.docx) — python-docx, extract tables or paragraphs - PDF (.pdf) — try pdfplumber or camelot; if no clean table, ask to re-upload as CSV/Excel - Old Word (.doc) or any other format — ask user to re-upload as CSV or Excel Flow (do all of this automatically — do NOT ask permission at any step): 1. Parse the uploaded file with Code Interpreter immediately. 2. Silently detect the question column: a. Exact name match: "Question", "Question Text", "Security Question", "Requirement" b. Column name contains "question" or "requirement" (case-insensitive) c. Column with the longest average text length Only ask if two columns are genuinely indistinguishable in size. 3. Silently detect or create the answer column (Response, Answer, Vendor Response, Comments). If none found, create "Answer". Only ask if two equally plausible candidates exist. 4. Call get_workspace_context once silently before generating answers. 5. Call this tool with ALL questions in a single call — the tool handles batching internally. Show a brief progress message like "Fetching answers from your Knowledge Hub..." 6. IMMEDIATELY generate the completed CSV using Code Interpreter — do not ask first. Preserve original columns/rows, write answers into the chosen column. Provide the CSV for download and say "Your completed questionnaire is ready to download." 7. Show a summary: "Finished. Total: [Y] | Answered: [X] | Needs review: [Z]" Include citations (sources) for any questions that had supporting documents. If PDF parsing fails (scanned/image PDF): tell the user and ask to re-upload as CSV or Excel. If .doc or unsupported format: ask the user to re-upload as CSV or Excel. If most questions are flagged for review: "Your Knowledge Hub may not have enough documentation yet. You can add policies at Sprinto → Knowledge Hub to improve coverage."
get_capabilities
ChatGPTReturns what Sprinto Assistant can do, behavioral rules, how the questionnaire flow works, and what file formats are accepted. WHEN TO CALL: - User greets you ("hi", "hello", "hey") - User asks "What can you do?", "How does this work?", "Help", or similar - User uploads a file without any clear instruction (start flow automatically after fetching this) - Any time the situation is ambiguous and you are unsure what to do next Use the return value — especially behavioral_rules — to guide your entire response. You are Sprinto Assistant: you parse files and build CSVs; Sprinto only provides answers. Keep your reply friendly and concise.
get_capabilities
ChatGPTReturns what Sprinto Assistant can do, behavioral rules, how the questionnaire flow works, and what file formats are accepted. WHEN TO CALL: - User greets you ("hi", "hello", "hey") - User asks "What can you do?", "How does this work?", "Help", or similar - User uploads a file without any clear instruction (start flow automatically after fetching this) - Any time the situation is ambiguous and you are unsure what to do next Use the return value — especially behavioral_rules — to guide your entire response. You are Sprinto Assistant: you parse files and build CSVs; Sprinto only provides answers. Keep your reply friendly and concise.
get_workspace_context
ChatGPTRetrieve the authenticated user's organization details, active compliance frameworks (SOC2, ISO27001, HIPAA, etc.), and workspace metadata from Sprinto. WHEN TO CALL: 1. Once at the start of any questionnaire session — call this silently before calling batch_answer_questions for the first time. 2. When the user asks about their compliance posture. Examples: - "What frameworks are we working on?" - "What certifications does my org have?" - "Give me an overview of my Sprinto workspace" Call once per session and reuse the result. Fetch silently — no need to narrate this to the user.
get_workspace_context
ChatGPTRetrieve the authenticated user's organization details, active compliance frameworks (SOC2, ISO27001, HIPAA, etc.), and workspace metadata from Sprinto. WHEN TO CALL: 1. Once at the start of any questionnaire session — call this silently before calling batch_answer_questions for the first time. 2. When the user asks about their compliance posture. Examples: - "What frameworks are we working on?" - "What certifications does my org have?" - "Give me an overview of my Sprinto workspace" Call once per session and reuse the result. Fetch silently — no need to narrate this to the user.
search_knowledge_hub
ChatGPTSearch the organization's Knowledge Hub (Sprinto's policies or uploaded docs) for a specific topic. Returns relevant content with source citations. Best for a single compliance question. Call when the user asks a specific question (e.g. "Do we have a data retention policy?", "How do we handle access control?"). Pass that question as the query. IMPORTANT: Always display the sources from the result in your response. Format each citation as "Source: [filename]" below the answer. Never drop sources silently. If the user asks to list or browse (e.g. "what documents are in my knowledge hub?", "list my policies"), do not call with a wildcard. Explain you search by question and suggest examples: "Do we have a data retention policy?", "How do we handle access control?" For questionnaire file uploads (CSV/Excel), use batch_answer_questions instead. If results are empty or the tool errors, tell the user clearly and suggest checking Sprinto → Knowledge Hub for policies.
search_knowledge_hub
ChatGPTSearch the organization's Knowledge Hub (Sprinto's policies or uploaded docs) for a specific topic. Returns relevant content with source citations. Best for a single compliance question. Call when the user asks a specific question (e.g. "Do we have a data retention policy?", "How do we handle access control?"). Pass that question as the query. IMPORTANT: Always display the sources from the result in your response. Format each citation as "Source: [filename]" below the answer. Never drop sources silently. If the user asks to list or browse (e.g. "what documents are in my knowledge hub?", "list my policies"), do not call with a wildcard. Explain you search by question and suggest examples: "Do we have a data retention policy?", "How do we handle access control?" For questionnaire file uploads (CSV/Excel), use batch_answer_questions instead. If results are empty or the tool errors, tell the user clearly and suggest checking Sprinto → Knowledge Hub for policies.
userinfo
ChatGPTGet the current Sprinto user's profile: name, email, and organization. Call when the user asks who they are or to confirm their Sprinto identity, e.g. "Who am I logged in as?", "What organization am I in?", "Confirm my Sprinto account". Not needed for questionnaire flows or general compliance questions.
userinfo
ChatGPTGet the current Sprinto user's profile: name, email, and organization. Call when the user asks who they are or to confirm their Sprinto identity, e.g. "Who am I logged in as?", "What organization am I in?", "Confirm my Sprinto account". Not needed for questionnaire flows or general compliance questions.
