> ## Documentation Index > Fetch the complete documentation index at: https://vaquill.ai/docs/llms.txt > Use this file to discover all available pages before exploring further. # Start an upload > Open an upload session and return a presigned URL for every part. `sizeBytes` must be exact: it fixes the part size and part count for the life of the session, and a wrong value is not detected until assembly fails. Split the file on exactly `partSize` and PUT each part to its own URL, keeping the `ETag` each one returns. Bytes go straight to storage and never through this API. A small file is a one-part upload rather than a separate code path, so there is one flow to implement at every size. ## OpenAPI ````yaml https://api.vaquill.ai/workspace/openapi/v1.json post /v1/uploads openapi: 3.1.0 info: title: Vaquill Legal Workspace API description: >- Organization-scoped API for driving your legal workspace from your own backend: matters, documents, drafting, review, compare and matrices. **Authentication**: `Authorization: Bearer vq_ws_...`. Credentials are issued by an organization owner from the automation console at `/automation` and are shown once. The organization is resolved from the credential and can never be named in a request, so there is no `organizationId` field anywhere in this API and sending one is refused. This is NOT the Vaquill Data API. That one holds a `vq_key_` credential and serves the public legal corpus; the two share no credential, host or endpoint, and sending a `vq_key_` here is refused with an error naming the other product. ## Long-running work Anything that costs real work answers `202` with an **operation**. Poll `GET /v1/operations/{operationId}` until `status` is `succeeded`, `failed` or `cancelled`, honouring `Retry-After` between polls. There are no webhooks, so polling is the only completion signal. The five statuses are `queued`, `running`, `succeeded`, `failed`, `cancelled`, and there is no sixth: every capability reports through the same envelope. Send an `Idempotency-Key` on any launch. A retry with the same key and the same body returns the FIRST operation rather than starting a second billable job, which is what makes recovering from a timeout free. ## Identifiers Every public id is `_<32 lowercase hex>`, for example `mat_9f2c8b1e4a7d43c9b6e0f1a2c3d4e5f6`. The prefix names the resource type. Ids are opaque: take them from responses rather than building them, and never pass a bare database uuid. ## Lists Every collection answers `{data, pagination}`, never a bare array. Page with `limit` and `offset`, and read `pagination.total` to size a job before running it. `pagination.hasMore` is the reliable signal that rows remain. ## Errors Errors are RFC 9457 problem documents (`application/problem+json`), including for a wrong path and a wrong method. Branch on `type`; it is the stable identifier and it resolves to a page describing the failure. `title` and `detail` are written for people and may be reworded at any time. Every response carries `X-Request-ID`, and every error repeats it as `requestId`: quote it when contacting support. A `404` is returned identically for a resource that does not exist, one belonging to another organization, and one outside your installation's matter allowlist. That is deliberate, so the status code cannot be used to discover which ids exist elsewhere. ## Rate limits Limits are per credential and are set by what an operation COSTS rather than by its HTTP method, so a `GET` that runs a retrieval is not a read. Each operation's tier is fixed; a `429` carries `Retry-After`. | Tier | Per minute | Per hour | Per day | |---|---|---|---| | `poll` | 120 | 3000 | 30000 | | `read` | 60 | 1200 | 15000 | | `write` | 30 | 600 | 5000 | | `launch` | 10 | 200 | 2000 | | `download` | 20 | 300 | 3000 | | `upload` | 5 | 100 | 500 | ## Scopes A credential carries an explicit scope set, chosen when it is issued. A call outside them is `403` with `insufficient-scope`, naming the scopes the operation needed. Read and run are separate throughout, so a credential can be allowed to read results without being allowed to spend money producing them. `chronology:read`, `chronology:write`, `clients:read`, `clients:write`, `compare:read`, `compare:run`, `compliance:read`, `compliance:run`, `credentials:rotate`, `documents:download`, `documents:read`, `documents:write`, `drafting:read`, `drafting:run`, `exports:create`, `exports:read`, `facts:read`, `facts:run`, `matrices:read`, `matrices:run`, `matrices:write`, `matters:read`, `matters:write`, `nda:read`, `nda:run`, `operations:read`, `playbooks:read`, `playbooks:run`, `playbooks:write`, `research:read`, `research:run`, `review:read`, `review:run`, `summaries:read`, `summaries:run`, `webhooks:read`, `webhooks:write`, `workflows:read`, `workflows:run`, `workflows:write` version: 1.0.0 servers: - url: https://api.vaquill.ai/workspace description: Vaquill Legal Workspace API (production) - url: /workspace description: Vaquill Legal Workspace API (relative to the mount) security: - WorkspaceAuth: [] tags: - name: Operations description: >- The one job envelope. Every long-running call answers `202` with an operation, and polling this resource is the only way to learn it finished: there are no webhooks. Five statuses, no synonyms. - name: Clients description: The people and companies matters are filed under. Synchronous CRUD. - name: Matters description: >- The unit of work everything else hangs off. A