GENERATION_TIMEOUT

What triggered it

Every generation runs against a hard 10-minute wall (configurable per tier — Studio gets 15 minutes). If the LLM hasn’t emitted its complete event by then, we cancel and return GENERATION_TIMEOUT.

Three things cause this:

• The model got stuck in a loop. Open-weight models occasionally re-write the same file 4-5 times instead of moving on.
• The prompt is gigantic. “Build a full economy + leveling + moderation + AI chat bot in one go” can take 12-15 minutes — too much for a single generation.
• Upstream provider is slow. OpenRouter sometimes queues requests when traffic spikes (e.g. weekends, post-launch days).

Retry costs nothing

Failed timeouts don’t consume your monthly generation quota. The retry button is free.

What to try

1. Hit retry. In the chat, hit Retry or just send the same prompt again. Most timeouts succeed on the second try.

2. Split the prompt. If the prompt is “build everything,” try:

• First generation: core /ping, /help, and welcome message handler
• Iterate: add moderation commands
• Iterate: add the economy

Each iteration is a fresh 10-min budget. Three sequential iterations beat one giant timeout.

3. Check the status page. If OpenRouter is degraded, we’ll have an active incident. Wait 5-10 minutes and retry.

4. Switch model. Settings → Model. If MiniMax M2.5 keeps timing out on a specific prompt, try Z.ai GLM-5 or Kimi K2.5 — different model behaviors.

Back to the app

When to contact support

If retries keep timing out on simple prompts and the status page is green:

• Drop a screenshot of the chat in Discord
• Or email support@vibecord.dev with your project name and approximate timestamp

We can pull the generation logs and tell you exactly where it got stuck. Usually we can re-run with a different model and ship you the result manually.