get_file_content
ChatGPTOpen and read the full content of a document PDF stored in Omissis. Use this whenever the user wants to read, open, extract text from, or analyse a specific file. Requires the document_id obtained from search_files.
Productivity
Overview
Edito-Pdf lets you search, read, and analyze documents that have been pseudonymized or anonymized by the Omissis system. Using ChatGPT, you can locate files across your personal drive and shared drives, extract their text, and generate professional reports or summaries. When a document needs to be sent with real data restored, Omissis performs de-pseudonymization automatically and delivers the result to your email — without ever exposing sensitive information in the conversation.
Tools
get_file_content
ChatGPTOpen and read the full content of a document PDF stored in Omissis. Use this whenever the user wants to read, open, extract text from, or analyse a specific file. Requires the document_id obtained from search_files.
get_file_content
ChatGPTOpen and read the full content of a document PDF stored in Omissis. Use this whenever the user wants to read, open, extract text from, or analyse a specific file. Requires the document_id obtained from search_files.
search_files
ChatGPTSearch for documents processed by the Omissis redaction system. You can filter by file name and by folder path (e.g. "/MyFolder/SubFolder"). Returns a list of files with their document_id. Use get_file_content with the document_id to read the full content of any file. VERY IMPORTANT: Never display or mention the document_id to the end user. It is for internal use only.
search_files
ChatGPTSearch for documents processed by the Omissis redaction system. You can filter by file name and by folder path (e.g. "/MyFolder/SubFolder"). Returns a list of files with their document_id. Use get_file_content with the document_id to read the full content of any file. VERY IMPORTANT: Never display or mention the document_id to the end user. It is for internal use only.
search_files
ChatGPTSearch for documents processed by the Omissis redaction system. You can filter by file name and by folder path (e.g. "/MyFolder/SubFolder"). Returns a list of files with their document_id. Use get_file_content with the document_id to read the full content of any file. VERY IMPORTANT: Never display or mention the document_id to the end user. It is for internal use only.
send_file_for_depseudonymization
ChatGPTTakes an HTML document (or other formatted content) that YOU compose, replaces all pseudonym labels in it with the real values from the documents' entity dictionaries, and delivers the result by email to the user. The user's email is resolved server-side — never ask the user for it. WHEN TO CALL: whenever the user asks to depseudonymize a document, send a report, generate a summary from Omissis documents, or receive any document with real data. MANDATORY STEPS BEFORE CALLING THIS TOOL: 1. Call search_files to locate the relevant documents. 2. Call get_file_content to read the pseudonymized text of each document. 3. YOU MUST COMPOSE the full output document yourself — write a complete, well-styled HTML document (report, summary, full text, or whatever the user requested) using the pseudonymized content you read. Do NOT pass the raw text from get_file_content; compose a new document around it. 4. Only then call this tool, passing the HTML you wrote as the "content" parameter. HTML REQUIREMENTS for docx output (default): - Write a complete <!DOCTYPE html>...<html>...<body>...</body></html> document. - Use INLINE STYLES on every element (font-family, font-size in pt, color, padding, margin). - Tables: wrap rows in <thead>/<tbody>; apply inline styles to each <th> and <td>. - Alternate table row colors with style="background-color:..." on each <tr>. - For xlsx: pass CSV (comma-separated, first row = headers). For txt/md: plain text. AFTER THIS TOOL SUCCEEDS: say only one sentence confirming the email was sent and how many entities were replaced. Do NOT narrate the steps, do NOT mention HTML/DOCX/format names, do NOT reveal any real entity value. ERROR: if the tool returns "DEPSEUDONYMIZATION_IMPOSSIBLE" — stop immediately, do not retry, explain the reason to the user in plain language. PRIVACY: real entity values are sent ONLY via email. Never display, quote, or infer them in the conversation.
send_file_for_depseudonymization
ChatGPTTakes an HTML document (or other formatted content) that YOU compose, replaces all pseudonym labels in it with the real values from the documents' entity dictionaries, and delivers the result by email to the user. The user's email is resolved server-side — never ask the user for it. WHEN TO CALL: whenever the user asks to depseudonymize a document, send a report, generate a summary from Omissis documents, or receive any document with real data. MANDATORY STEPS BEFORE CALLING THIS TOOL: 1. Call search_files to locate the relevant documents. 2. Call get_file_content to read the pseudonymized text of each document. 3. YOU MUST COMPOSE the full output document yourself — write a complete, well-styled HTML document (report, summary, full text, or whatever the user requested) using the pseudonymized content you read. Do NOT pass the raw text from get_file_content; compose a new document around it. 4. Only then call this tool, passing the HTML you wrote as the "content" parameter. HTML REQUIREMENTS for docx output (default): - Write a complete <!DOCTYPE html>...<html>...<body>...</body></html> document. - Use INLINE STYLES on every element (font-family, font-size in pt, color, padding, margin). - Tables: wrap rows in <thead>/<tbody>; apply inline styles to each <th> and <td>. - Alternate table row colors with style="background-color:..." on each <tr>. - For xlsx: pass CSV (comma-separated, first row = headers). For txt/md: plain text. AFTER THIS TOOL SUCCEEDS: say only one sentence confirming the email was sent and how many entities were replaced. Do NOT narrate the steps, do NOT mention HTML/DOCX/format names, do NOT reveal any real entity value. ERROR: if the tool returns "DEPSEUDONYMIZATION_IMPOSSIBLE" — stop immediately, do not retry, explain the reason to the user in plain language. PRIVACY: real entity values are sent ONLY via email. Never display, quote, or infer them in the conversation.
send_file_for_depseudonymization
ChatGPTTakes an HTML document (or other formatted content) that YOU compose, replaces all pseudonym labels in it with the real values from the documents' entity dictionaries, and delivers the result by email to the user. The user's email is resolved server-side — never ask the user for it. WHEN TO CALL: whenever the user asks to depseudonymize a document, send a report, generate a summary from Omissis documents, or receive any document with real data. MANDATORY STEPS BEFORE CALLING THIS TOOL: 1. Call search_files to locate the relevant documents. 2. Call get_file_content to read the pseudonymized text of each document. 3. YOU MUST COMPOSE the full output document yourself — write a complete, well-styled HTML document (report, summary, full text, or whatever the user requested) using the pseudonymized content you read. Do NOT pass the raw text from get_file_content; compose a new document around it. 4. Only then call this tool, passing the HTML you wrote as the "content" parameter. HTML REQUIREMENTS for docx output (default): - Write a complete <!DOCTYPE html>...<html>...<body>...</body></html> document. - Use INLINE STYLES on every element (font-family, font-size in pt, color, padding, margin). - Tables: wrap rows in <thead>/<tbody>; apply inline styles to each <th> and <td>. - Alternate table row colors with style="background-color:..." on each <tr>. - For xlsx: pass CSV (comma-separated, first row = headers). For txt/md: plain text. AFTER THIS TOOL SUCCEEDS: say only one sentence confirming the email was sent and how many entities were replaced. Do NOT narrate the steps, do NOT mention HTML/DOCX/format names, do NOT reveal any real entity value. ERROR: if the tool returns "DEPSEUDONYMIZATION_IMPOSSIBLE" — stop immediately, do not retry, explain the reason to the user in plain language. PRIVACY: real entity values are sent ONLY via email. Never display, quote, or infer them in the conversation.
Capabilities
Writes
