Repository navigation
Troubleshooting
Deutsch: Fehlerbehebung
When something goes wrong, AI Studio shows a message at the edge of the window, red for errors and yellow for warnings. The message usually names the cause, often with the text the provider sent. This page lists the most common problems and what helps.
AI Studio retries by itself when a provider is busy or has a temporary problem: up to six times, waiting longer each time. If the request still fails, the message appears, and your message stays in the chat. Click "Edit" on it to put the text back into the input field, choose another provider if you like, and send it again. If the answer broke off, "Regenerate" asks the model again.
| The message says | What helps |
|---|---|
| "The API key might be invalid" or the key "is missing or was rejected" | Enter the key again: "Settings" → "Configure LLM Providers" → "Edit". |
| "You might not be able to use this provider from your location" or the provider "refused the request" | Your account may not be allowed to use the model, or the provider doesn't serve your region. Choose another model or provider. |
| "too many requests were sent" | Wait a moment and try again. If it happens often, check the limits of your plan with the provider. |
| You have no "API credits left" | Add credits to your account with the provider. |
| The provider "could not be reached", or "There were some problems with the request" | Check your network, VPN, and proxy. For a provider you run yourself, check that the server runs and that the address in the provider settings is right. |
| The request "timed out" | Try again. For long tasks, raise "Request timeout" in the Settings. |
| The provider "does not know the selected model", "Something was not found", or the model "is no longer available" | Edit the provider, click "Load models", and choose a current model. |
| The data is "too large for the selected model", or the text "was longer than the selected model accepts" | Remove attachments, start a new chat, or choose a model with a larger context window. |
| "The selected model is not able to use tools" | Choose a model that can, or switch tools off; the message tells you how. |
| "The server might be down" or "The provider is overloaded" | Try again later, or choose another provider. |
Below the provider selection, AI Studio explains why it is empty:
- "No LLM providers are configured yet." Add one: see Providers and Models.
- "The use of all your LLM providers has been prohibited by your organization." Add a provider your organization allows, or ask your IT department.
- "No LLM providers meet the confidence requirements." A minimum confidence hides your providers, either for the whole app or for this assistant. Lower it, or add a provider you trust more; see Provider Confidence. Your organization may have set the minimum.
"Load models" works once you entered an API key, or for a provider you run yourself, once you chose its host. When loading fails, the message names the reason: a missing, invalid, or expired key, missing permissions, no credits, too many requests, or a provider that can't be reached.
For a provider you run yourself, "due to an unknown error" usually means that the server isn't running or the address is wrong. The address must start with http:// or https://, for example, http://localhost:1234.
"Failed to store the API key in the operating system" names the reason. On Linux, AI Studio needs a password manager with the FreeDesktop Secret Service and an unlocked default collection:
- "no default collection is configured": create a collection in your password manager, unlock it, and set it as the default.
- "the default collection is locked": unlock it in your password manager.
- "No compatible secure-storage service is available": set up a password manager that offers the Secret Service.
See Secure storage for API keys. On Windows and macOS, the details are in the log; see below.
-
A
.docfile: convert it to.docx, for example, with Microsoft Word. - HEIC, AVIF, or SVG images: convert them to JPG or PNG.
- "Images are not supported by the selected provider and model": choose a model that reads images.
- A file "needs Pandoc": install Pandoc; see Pandoc and File Formats.
- "No text could be read from the file": often a scanned PDF without a text layer. Run text recognition on it first.
- The file "is currently open in another program": close it there and try again.
- Audio and video need a transcription provider; see Voice Input.
- "Some files could not be accessed" after dropping them: choose them with the file chooser instead.
- No embedding provider: local data sources need one; add it right where AI Studio asks, or see Data Sources.
- A data source "Needs attention": open the data sources and look at its message. If its index can't be read anymore, "Repair this data source by indexing it anew" rebuilds it.
- A folder "cannot be reached": connect the drive it is on. AI Studio removes nothing from the index meanwhile.
- "Your data sources cannot be used with the selected providers": the data sources demand a provider you trust more. Choose another provider.
- "An ERI server reports 'Connection failed.'": check the address and port, and ask your IT department.
- "Failed to check for updates." AI Studio asks GitHub for new versions. A firewall or proxy may block that; download the new version from the website instead.
- AI Studio can't update itself, for example, as a Flatpak or when your organization manages it. "Information" → "How to update" explains what to do.
- "Your settings were created by a newer AI Studio version." You started an older version after a newer one. Install the latest update; until then, AI Studio doesn't save your changes.
For problems with the installation, see Installing AI Studio.
- On the "Information" page, note the version under "Versions".
- Under "Logbook", open the log files, or click "Open Log Viewer"; see Log Viewer. The logs contain no passwords, but they may contain file paths and other details of your work, so look through the lines you share.
- Use the link "Report your concern here" on the "Information" page, or open an issue on GitHub.
If your organization provides AI Studio, ask your IT department first.
Describes MindWork AI Studio v26.10.1 · Found a mistake? Open an issue
User guide
Basics
Chat
Assistants
- Overview
- General
- Business
- Learning
- Software Engineering
- AI Studio Development
More
Enterprise IT
Deployment
Configuring
Topics
- Providers and API Keys
- Confidence and Data Security
- Assistants and Templates
- Tools and Web Access
- Labeling and Usage Statistics
- Managing Backups
Scenarios and help
Developers
Deutsch