Bot Configuration

Every bot in AI Supreme Council is defined by a configuration object that controls its provider, model, behavior, and appearance. This page documents the full configuration schema and explains each field.

The Config Panel

The config panel slides out from the right side of the screen. It is visible by default on desktop and accessible via the settings icon on mobile. The panel contains:

• Provider and model selectors at the top
• System prompt text area
• Generation parameters (temperature, max tokens, etc.)
• Advanced settings (expandable section)
• Persona settings (icon, description, color)
• Per-bot API key field

Changes are saved automatically as you edit.

Core Fields

Provider

The LLM provider that hosts the model. Select from the dropdown:

Provider	API Key Source	Notes
Anthropic	console.anthropic.com	Claude models -- advanced reasoning
OpenAI	platform.openai.com	GPT-4o, o3, DALL-E
xAI	console.x.ai	Grok models
Google Gemini	aistudio.google.com	Free Flash tier available
OpenRouter	openrouter.ai	300+ models, free tier
Groq	console.groq.com	Ultra-fast LPU inference
DeepSeek	platform.deepseek.com	R1 and V3 reasoning models
Mistral	console.mistral.ai	Large, Small, Codestral
Ollama	N/A (local)	Runs locally, no API key needed
Custom	User-defined	Any OpenAI-compatible endpoint

Changing the provider updates the model dropdown to show available models for that provider.

Model

The specific model from the selected provider. The dropdown is populated from the community model registry and shows:

• Model name and ID
• Pricing tier (free or paid)
• Context window size
• Capabilities (vision, reasoning, tools)

Model info card

When you select a model, an info card appears below the dropdown showing pricing (input/output per million tokens), context window size, max output tokens, and capability tags. This helps you choose the right model for your task.

System Prompt

Instructions sent to the model at the beginning of every conversation. Defines the bot's behavior, personality, and constraints. See System Prompts for detailed guidance and templates.

The system prompt field includes a template dropdown with 5 built-in presets. Select a template and customize it to your needs.

Temperature

Controls the randomness/creativity of responses.

Value	Behavior
0	Deterministic -- always picks the most likely token
0.3	Conservative -- focused, consistent responses
0.7	Balanced (default) -- good mix of creativity and coherence
1.0	Creative -- more varied responses
1.5-2.0	Very creative -- unpredictable, experimental

Max Tokens

Maximum number of tokens the model can generate in a single response. Range: 128 to 16,384 (or higher for models that support it). Default: 4,096.

tip

Setting max tokens too low may cause responses to be cut off mid-sentence. Setting it too high may increase costs for paid models. 4,096 is a good default for most use cases.

Advanced Fields

Expand the Advanced section in the config panel to access these parameters.

Reasoning / Thinking Effort

For models that support extended thinking (Claude, GPT o-series, DeepSeek R1, Gemini thinking models), this controls how much computation the model spends reasoning before responding.

Value	Description
Off	No extended thinking (standard response)
Low	Minimal reasoning, fastest responses
Medium	Moderate reasoning depth
High	Deep reasoning, more thorough
Highest	Maximum reasoning effort
Custom	Set a custom budget token count

note

Not all models support reasoning/thinking mode. The field only appears when the selected model or provider supports it. Models known to support reasoning include Anthropic Claude (extended thinking), OpenAI o1/o3 series, DeepSeek R1, xAI Grok, and Google Gemini thinking models.

Top P (Nucleus Sampling)

Controls the cumulative probability threshold for token selection. Default: 1.0 (no filtering).

• 1.0 -- consider all tokens (default)
• 0.9 -- consider tokens comprising the top 90% of probability mass
• 0.1 -- very restrictive, almost deterministic

warning

It is generally recommended to adjust either Temperature or Top P, not both at the same time. Changing both can produce unpredictable results.

Frequency Penalty

Reduces the likelihood of the model repeating tokens that have already appeared. Range: 0 to 2. Default: 0.

• 0 -- no penalty (default)
• 0.5 -- mild repetition reduction
• 1.0-2.0 -- strong repetition reduction

Presence Penalty

Encourages the model to talk about new topics by penalizing tokens that have appeared at all. Range: 0 to 2. Default: 0.

• 0 -- no penalty (default)
• 0.5 -- mild topic diversity encouragement
• 1.0-2.0 -- strong push toward new topics

Seed

An integer value for reproducible outputs. When the same seed, prompt, and parameters are used, the model attempts to return the same response. Not all providers support this feature.

Stop Sequences

A comma-separated list of custom stop tokens. When the model generates any of these sequences, it stops producing further output. Useful for structured output or role-playing scenarios.

Example: END,---,</answer> causes the model to stop when it generates END, ---, or </answer>.

Response Format

Controls the output format:

Value	Description
text	Standard text output (default)
json	Forces JSON output (model must support structured output)

note

JSON mode requires the system prompt to instruct the model to produce JSON. Simply setting the response format to JSON without mentioning it in the prompt may result in errors or malformed output.

Tool Calling

Configure function/tool calling for models that support it. This is an advanced feature for building bots that can invoke external tools or structured actions.

Per-Bot API Key

Each bot can use a different API key from the global key:

1. Find the API Key field in the config panel (below the provider selector)
2. Enter a key specific to this bot
3. This key overrides the global key for this bot only

Per-bot keys are stored locally and are never included in shared bot URLs.

Config Short Keys

Internally, bot configurations use short keys for efficient URL compression. This is useful if you are building integrations or debugging shared URLs:

Short Key	Full Name	Type	Default
n	Name	string	"New Bot"
p	Provider	string	"anthropic"
m	Model	string	--
s	System Prompt	string	""
t	Temperature	number	0.7
x	Max Tokens	number	4096
tp	Top P	number	1.0
fp	Frequency Penalty	number	0
pp	Presence Penalty	number	0
se	Seed	number	--
re	Reasoning Effort	string	--
st	Stop Sequences	string	--
rf	Response Format	string	"text"
tc	Tool Choice	string	--
tl	Tool List	array	--
pi	Persona Icon	string	--
pd	Persona Description	string	--
pc	Persona Color	string	--

Council Config (c key)

When the bot is a council (2+ members), the c key contains:

Short Key	Full Name	Type	Default
cs	Council Style	string	"council"
ms	Members	array	--
ch	Chairman	number	0
vm	Voting Mode	string	"weighted"
sr	Skip Review	boolean	false
mx	Max Tokens (per member)	number	1024
dr	Deliberation Rounds	number	2

Chat Settings

These per-bot settings control the chat experience:

Setting	Short Key	Description	Default
Context Limit	cl	Max messages sent to the API (0 = unlimited)	Unlimited
Streaming	sm	Stream tokens in real-time	Enabled
Auto-Title	at	Auto-generate conversation titles	Disabled
Markdown	mr	Render markdown in responses	Enabled
Show Token Count	stc	Display token usage per message	Disabled

Editing Raw Config

The config panel provides a visual editor for all fields. You do not need to edit raw JSON under normal use. However, the internal format is documented here for developers building integrations, custom providers, or tools that generate bot configurations programmatically.

Example raw config:

{
  "n": "Code Assistant",
  "p": "anthropic",
  "m": "claude-sonnet-4-20250514",
  "s": "You are an expert software engineer. Provide clear, well-documented code with explanations.",
  "t": 0.3,
  "x": 8192,
  "pi": "💻",
  "pd": "Software engineering assistant",
  "pc": "#4a90d9"
}