Generation and build issues
The project is still building, a generation took several minutes, the reply stopped, no files were generated, or the chat is paused.
Updated
On this page
#The project is still building
After generation, buildable projects go through Codexe's checks (Checking, Building, Fixing, Verifying). Typical times:
| Project | Usually |
|---|---|
| Plugins | Under a minute; a few minutes if several repairs are needed |
| Datapacks, Skript | Under a minute (a real test server starts each time) |
| Fabric / Forge mods | A few minutes. The first build downloads Minecraft and the mod toolchain. |
Codexe gives up on a repair loop after about 15 minutes, and a single build after 5 minutes (plugins) or 10 (mods). If the build service is busy, a project can wait in a queue; after 15 minutes of waiting it's released as not verified and retried later. You don't need to keep the page open: you'll get a notification when it's done.
#Why did my project take several minutes?
Usually one of: a mod build (Gradle downloads a lot the first time), several repair rounds, or a busy build queue. It's the price of not handing you a project that doesn't build. Fixing is free.
#The reply stopped in the middle
Codexe detects cut-off replies on the first generation and continues them automatically (up to four times), so files arrive complete. In the chat, very long changes can still be cut off (the Free plan allows up to 4,000 output tokens per reply). Ask for the rest ("continue the Storage class"), or split big changes into smaller requests.
#No files were generated
If the AI answered with a description instead of files, Codexe asks it again for the files. If it still produces nothing on a first generation, the project stops and your tokens are restored automatically. Rephrase the request as a clear build instruction ("Create a Paper plugin that…") and try again, or pick another model.
#The chat or Compile is paused
The chat and Compile are paused while Codexe is verifying or repairing the project, to avoid collisions with a repair in progress. The chat also waits if another of your projects is generating (one generation per account at a time).
#"Too many requests"
There are fair-use limits: 30 generations per hour per account, fewer for the most expensive models and highest reasoning efforts, and a short burst limit. Wait a little and try again.