If you have not added an API key yet, see Use your own AI API keys. If a request fails while using your own API key, follow the checks below.
1. Confirm the AI source
First, confirm that the failed request was using your own API key rather than Heptabase AI or ChatGPT Plan. When you use your own API key, the provider manages model access, usage limits, and billing.
Keep in mind:
A ChatGPT subscription and OpenAI API billing have separate usage limits and billing.
A consumer Gemini plan does not provide quota for a Gemini API project.
Make sure you are checking the same account, project, or organization that created the API key.
An API key shown as Verified only means that Heptabase could use the credentials when they were checked. It does not mean that every model is available or that the account still has enough quota.
2. Follow the error message
Invalid, expired, or suspended API key
Open the AI provider's settings and confirm that the key is still active. If you cannot confirm it, create a new API key and reconnect it in Heptabase. Never send your API key to Heptabase support.
Rate limit, quota, or billing error
When you use your own API key, the AI provider sets rate limits based on the account, model, and usage tier. These are not Heptabase usage limits. Free and lower usage tiers usually have stricter limits, so even short requests may be rejected. Heptabase cannot increase or reset these limits.
Check the following in the AI provider's dashboard:
Usage
Rate limits
Billing or credit balance
The account, project, or organization that owns the API key
Whether the selected model is available on the current usage tier
Rate limits may be calculated separately for each project, model, and time window. If a limit has been reached, wait for it to reset, update the provider's billing or limits, or choose another available model. If the dashboard looks normal but the same error continues, contact Heptabase support.
Model unavailable, access denied, or organization verification required
Confirm that the account, project, or organization that owns the API key can use the selected model. You can:
Choose another available model
Check model access in the AI provider's dashboard
Complete organization verification only when the error specifically requires it
Confirm that the required API is enabled
Temporarily overloaded, unavailable, or internal error
These errors mean that the AI provider could not process the request at that time. They do not necessarily mean that the API key or Heptabase settings are incorrect.
When you use your own Gemini API key, Google decides whether to accept the request based on the usage tier, rate limits, and current service availability for that key. Free and lower usage tiers are more likely to encounter overloaded or rate limit errors. Heptabase cannot increase or reset these limits.
If the same request works with Heptabase AI, check the quota, billing, and usage tier for your Gemini API key in Google AI Studio. You can also try again later or temporarily use another model or AI source. If the error continues, contact Heptabase support.
Request too large, context window, or token limit
The conversation, attachments, or request may exceed the model's limits. You can:
Shorten the latest message
Remove large attachments
Start a new chat
Choose a model with a larger context window
Something went wrong
If you only see a general error message:
Try the request again.
Send a short message in a new, empty chat.
Try another available model.
Try another AI source, such as Heptabase AI.
If you can reproduce the problem consistently, contact Heptabase support.
Contact Heptabase support
Please include:
The complete error message or a screenshot
The AI provider and model
Whether you were using your own API key, Heptabase AI, or ChatGPT Plan
The approximate time and timezone
Whether the problem occurs in a new, empty chat
Whether another model or AI source works
Never include your API key.