matter scopes documents, drafts, reviews, comparisons and matrices, and it is what a credential's access is checked against. - name: Folders description: >- Organization inside and across matters. The same folders the web app shows, so one created here appears in the customer's browser. - name: Documents description: >- Files in a matter: their metadata, their ingested text, and the original bytes. A document is only usable by retrieval, drafting and review once its status is `succeeded`. - name: Uploads description: >- Getting files in. One presigned multipart flow at every size, with the bytes going straight to storage and never through this API. Initiate, PUT each part, then complete. - name: Research description: >- Asking legal questions and getting grounded, cited answers. This API does NOT stream: an ask answers `202` and you poll the operation, then read the answer off the message it points at. Conversation state is kept here, so you hold a chat id rather than replaying a transcript. Also covers the matter settings behind an answer, the skills you can ask through, and the standalone web-research tools. - name: Drafting description: >- Generating, revising and exporting draft documents. Bodies come out as sections and go in as markdown; the editor's own format is never on the wire. Rendered files come from the export route. - name: Playbooks description: >- The organization's negotiating positions, per contract type, and the starter templates to build them from. A playbook is the input a contract review is measured against. - name: Reviews description: >- Reviewing one contract against a playbook: clause analysis, redlines, flags, liability exposure and a reported sign-off gate. Export applies the redlines as native Word tracked changes. - name: Facts description: >- The cross-document fact ledger for a matter: what every document in it asserts, coalesced and cited back to its source passages. - name: Matter summary description: >- A citation-backed summary of everything filed to a matter, generated on demand and readable claim by claim. - name: Chronology description: >- The dated events across a matter's documents, as one timeline, with duplicate and date-conflict flags. - name: NDA triage description: >- Screening one inbound NDA against ten standard criteria and, where you name one, your own NDA playbook. Answers `green`, `yellow` or `red` with the reasoning per criterion, plus a report you can forward. - name: Compliance description: >- Checking one document against one regulation's requirement checklist: a verdict per requirement with the article it comes from, the gaps, and what to do about them. Only regulations with a real checklist behind them are accepted. - name: Comparisons description: >- Diffing two documents in a matter into structural changes, with a substantive-versus-cosmetic judgment and a downloadable redline. - name: Matrices description: >- Spreadsheet-style extraction across many documents: rows are documents, columns are questions, cells are answers with verified citations. Building a grid is free; running it is what costs. - name: Workflows description: >- Multi-step analyses from a fixed catalogue. Read a definition to learn what documents and inputs it takes, launch a run inside a matter, then download the artifacts it produces. externalDocs: description: Getting started guide and error reference url: https://vaquill.ai/docs/workspace-api paths: /v1/uploads: post: tags: - Uploads summary: Start an upload description: >- Open an upload session and return a presigned URL for every part. `sizeBytes` must be exact: it fixes the part size and part count for the life of the session, and a wrong value is not detected until assembly fails. Split the file on exactly `partSize` and PUT each part to its own URL, keeping the `ETag` each one returns. Bytes go straight to storage and never through this API. A small file is a one-part upload rather than a separate code path, so there is one flow to implement at every size. operationId: uploads.initiate requestBody: required: true content: application/json: schema: $ref: '#/components/schemas/UploadInitiateRequest' example: matterId: mat_9f2c8b1e4a7d43c9b6e0f1a2c3d4e5f6 filename: msa-acme-v3.docx contentType: application/pdf sizeBytes: 248193 folderId: fld_9f2c8b1e4a7d43c9b6e0f1a2c3d4e5f6 handwritten: false responses: '201': description: Successful Response content: application/json: schema: $ref: '#/components/schemas/UploadSession' example: uploadId: upl_9f2c8b1e4a7d43c9b6e0f1a2c3d4e5f6 partSize: 5242880 totalParts: 3 parts: - partNumber: 1 url: >- https://storage.vaquill.ai/exports/msa-acme-v3-redline.docx?signature=... expiresAt: '2026-08-19T14:32:10Z' expiresAt: '2026-08-19T14:32:10Z' '401': description: >- The credential is missing, malformed, unknown, revoked or expired. `type` is `invalid-credential`, or `wrong-product-credential` when a `vq_key_` Data API key was sent to this API. headers: X-Request-ID: description: >- The id of this request. The same value appears as `requestId` in the body. Quote it when contacting support. schema: type: string examples: - req_5f2c8b1e4a7d43c9b6e0f1a2c3d4e5f6 WWW-Authenticate: description: RFC 9110 authentication challenge. schema: type: string content: application/problem+json: schema: $ref: '#/components/schemas/Problem' '403': description: >- The credential does not carry a scope this operation requires (`insufficient-scope`), or the organization's installation