Knowledgebase

When a request is refused, the answer carries an HTTP status and a short code. A refused request costs no credits.

• 401 invalid_api_key: the key is missing, mistyped or replaced by a new one. Copy it again from the Virsilipas AI key page. If you got a new key, update ~/.virsilipasKey and your apps.
• 402 insufficient_quota: your balance cannot cover the request. A request needs more balance to start than it finally costs, so a long prompt can be refused while some credits remain. Shorten it, or wait for your next paid invoice to add credits.
• 403 key_disabled: AI credits are switched off for your account. Contact support.
• 404 model_not_found: the model name is not an alias. Use one from the model list.
• 404 not found: a wrong path. Check that the base URL is https://virsilipas.mcx.fi/v1.
• 400 unsupported_content: the request had an image, audio or file. Send text only.
• 400 unsupported_tool: the app asked for a tool that runs on a vendor's side, such as web search. Turn that tool off in the app.
• 400 context_length_exceeded: the conversation is longer than the model accepts. Start a new one, or pick a model with a longer context.
• 413: the request body is too large.
• 429: too many requests at once or in a short time, or the service is busy. Wait the number of seconds in the Retry-After header and try again. Many apps and libraries retry by themselves.
• 503 server_busy, spend_cap_reached or upstream_unavailable: the service is busy or cannot reach a model right now. Wait the Retry-After seconds and try again.
• 502 upstream_error or 504 upstream_timeout: the model failed or took too long. Try again a little later.

Reporting a problem

Open a ticket with the time (UTC), the model, the status and code, and the X-Request-Id header of the answer if you have it. Never send your key.

Every error with its exact message: API reference.