Definely

Fallback for clients without a shell (e.g. ChatGPT) where the Step 2 `curl` upload in `definely_upload_document` cannot be run. Given a dragged-in file's download_url, this fetches the file and uploads it to Definely server-side, returning a fileId; then call `definely_prepare_document` with that fileId. Only use this when you cannot perform the Step 2 curl upload yourself — clients that can run curl must use the standard `definely_upload_document` flow instead. If this returns `retryable: true`, call it again (a fresh download_url will be provided) before giving up.

Open the Definely upload widget so the user can pick a local document to analyze. Use this in ChatGPT whenever the user wants to upload a local file — the curl-based `definely_upload_document` flow cannot run there. The widget registers the file with ChatGPT and calls `definely_ingest_document`, which returns a fileId; then call `definely_prepare_document` with that fileId to start processing.

Required workflow: Step 3 of 3 — call this immediately after the curl upload in Step 2. Do not call any other Definely tool until this completes successfully. Runs the full Definely preparation sequence: a. Submits the rules-based analysis job, then the proofreading job with the RB job's ID. b. Polls the rules-based job until it completes (progress == 100%%). c. Stores the document in the Definely document store (Elasticsearch) once the RB job is done. Workflow: 1. `definely_upload_document` — obtain fileId and pre-signed uploadUrl. 2. Bash curl upload — PUT the file to the uploadUrl. 3. `definely_prepare_document` (this tool) — call with the fileId from Step 1 and the original document_name (including extension, e.g. 'contract.docx') to prepare the document for all further Definely tools.

Required workflow: Step 1 of 3 — call this first for every Definely or contract-related request. Do not call any other Definely tool before this. Obtains a pre-signed S3 uploadUrl and fileId for the document. Proceed to Step 2 then Step 3: Step 2 — Upload the file via Bash: curl -X PUT '<uploadUrl>' --data-binary '@/path/to/file' -H 'Content-Type: <content_type>' Content-Type by extension: .pdf -> application/pdf, .docx -> application/vnd.openxmlformats-officedocument.wordprocessingml.document, .doc -> application/msword, .xml -> application/xml. Step 3 — Call `definely_prepare_document` with the fileId. UX note: Do not surface the uploadUrl, curl command, or fileId to the user. Perform the upload silently and only confirm to the user once all three steps are complete.

Use this tool to find a section, clause, defined term, or other document item by its name. The tool searches the document for tokens whose name matches the text you provide and returns the matching results. This is useful when a document contains cross-references and you want to quickly see the text of the referenced item. The tool returns a list of token objects (a token is a meaningful structural unit in a document, such as a section, clause or defined term), each containing: - name: the name of the matching item - text: the text content of the matching item - hash: the unique identifier of the matching item within the document - tokenType: the kind of matching item: section, clause, definition, undefined_term or other Input: - documentUuid (string, required): the unique identifier of the document - tokenName (string, required): the name of the token to search for

Use this tool to find the sections, clauses, defined terms, and other document items that are referred to within a specific part of a document. The tool looks inside the specified section and returns the document items that are referenced within it. The tool returns a list of token objects (a token is a meaningful structural unit in a document, such as a section, clause or defined term), each containing: - name: the title, label, or term being referenced - text: the full text content of the referenced item - hash: the unique identifier of that referenced item within the document - tokenType: the kind of referenced item: section, clause, definition, undefined_term or other This tool is useful for understanding cross-references, checking what a clause depends on, and exploring the context of a provision. Input: - documentUuid (string, required): the unique identifier of the document - tokenHash (string, required): the unique hash identifier of the token whose internal references you want to retrieve

Use this tool to search the Vault for clauses, definitions, and other document items whose name matches your query. Best for: named clauses or terms (e.g. 'Leakage', 'Change of Control') where you know the exact or partial name. The Vault is the firm's repository of precedent clauses, definitions, and other reusable drafting materials. This tool performs a lexical search, which means it matches exact words or phrases in the item name. It can return results where the name fully matches or partially matches the search term you provide. You can refine the search using metadata filters such as contract type, governing law, parties, authors, document name, token type, and other available filters. The tool returns items (up to 5), total, limit and offset. Each item can include: - name and text: the heading and text of the matching item - tokenType: the kind of item: section, clause, definition or other - documentName, folderName, folderAncestors and tokenAncestors: where the item sits - contractType, jurisdiction, parties, labels and authors: document metadata, as names - isGoldStandard, vaultType, isInsertedManually, insertionCount, isFrequentlyUsed, createdAt and lastModifiedAt Use this tool when you want to retrieve a clause or defined term from the Vault and you already know its name, or at least part of its name. Do not use this tool when you do not need Vault content, when you do not know any exact part of the item name, or when you want to search by meaning rather than exact wording. Tips: - Set primaryDocumentUuid to the fileId of the uploaded document — this surfaces contextually relevant results higher. Without a document, use 00000000-0000-0000-0000-000000000000. - When the user names a specific precedent document, pass documentNames: ["<exact name>"] to scope the search to that document only. - People who drafted, wrote or authored a document go in authors; the companies or individuals that entered into the agreement go in parties. - Vault search returns up to 5 results by default. If the first batch does not include the right result, increase offset to paginate. - The vault contains both Personal (individual lawyer) and Company (firm-wide) documents. Leave vaultTypes empty to search both. Only filter to one type if the user explicitly requests it. - If this lexical search returns no results or off-topic results, follow up with find_similar_items_in_vault using a description of what the clause does as the text parameter. - If the user mentions a specific document name that results do not surface, try find_precedent_in_vault with the document name to locate it directly, then filter subsequent searches by that documentName. Input: - name (string, required): the name or partial name to search for - primaryDocumentUuid (string, required): the fileId of the current document, or 00000000-0000-0000-0000-000000000000 - offset (integer, optional, default 0): the number of results to skip; each call returns up to 5 results - contractTypes, labels, parties, governingLaws, documentNames (list of strings, optional): only search items with these values - authors (list of strings, optional): only search items from documents drafted by these people; several are OR'ed - vaultTypes (list of Personal or Company, optional): leave empty to search both - tokenType (one of All, References, Definitions, Clauses, SubClauses, optional, default All): References are section headings - goldStandardOnly, manuallyAddedOnly, bookmarkedOnly (boolean, optional, default false)

