Guides
Request vs Project Settings
Many reply settings can be set in two places:
- In the request, with
replyParameterson each Enqueue Conversation call. - In the project, on your project settings page in the dashboard.
When both are set and disagree, this page tells you which one applies.
The rule
For each reply, DumGum picks the first value it finds:
- The value sent with the request. It always wins,
falseincluded. - Your project setting, when the request leaves the field out.
- The platform default, when neither sets it.
A value sent with a request applies to that reply only. It never changes your project settings.
Settings you can set in both places
| What it controls | In the request | In the project |
|---|---|---|
| Several messages per reply | multipleMessagesPerReply | Allow Multiple Messages Per Reply |
| Calm tone with rude users | preventAggressiveTone | Prevent Aggressive Tone |
| No agreeing with complaints about your service | serviceLoyalty | Service Loyalty |
| How far the Persona goes sexually | personaProfileOverrides.sexualDrive | Project configuration |
| Photo sharing | pictureSharingMode | Picture Sharing |
| Which guardrails run | guardrails | Core Guardian |
| Language rewrites | postProcessing | Language-specific formatting |
| Photo, video and voice analysis | vision, audio | Your project's capabilities |
Projects created in the dashboard start with Allow Multiple Messages Per Reply and Prevent Aggressive Tone on. If neither the request nor the project sets sexualDrive, the Persona's SoulEngine personality decides, and NORMAL applies when there is none.
Settings that are not in this table exist in one place only, so they cannot conflict.
Three cases to watch
Objects replace, they don't merge
guardrails and postProcessing are objects. Sending one replaces your whole project setting for that reply. Anything you leave out of the object takes the platform default, not your dashboard value.
For example, if you turned Malicious Content on in Core Guardian and a request sends guardrails without it, Malicious Content is off for that request. Send the complete object, or leave it out to use your project setting.
vision and audio work the other way: each field is read on its own, and a field you leave out follows your project.
Sending pictureSharingMode turns photo sharing on
pictureSharingMode sets how willingly the Persona shares photos (see Control Sharing and Sexual Intensity), but sending it also turns sharing on for that reply, even if Picture Sharing is off in your project. The Persona still needs photos in shareablePictures to send one. To keep photos off, leave pictureSharingMode out.
A dashboard switch that seems to do nothing
You switch Allow Multiple Messages Per Reply off, and replies still arrive in several messages. The cause is almost always a request that sends "multipleMessagesPerReply": true:
| Your request sends | Project option | Result |
|---|---|---|
true | on or off | The reply may be split |
false | on or off | Always one message |
| nothing | on | The reply may be split |
| nothing | off | Always one message |
The same goes for every setting in the table above: check what your requests send before changing the dashboard.
Check what a reply used
Open the reply in the Response Explorer. Its detail panel shows the reply parameters behind it, so you can see what your integration actually sent.