is suspended (`installation-inactive`). headers: X-Request-ID: description: >- The id of this request. The same value appears as `requestId` in the body. Quote it when contacting support. schema: type: string examples: - req_5f2c8b1e4a7d43c9b6e0f1a2c3d4e5f6 WWW-Authenticate: description: RFC 9110 authentication challenge. schema: type: string content: application/problem+json: schema: $ref: '#/components/schemas/Problem' '422': description: >- The request does not match the published schema. `errors` names each rejected field and why. The submitted value is never echoed back. headers: X-Request-ID: description: >- The id of this request. The same value appears as `requestId` in the body. Quote it when contacting support. schema: type: string examples: - req_5f2c8b1e4a7d43c9b6e0f1a2c3d4e5f6 content: application/problem+json: schema: $ref: '#/components/schemas/ValidationProblem' '429': description: Too many requests for this credential's tier. Honour `Retry-After`. headers: X-Request-ID: description: >- The id of this request. The same value appears as `requestId` in the body. Quote it when contacting support. schema: type: string examples: - req_5f2c8b1e4a7d43c9b6e0f1a2c3d4e5f6 Retry-After: description: Seconds to wait before retrying. schema: type: integer content: application/problem+json: schema: $ref: '#/components/schemas/Problem' '500': description: >- An unexpected error on our side. The body carries a stable `type` and the request id and nothing else; the cause is in our logs. headers: X-Request-ID: description: >- The id of this request. The same value appears as `requestId` in the body. Quote it when contacting support. schema: type: string examples: - req_5f2c8b1e4a7d43c9b6e0f1a2c3d4e5f6 content: application/problem+json: schema: $ref: '#/components/schemas/Problem' '503': description: >- A dependency this request needs is unavailable, so nothing was done. Retryable. Authentication fails closed rather than admitting the request, so this is never a statement about your credential. headers: X-Request-ID: description: >- The id of this request. The same value appears as `requestId` in the body. Quote it when contacting support. schema: type: string examples: - req_5f2c8b1e4a7d43c9b6e0f1a2c3d4e5f6 content: application/problem+json: schema: $ref: '#/components/schemas/Problem' components: schemas: UploadInitiateRequest: properties: matterId: type: string title: Matterid description: >- `mat_` identifier of the matter to upload into. Must belong to your organization. examples: - mat_9f2c8b1e4a7d43c9b6e0f1a2c3d4e5f6 filename: type: string maxLength: 255 minLength: 1 title: Filename description: >- Filename to store, including its extension. Trimmed; whitespace alone is refused. This exact string is used to build the stored object, so it cannot be changed afterwards. examples: - msa-acme-v3.docx contentType: type: string maxLength: 255 minLength: 1 title: Contenttype description: >- IANA media type of the file, for example `application/pdf`. Must be on the supported list or the request is refused before any URL is signed. examples: - application/pdf sizeBytes: type: integer maximum: 2147483648 exclusiveMinimum: 0 title: Sizebytes description: >- Exact size of the file in bytes. Not advisory: it fixes the part size and part count for the life of the session, and a wrong value fails at complete time. Must be greater than zero. examples: - 248193 folderId: anyOf: - type: string - type: 'null' title: Folderid description: >- `fld_` identifier to file the document under on the way in. Saves a second call. Must belong to your organization. examples: - fld_9f2c8b1e4a7d43c9b6e0f1a2c3d4e5f6 handwritten: type: boolean title: Handwritten description: >- Set true for scanned or handwritten pages to route the file through vision OCR instead of the text-layer reader. Leaving it false on a scanned exhibit ingests it badly. default: false examples: - false additionalProperties: false type: object required: - matterId - filename - contentType - sizeBytes title: UploadInitiateRequest description: >- Ask for somewhere to put a file. `sizeBytes` is required and is not advisory: it decides the part size and the part count, both of which are fixed for the life of the session. R2 cannot be asked to assemble parts it was not told to expect, so a size that turns out to be wrong fails at complete time rather than corrupting a document. UploadSession: properties: uploadId: type: string title: Uploadid description: >- Public identifier for this upload session, `upl_` followed by 32 hex characters. Use it to presign a part and to complete the upload. examples: - upl_9f2c8b1e4a7d43c9b6e0f1a2c3d4e5f6 partSize: type: integer title: Partsize description: >- Size in bytes of every part except the last. Split the file on exactly this boundary or the assembled object will not match. examples: - 5242880 totalParts: type: integer title: Totalparts description: >- How many parts to upload. Always at least 1: a small file is a one-part upload rather than a separate code path. examples: - 3 parts: items: $ref: '#/components/schemas/UploadPartLink' type: array title: Parts description: One presigned URL per part, in order. expiresAt: type: string format: date-time title: Expiresat description: >- When the whole SESSION expires (RFC 3339), which is much later than any individual part URL. Past this, storage discards the parts and the upload has to start again. examples: - '2026-08-19T14:32:10Z' additionalProperties: