Personality & Identity
Personality & Identity
Every character in Qui Anima has an identity — who it is, how it speaks, and what rules it follows. This is configured through the core node and the Definition node in the Visual Builder.
Identity
Click the core Anima node on the canvas to set the character's basic identity:
- Name — the display name used in conversations and across the ecosystem
- Description — a brief summary visible in character lists and selection menus
The name is important: the system tells the LLM "Your name is [name]" as part of every conversation. The character will introduce itself by this name and refuse to identify as Claude, ChatGPT, Gemini, or any other AI assistant.
System Prompts
Click the Definition node (square shape, green) to configure system prompts — the instructions that shape your character's behavior.
System prompts are the most powerful tool for defining character behavior. They tell the LLM what to do, how to do it, and what boundaries to respect.
Writing Effective System Prompts
A good system prompt covers:
- Role — what the character does ("You are a code reviewer specializing in Python")
- Behavior — how it responds ("Be concise. Lead with the answer. Use code examples.")
- Boundaries — what it should not do ("Never execute destructive commands without confirmation")
- Knowledge — domain-specific context ("The project uses FastAPI with asyncpg for database access")
Tip: Be specific. "Be helpful" is vague. "When the user asks about database queries, always show the asyncpg syntax with parameterized queries ($1, $2)" is actionable.
Multiple System Prompts
Characters support multiple system prompt entries. Each entry has a label (for organization) and content (the actual instruction). All entries are assembled into a single system prompt at runtime, separated by dividers.
Use multiple entries to organize instructions by topic:
- "Role & Personality" — who the character is
- "Coding Standards" — project-specific rules
- "Communication Style" — tone and formatting preferences
Human vs Self-Authored Prompts
System prompts have two sources:
- Human-authored — written by you in the builder. These are immutable — the character can never modify them, even if self-modifying is enabled.
- Self-authored — created by the character itself at runtime (requires the Self-Modify node). Tagged with
source: "self"and subject to rate limits and entry caps.
Both coexist in the same prompt list. Human entries always take precedence.
Personality Traits
The Definition node also lets you set personality traits that influence the character's communication style. These are descriptive keywords or phrases that get woven into the system prompt.
Examples:
- "analytical and precise"
- "warm and encouraging"
- "direct, no small talk"
- "uses technical jargon freely"
Traits provide consistent tonal guidance across all conversations without needing to spell out every aspect in the system prompt.
Token Limits
Click the Token Limits node (square shape, red) to configure how much the character can process and generate:
| Setting | What It Controls |
|---|---|
| Max Input Tokens | Maximum tokens in the conversation context sent to the LLM |
| Max Output Tokens | Maximum tokens in each response. Overrides the default per call. Also enforced by your billing balance — the system caps output to what you can afford. |
| Context Window | Total context window size for the model |
Setting appropriate token limits helps control costs. A character that only needs short answers can have a low max output (e.g., 500 tokens), while one that generates detailed analysis might need 4000+.
How Identity Flows Through the System
When a character generates a response, the system assembles a complete prompt from multiple sources in this order:
- Identity — name, description, personality traits
- Consciousness context — if Qonscious is enabled (coherence, emotional state)
- System prompts — all human-authored and self-authored entries
- Knowledge — relevant documents from the knowledge base
- Memory — recalled semantic memories from past conversations
- Emotional context — from the Qonscious UI indicator
- Deep reasoning — from Autothink strategies (if active)
- Tool descriptions — for each enabled capability node
This assembled context ensures the character responds consistently with its full identity, knowledge, and capabilities — every time, from every interface.
