Character chat setup
Chub provides the character and chat interface; Chutes can supply the model response when your chosen connection method supports it. This guide shows what to check before changing settings, how to test the connection, and where to look when replies fail.
A Chub character, a selected chat connection, and a model each play a different role. Identify which part you want to change before touching a working conversation.
You already have a character card and want to compare its greeting and voice across model responses.
Keep the same card and opening prompt while changing only the model connection. For another character-chat integration, see how to use chutes on janitor ai.
You have an ongoing Chub conversation and do not want a connection test to disrupt its context.
Start a separate test chat first, then return to the original thread if the responses look right. The guide to how to use chutes on janitor ai covers the equivalent check on a different chat surface.
A character loads in Chub, but sending a message produces an error or no useful reply.
Check the endpoint, authentication requirements, and model identifier separately. If you also use Janitor AI, how to use chutes on janitor ai explains its distinct connection path.
Chutes is the model-serving side of this arrangement, not the character library. The available connection fields and supported providers depend on Chub's current interface.
Character
Choose the character you want to test and note its opening message, instructions, and any chat-specific settings. A model change does not rewrite the character card. Keeping those inputs steady makes it easier to tell whether a different response came from the connection or from a changed prompt.
Connection
Look for Chub's external provider or compatible API setting. Enter connection details only when Chub offers fields that match the access method documented for the Chutes model you intend to use. Do not assume that a browser chat address is an API endpoint or that every provider option accepts the same format.
Model
Use the exact model identifier shown by the model provider rather than a display name copied from a conversation. Model availability can change. If Chub rejects the identifier, verify it against the provider's current documentation before changing unrelated character settings.
The character stays on Chub; the response path changes only if the selected provider configuration works.
| Existing Chub chat | Chub chat using Chutes | |
|---|---|---|
| Character card | Stored and selected in Chub | Still stored and selected in Chub |
| Chat interface | Read and send messages in Chub | Read and send messages in Chub |
| Response source | The currently selected chat provider | A compatible Chutes model, if connected |
| Connection fields | Existing provider settings | Endpoint, access details, and model identifier as required |
| First test | Send a message through the current provider | Send the same message in a separate test chat |
| Failure check | Inspect the existing provider setup | Inspect field format, access, model name, and provider compatibility |
A successful setup is a repeatable character-chat response, not merely a saved settings screen.
In Chub, inspect the chat's provider settings. Confirm that the offered external connection method matches the current instructions for the Chutes model you want to use; interface labels may vary.
Copy the documented endpoint, required access information, and exact model identifier into their matching fields. Never paste a secret into the character card, public description, or chat message.
Open a separate conversation with the chosen character and send a short prompt. A useful result is a visible reply that fits the character; an error means the connection still needs checking.
Use these checks to separate a character-writing issue from a connection issue without changing everything at once.
Test whether Chub can return a response through the selected Chutes connection before evaluating character quality. A brief greeting makes failures easier to identify.
Once the connection responds, repeat a prompt that asks for a recognizable detail from the character card. Compare the reply with the card's instructions, not with an unrelated model demo.
If replies stop, change one connection variable at a time. This helps distinguish a mistyped model identifier from an unsupported endpoint or missing access detail.
A short, repeatable character prompt gives you something concrete to test in Chub. Draft one now, then use it to check whether your selected Chutes model responds and follows the character's setup.
Keep the character selected in Chub and inspect that chat's provider settings. If Chub offers a compatible external connection, enter the details required for your chosen Chutes model and test it in a separate conversation. The character card itself does not need to be moved.
Use the endpoint format, access details, and model identifier documented for the specific Chutes connection you intend to use. Match each value to the corresponding field in Chub rather than guessing from a browser address. Available fields and supported methods can change.
Loading a character confirms that Chub can display the card; it does not confirm that the model connection works. Check the visible error, then verify the endpoint, required access, model identifier, and compatibility of the selected provider option. Retry with a short message after each correction.
Changing a chat's response provider should not require rewriting the character card. The replies may still differ because models interpret the same instructions differently. Test in a separate chat if you want to preserve an existing conversation.