false type: object required: - uploadId - partSize - totalParts - parts - expiresAt title: UploadSession description: >- An open upload, and every URL needed to fill it. Two things are deliberately not here. The R2 object KEY. It is inside each presigned URL and cannot not be, since an S3 presigned URL is a signed PATH. What is refused is publishing it as a FIELD: a field is a stable, structured handle a client would build on and we could never withdraw, whereas the URL is a fifteen-minute capability for one object the caller was already granted. Nothing cross-tenant is disclosed either way; the uuids in that path are the caller's own. The DOCUMENT ID. There is no document until the bytes land, only an intention to create one, and publishing an id for a row that may never be filled invites a customer to store it and then poll something that never starts. It arrives on the operation's `resource` at complete time, which is the first moment it means anything. Problem: type: object title: Problem description: >- An RFC 9457 problem document. Branch on `type`, which is stable; `title` and `detail` are written for people and may be reworded. Some problems carry extra members (`requiredScopes`, `limit`, `expectedVersion`), which is why this object is open. required: - type - title - status - detail - instance properties: type: type: string format: uri description: >- The stable identifier for this error, and the one field to branch on. Resolves to a page describing it. examples: - https://vaquill.ai/docs/workspace-api/errors/insufficient-scope title: type: string description: A short human-readable summary. examples: - Insufficient scope status: type: integer description: The HTTP status code, repeated. examples: - 403 detail: type: string description: >- What went wrong on this specific request. May be reworded at any time. examples: - >- This credential carries matters:read. This operation needs matters:write. instance: type: string description: The path this problem occurred on. examples: - /workspace/v1/matters/mat_9f2c8b1e4a7d43c9b6e0f1a2c3d4e5f6 requestId: type: string description: >- The id of this request, identical to the `X-Request-ID` response header. Quote it when contacting support. examples: - req_5f2c8b1e4a7d43c9b6e0f1a2c3d4e5f6 additionalProperties: true ValidationProblem: type: object title: ValidationProblem description: >- A problem document for a schema rejection. `errors` lists the fields that were refused. The value you submitted is deliberately not echoed, so a validation failure cannot copy your content into an error response or into either side's logs. required: - type - title - status - detail - instance - errors properties: type: type: string format: uri description: >- The stable identifier for this error, and the one field to branch on. Resolves to a page describing it. examples: - https://vaquill.ai/docs/workspace-api/errors/insufficient-scope title: type: string description: A short human-readable summary. examples: - Insufficient scope status: type: integer description: The HTTP status code, repeated. examples: - 403 detail: type: string description: >- What went wrong on this specific request. May be reworded at any time. examples: - >- This credential carries matters:read. This operation needs matters:write. instance: type: string description: The path this problem occurred on. examples: - /workspace/v1/matters/mat_9f2c8b1e4a7d43c9b6e0f1a2c3d4e5f6 requestId: type: string description: >- The id of this request, identical to the `X-Request-ID` response header. Quote it when contacting support. examples: - req_5f2c8b1e4a7d43c9b6e0f1a2c3d4e5f6 errors: type: array description: One entry per rejected field. items: type: object required: - location - message - type properties: location: type: string description: >- Dotted path to the rejected field, for example `body.contentMarkdown`. message: type: string description: Why it was rejected. type: type: string description: The validation rule that failed. additionalProperties: true UploadPartLink: properties: partNumber: type: integer minimum: 1 title: Partnumber description: >- Which part this URL is for. Parts are numbered from 1 and must be uploaded under the number given here. examples: - 1 url: type: string title: Url description: >- Presigned URL to `PUT` this part's bytes to, directly to storage. No bytes pass through the Vaquill API. Keep the `ETag` the response returns. examples: - >- https://storage.vaquill.ai/exports/msa-acme-v3-redline.docx?signature=... expiresAt: type: string format: date-time title: Expiresat description: >- When this part URL stops working (RFC 3339). Re-sign an expired part with `uploads.presignPart`; parts already uploaded stay valid. examples: - '2026-08-19T14:32:10Z' additionalProperties: false type: object required: - partNumber - url - expiresAt title: UploadPartLink description: >- Where to PUT one part, and how long that stays true. `expiresAt` is on every part rather than on the session, because a large upload outlives a single signature: a client re-signs the parts it has not finished through `uploads.presignPart` and keeps the ones it has. securitySchemes: WorkspaceAuth: type: http scheme: bearer bearerFormat: vq_ws_* description: >- Workspace credential issued from the automation console at `/automation`. Send it as `Authorization: Bearer vq_ws_...`. This is NOT a Data API key: a `vq_key_` credential is refused here and names the other product in the error. ````