Skip to main content

Files, Uploads, Exports, and Generated Files

Files, Uploads, Exports, and Generated Files

Use this guide when a file will not upload, Hatz cannot read a file as expected, a generated file cannot be downloaded, or an export does not appear where you expected.

Start here

  • Check the current supported file types: use the Supported File Types article as the source for file-family support. If a format is not listed there, convert it to a listed format before retrying.

  • Confirm the surface: upload behavior can differ between Chat, Workflow inputs, Agents, Apps, generated files, exports, email attachments, and support-chat attachments.

  • Check the action: uploading, parsing, generating, exporting, downloading, and attaching to a support conversation are different actions.

  • Separate source files from generated files: uploaded files are inputs for Hatz to read; generated files are outputs Hatz creates. Their limits, retention, and recovery paths can differ.

  • Remove unnecessary sensitive data: share metadata, screenshots, and error text first. Do not send raw sensitive files to Support unless Hatz specifically asks for them through an approved path.

Common symptoms

  • Upload is blocked: the file type, size, count, browser state, permissions, or product surface may not support that upload path.

  • File is still processing: wait briefly before retrying. If processing changes to a failed or could-not-be-processed state, re-upload the source file.

  • File was accepted but not read correctly: the file may contain images, scans, macros, tables, formatting, protected content, or content that needs conversion before analysis.

  • File is missing or ambiguous: attach the file to the current chat. If multiple attachments have the same filename, remove the duplicate, rename one file, or clearly identify the intended file.

  • Generated file is missing: the run may have completed text output but failed while creating, saving, or linking the generated file. Turn on Auto-Tool or Code Execution and ask for an actual downloadable output. If the message says the generated-file count was exceeded, see Generated files could not be saved.

  • Download fails: stay signed in to the same workspace and use Download on the Generated file card. Refresh the page rather than copying an old file URL. The file may have expired, the export may have failed, or browser/network settings may be blocking the download.

  • Email or chat attachment is not reflected in output: attachments sent through a support channel are not always the same as product inputs to a workflow, agent, or chat run.

API Uploads and Sandbox Outputs

For API workflows, upload source files through the Hatz File Upload API, wait for the file to be ready, and pass the returned file_id into chat, app, workflow, or agent requests. See File Upload and Chat Completions.

A file path produced inside a code execution sandbox is not the same thing as an uploaded file ID or a public download URL. If Hatz returns a sandbox path, ask Hatz to create a downloadable file for that output, or contact Support with the chat/run link, timestamp, and any visible file or request IDs.

Before retrying

  • Convert unsupported or unusual formats to a documented supported format when possible.

  • Use a smaller test file if the original file is large or contains many pages, sheets, images, or embedded objects.

  • For scanned PDFs or image-heavy files, confirm the text is selectable or provide a text-based version if available.

  • For image-heavy source material in Agents, prefer a text summary when possible. A practical pattern is to ask Hatz to summarize the image into text, save that text in a supported document format, and use the text document as the source.

  • For spreadsheets, simplify hidden sheets, macros, external links, and complex formatting before retrying.

  • Retry once in the same surface and once in a clean browser session before treating it as a product defect.

What to include in a support ticket

Include enough context for Support to route the issue without asking for the full file contents first.

  • Workspace or tenant.

  • Product surface: Chat, Workflow, Agent, App, API, generated file, export, support chat, email, Microsoft 365 file picker, OneDrive, SharePoint, or another area.

  • File family and extension, such as PDF, DOCX, XLSX, CSV, PNG, ZIP, or another format.

  • Approximate file size bucket, such as small, medium, large, or larger than the published limit, if known.

  • Action attempted: upload, parse, generate, export, download, attach, open, save to Microsoft 365, or use as an Agent source.

  • Time and timezone.

  • Expected output and actual result.

  • Displayed error, screenshot, generated-file panel, workflow run link, item URL, source URL, or export/download state, with secrets and unnecessary customer data removed.

  • For OneDrive or SharePoint issues, include both the Hatz link and the Microsoft file, folder, or site URL when safe to share.

  • Whether the issue repeats with the same file and whether a smaller or converted file works.

For product UI issues, share visible file, generated-file, workflow, chat, app, agent, or item links when available, plus timestamp, screenshot, and error text. For API issues, include the response X-Request-ID and visible file_id when available.

When to contact Support

  • A file type listed as supported fails reproducibly after reasonable retry and conversion checks.

  • A generated file or export repeatedly fails for multiple users, tenants, runs, or browsers.

  • The UI shows a failure state, missing file, expired link, or download problem without enough information for the customer to recover.

  • You have the file family, sanitized size bucket, surface, timestamp, error, and reproduction path needed for Support to route the issue.

Things to know

  • A support-chat or email attachment is not always the same as a product input to a workflow, agent, or chat run.

  • A generated file may not always be recoverable after a link expires or a generation step fails.

  • Large source sets, very large files, scanned files, or image-heavy sources can make responses slower, less reliable, or more credit-intensive because Hatz has more material to process.

  • Exact limits, retry behavior, artifact lifetime, generated-file retention, and parser behavior can vary by surface and file type. Use the current supported-file-types article and in-product error messages as the source of truth.

Did this answer your question?