Category: Participant Management
Welcome back to the MiocAI documentation series. We have covered setting up your character and navigating the interface. Now it is time to tackle one of the most dynamic and complex features on the
MiocAI هو بديل لـ Character.AI و Candy AI وتطبيقات الدردشة الأخرى مع شخصيات الذكاء الاصطناعي.
Welcome back to the MiocAI documentation series. We have covered setting up your character and navigating the interface. Now it is time to tackle one of the most dynamic and complex features on the platform: Group Chats.
Putting multiple AI characters into a single room completely changes how the backend processes information. The system no longer assumes a simple back-and-forth dialogue between you and a bot. Instead, it must juggle multiple personalities, private memories, and speaking orders.
If you just throw five characters into a chat and hope for the best, the narrative will likely derail. Here is exactly how group chats function under the hood and how you can manage your participants to keep the story cohesive.
To understand how to manage your participants, you need to understand what the AI actually sees when you hit send.
In a standard 1-on-1 chat, the backend formats the conversation as a simple exchange between "user" and "assistant". In a group chat, the formatting is entirely different. The backend script explicitly attaches the author's name to every single message.
If you say "Hello" and Character A responds, the AI model receives the text exactly like this:
YourName: Hello
Character A: How are you doing?
Furthermore, the system dynamically alters the core prompt for whoever is speaking. It automatically injects a specific instruction: "This is a group chat. Other participants include: [Names of other bots]." This is how the AI knows who else is in the room and prevents it from hallucinating random people.
You manage your roster via the Participant Manager, located in the Chat Settings panel.
This menu displays every character currently in the chat. You will see their avatar, their name, and a status badge indicating whether they are an AI or a human user.
You have three main controls here: * Add Character: The button at the bottom allows you to pull a new character into the existing chat. * Remove Button (Trash Can): Permanently kicks the character out of the group chat. * Enable/Disable Toggle: This is your best tool for crowd control. If a character falls asleep in the story or leaves the room, you do not need to delete them. Simply toggle them off. The backend will temporarily exclude them from the "Other participants include" list, stopping the active bots from interacting with them until you toggle them back on.
If you have three bots in a room, the system needs to know who should reply to your message. By default, MiocAI uses a "Next Turn" queue system, rotating through the active participants.
However, you have absolute control over this order. You do not have to wait for the rotation.
The Next Turn Preview: At the bottom of the chat interface, you will see a small indicator showing which character is queued to respond next. If you want a different character to reply to the current context, click this preview. A dropdown will appear, allowing you to select the exact bot you want to speak.
The Send Override (Mobile/Desktop Shortcut): Alternatively, if you have typed out a message and want to direct it at a specific bot, hover over the Send button (or long-press it on mobile). This opens the Send Override Popover. Clicking a bot's avatar here will send your message and instantly force that specific character to generate the reply.
When you force a specific participant to reply, the backend halts its normal flow, loads that specific character's JSON file, builds their unique system prompt, and executes the generation exclusively from their perspective.
One of the most advanced features on MiocAI is how it handles cross-chat memory.
In your Group Chat settings, there is a toggle labeled Include Private Memories. If you have this enabled, the system does something highly complex during the generation phase. Before the chosen bot speaks, the backend quietly looks up your private, 1-on-1 chat history with that specific character.
It pulls the most recent messages and memory triggers from your private chat and silently injects them into the bot's groupchat memory.
This means if you are secretly plotting against the group with Character A in your private chat, Character A will remember that plot while talking in the group chat. The other bots will remain completely unaware, as the backend only injects private context for the specific participant who is currently generating a response.
If you are on a Premium plan or higher, you must utilize the Character Relations menu in the settings panel.
Without defined relationships, bots treat each other based solely on their core persona instructions, which can lead to generic interactions. By adding a relationship, you hardcode their dynamic into the system prompt.
When you set Character A and Character B as "Enemies" with an intensity of 8, the backend intercepts the prompt generation and injects a strict rule: "[Character A] is Enemies with [Character B]." The AI is forced to process this dynamic before it generates a single word, ensuring that participants interact with each other accurately instead of just focusing on you. Do note that you have to manually update these as we dont yet automatically update them.
While text generation in group chats is highly sophisticated, image generation is a completely different beast.
Image generation in group chats is currently hit or miss.
Image models (like SDXL or FLUX) are exceptionally good at generating a single subject based on a strict appearance prompt. When you generate an image in a group chat, the system struggles to perfectly blend multiple complex appearance prompts together. Ive previously pointed this out in our Act 1 of character creation, but this is due to constraints with the tokenizers of many models.
You may experience scenarios where: * The image only features the character who just spoke. * The characters' traits bleed into each other (for example, a character with blue hair and a character with red hair might result in two characters with purple hair). * The positioning is chaotic.
While we are constantly working to improve multi-character image rendering, you should expect less consistency with images in group chats compared to your 1-on-1 conversations. If you absolutely need a perfect image of a specific character, it is best to generate it in their private chat or the Studio.
Thats all for this guide! Hop on in with your favorite characters and... hold a meeting or something!