Use this tool to find the Vault document whose name best matches a given name and retrieve its reference report. The Vault is the law firm's repository of precedent contracts, clauses, defined terms, and other reusable drafting materials. A reference report is a structured outline of a document that shows its sections, clauses, and subclauses in a hierarchical view. This gives you a high-level skeleton of the precedent document, making it easier to compare structure, spot missing provisions, and understand how your current document differs from similar precedents. The tool returns: - documentName: the name of the related Vault document - references: the reference report showing the structure of the related document - needsReprocessingForReferences: true when the document must be reprocessed before its references are available If no Vault document matches, the tool returns no content. Tip: Use this tool when the user mentions a specific document name that find_items_by_name_in_vault did not surface — locate it here directly by name, then filter subsequent vault searches by that documentName. Input: - vaultDocumentName (string, required): the name of the Vault document to find - contractTypes, labels, parties, governingLaws (list of strings, optional): only search precedents with these values - vaultTypes (list of Personal or Company, optional): leave empty to search both - goldStandardOnly, bookmarkedOnly (boolean, optional, default false)

Use this tool to search the Vault for clauses, definitions, and other document items that are similar in meaning to your query. Best for: concepts or behaviours (e.g. 'how we cap liability on warranties') — pass a descriptive text to search by meaning. The Vault is the firm's repository of precedent clauses, definitions, and other reusable drafting materials. This tool performs a semantic search, which means it looks for content that is similar in meaning, context, or concept, not just exact wording. Search with a text snippet, and optionally a name. You can also refine the search using offset and metadata filters such as contract type, governing law, parties, authors, document name, token type, and other available filters. The tool returns items (up to 5), total, limit and offset. Each item can include: - name and text: the heading and text of the matching item - tokenType: the kind of item: section, clause, definition or other - documentName, folderName, folderAncestors and tokenAncestors: where the item sits - contractType, jurisdiction, parties, labels and authors: document metadata, as names - isGoldStandard, vaultType, isInsertedManually, insertionCount, isFrequentlyUsed, createdAt and lastModifiedAt Use this tool when you want to find similar provisions based on meaning or wording, or when you are looking for a clause or defined term but do not know its exact name. Do not use this tool when you do not need Vault content, or when you already know the exact name of the clause or defined term and a lexical search would be more appropriate. Tips: - Set primaryDocumentUuid to the fileId of the uploaded document — this surfaces contextually relevant results higher. Without a document, use 00000000-0000-0000-0000-000000000000. - When the user names a specific precedent document, pass documentNames: ["<exact name>"] to scope the search to that document only. - People who drafted, wrote or authored a document go in authors; the companies or individuals that entered into the agreement go in parties. - Vault search returns up to 5 results by default. If the first batch does not include the right result, increase offset to paginate. - The vault contains both Personal (individual lawyer) and Company (firm-wide) documents. Leave vaultTypes empty to search both. Only filter to one type if the user explicitly requests it. Input: - text (string, required): the wording or passage to search by meaning - primaryDocumentUuid (string, required): the fileId of the current document, or 00000000-0000-0000-0000-000000000000 - name (string, optional): the title, label, or term to search for - offset (integer, optional, default 0): the number of results to skip; each call returns up to 5 results - contractTypes, labels, parties, governingLaws, documentNames (list of strings, optional): only search items with these values - authors (list of strings, optional): only search items from documents drafted by these people; several are OR'ed - vaultTypes (list of Personal or Company, optional): leave empty to search both - tokenType (one of All, References, Definitions, Clauses, SubClauses, optional, default All): References are section headings - goldStandardOnly, manuallyAddedOnly, bookmarkedOnly (boolean, optional, default false)

Use this tool to find the most relevant similar document to current document in the Vault and retrieve its reference report. The Vault is the law firm's repository of precedent contracts, clauses, defined terms, and other reusable drafting materials. A reference report is a structured outline of a document that shows its sections, clauses, and subclauses in a hierarchical view. This gives you a high-level skeleton of the precedent document, making it easier to compare structure, spot missing provisions, and understand how your current document differs from similar precedents. The tool returns: - documentName: the name of the related Vault document - references: the reference report showing the structure of the related document - needsReprocessingForReferences: true when the document must be reprocessed before its references are available If no Vault document matches, the tool returns no content. This tool is useful for precedent comparison, gap analysis, and identifying clauses or sections that may be missing from the document you are working on. Input: - documentUuid (string, required): the fileId of the current document, as returned by definely_prepare_document - contractTypes, labels, parties, governingLaws (list of strings, optional): only search precedents with these values - vaultTypes (list of Personal or Company, optional): leave empty to search both - goldStandardOnly, bookmarkedOnly (boolean, optional, default false)