Connection guide
Chutes can provide a model endpoint while Janitor AI provides the chat interface, but the connection depends on the settings each service currently exposes. Use this guide to check compatibility, test one reply, and avoid placing private information in a test prompt.
The confusing part is separating the chat interface from the model that answers it. These starting points address different reasons a Janitor AI conversation may stall.
You see provider and model fields but are unsure which service belongs in each one.
Identify the interface, provider, and model before changing a chat setting. The Chub AI guide offers a second integration example for comparison.
A provider appears available, but you do not know whether the selected model suits dialogue.
Check the model's stated capabilities rather than assuming every endpoint handles a character conversation alike.
The connection saves, yet the first character reply does not follow the scene.
Test a short, neutral exchange before adding a long character setup or complex instructions.
A reply fails and it is unclear whether the endpoint, model selection, or chat settings caused it.
Isolate one variable at a time instead of repeatedly resending the same conversation.
Janitor AI and Chutes have different jobs. Compare the fields you can inspect before treating a saved configuration as a working conversation.
| Janitor AI chat interface | Chutes model endpoint | |
|---|---|---|
| Primary role | Displays characters, messages, and conversation controls. | Processes a compatible request and returns a model response. |
| What to check | Whether its current settings permit a custom or compatible provider. | Whether the chosen model exposes an endpoint suited to text chat. |
| Provider setting | May ask for provider, endpoint, or model details; available fields can change. | Supplies connection details only where the selected endpoint supports them. |
| Model selection | Uses the model identifier entered or selected in the chat settings. | Defines the model identifier and supported request behavior. |
| Character context | Holds the visible character setup and conversation flow. | Receives whatever context the interface sends with a request. |
| First test | Shows whether a reply appears in the conversation. | Shows whether the endpoint accepted the request and produced text. |
| If a reply fails | Inspect saved fields and any visible error message. | Check endpoint availability and whether the selected model matches the request. |
Start with the smallest possible test. Add character context only after a neutral message gets a usable response.
Open the current connection settings in Janitor AI and look for a supported way to specify an external provider or endpoint. If no suitable option is present, do not assume a Chutes model can be attached through that interface.
If a compatible route exists, take the endpoint and model identifier from the model's current documentation. Put each value in its corresponding Janitor AI field, then review the saved configuration for spelling or formatting errors.
Send a short fictional prompt, check that the response is coherent, and only then try a character exchange. If it fails, change one setting at a time and note whether the error occurs before or after a response begins.
A useful first test asks for a brief, constrained reply. You are checking that the model responds through the interface, not judging an entire character from one message.
The images illustrate the test stages; they are not screenshots of connection settings. A suitable text check would be a concise book recommendation that follows the requested two-sentence format. A blank reply or provider error calls for connection troubleshooting, not a longer prompt.
Before: untested connectionAfter: sample reply receivedFor how to use chutes on janitor ai in practice, follow the branch that matches what you can actually see. Menu names and supported connection methods may change.
Treat the visible fields as a checklist, not proof that every Chutes model will work. Match the endpoint and model identifier to current documentation, save the settings, and make one small request.
The connection may be functioning even when the dialogue is weak. Compare a neutral instruction with a short character scene, then adjust the scene rather than changing the endpoint immediately.
Separate connection failure from prompt failure. Preserve the visible error text, check that the configured model still exists, and retry once with a minimal message.
If your goal changes from connecting Janitor AI to testing another surface or choosing a model, these guides address those separate decisions.
A prompt sent through Janitor AI to a Chutes endpoint may pass through more than one service. Do not paste personal details, private character histories, secrets, or someone else's messages into a connectivity test. Review the current terms and privacy information for both services before using sensitive material, and follow each service's content rules. A successful sample reply confirms only that this particular request worked; it does not establish what either service retains or guarantee that every model and chat setting is compatible.
Check Janitor AI's current connection settings for a compatible external provider or endpoint option. Then compare the fields it requires with the current details for the Chutes model you intend to use; the presence of a field alone does not guarantee compatibility.
Send a short fictional message with a clear constraint, such as asking for a two-sentence recommendation. If a reply appears, try a brief character exchange next so you can distinguish connection success from dialogue quality.
A saved setting can still contain an incorrect endpoint or model identifier, or the selected model may not accept the request the interface sends. Read any visible error, recheck the current model details, and retry with a minimal prompt while changing one setting at a time.
Do not assume every listed model supports the same type of request or produces suitable dialogue. Check the chosen model's stated capabilities and test it through the connection method Janitor AI currently provides.
A message sent through an external model connection may be processed by both the chat interface and the model service. Use fictional text for setup tests and consult both services' current privacy information before sharing sensitive content.