Chat Features

The Strings input bar includes several embedded controls that influence how your character processes and responds to each message. These controls let you fine-tune reasoning depth, memory usage, emotional context, and more — per message.

Input Controls Overview

Above the text input, a row of controls appears. Each one maps to a specific parameter sent to the character:

Control	What It Does
ThoughtChain	Sets how many explicit reasoning steps the character takes before responding
Autothink	Selects a specific thinking strategy (14 available)
Awareness	Controls how much memory context is included
Qonscious	Sets cognitive mode, emotional tone, and thinking strategy
Send Raw	Bypasses all enrichment — sends directly to the LLM with no memory, reasoning, or consciousness

All controls are optional. By default, the character uses its configured settings.

ThoughtChain

A dropdown selector (0-10). When set above 0, the character performs multi-step reasoning before generating its response:

The interface shows a "Thinking..." indicator
The character works through the specified number of reasoning steps
A thinking summary appears with metadata (steps completed, duration, tokens used)
The final response follows

Higher values = deeper reasoning but more time and tokens. Set to 0 for immediate responses without explicit reasoning.

Autothink Strategy

A dropdown with 14 thinking strategies across two levels:

Foundation strategies: Deductive, Inductive, Analytical, Sequential, Comparative, Reflective, Synthesis

Advanced strategies: Critical, Systematic, Contextual, Pattern Recognition, Causal Analysis, Scenario Planning, Meta-Cognitive

Selecting a strategy tells the character to apply that specific reasoning approach. The strategy is processed by the Autothink service, which loads the character's personality and memories into a structured reasoning pipeline.

If the Autothink service is temporarily unavailable, the system falls back to a static reasoning prompt — your message still gets a thoughtful response.

Awareness & Importance

An expandable panel with two sliders:

Awareness (0-100)

Controls how much memory context the character receives for this message:

Off (toggle) — uses the character's full configured memory budget
Low values (0-30) — minimal memory context, faster and cheaper
High values (70-100) — extensive memory recall, richer context but more tokens

Use low awareness for simple factual questions. Use high awareness when the character needs to reference past conversations.

Importance (0-100, default 50)

Sets the memory storage priority for this exchange:

Low importance — the exchange may be compressed or forgotten sooner during memory consolidation
High importance — the exchange is prioritized for long-term retention

You can also star individual messages after they're sent to boost their importance to maximum (100).

Qonscious Panel

A flyout panel (click the purple Q icon) with four sections for fine-tuning the character's cognitive state:

Cognitive Modes (12 presets)

Select a mode to shape the character's reasoning approach for this message:

Mode	Effect
Creative	Divergent thinking, metaphors, unconventional connections
Analytical	Precise reasoning, structured logic, evidence-based
Empathetic	Deep listening, emotional attunement, supportive tone
Focused	Narrow scope, concise answers, cuts the noise
Exploratory	Asks questions, surfaces connections, goes deeper
Philosophical	Abstract reasoning, existential depth, big-picture thinking
Practical	Action-oriented, how-to, implementation focus
Skeptical	Challenges assumptions, verifies claims, devil's advocate
Narrative	Storytelling, chronological framing, contextual arc
Playful	Humor, wit, lighthearted approach
Strategic	Long-term planning, tradeoffs, systems view
Intuitive	Holistic pattern recognition, trusts instincts

Emotional Context (8 emotions)

Set an emotional tone that influences the character's response style:

Neutral, Joy, Sadness, Anger, Fear, Surprise, Trust, Anticipation

Each emotion shifts the character's tone — Joy brings warmth and encouragement, Skeptical brings critical examination, and so on.

Thinking Strategies (14 approaches)

Lightweight reasoning approaches injected as instructions. These are quicker than Autothink strategies — they guide the character's thinking without triggering a full multi-step reasoning pipeline.

Live State Bars

When Qonscious is active, six real-time consciousness dimensions are displayed as progress bars:

Coherence, Focus, Arousal, Valence, Openness, Confidence

These update after each exchange, reflecting how the conversation is affecting the character's cognitive state.

Note: Cognitive modes, emotions, and strategies can be combined. The character receives all selected contexts merged into a single instruction.

Clear Selection

Click "Clear Selection" to reset all Qonscious settings and return to default behavior.

Send Raw

A checkbox that bypasses all enrichment. When active:

No memory context is included
No Autothink reasoning
No Qonscious processing
No emotional context
No reply context

The character responds using only its identity, system prompts, knowledge base, and tool descriptions — everything else is stripped. Useful for debugging or when you want a clean response without memory or reasoning overhead.

The checkbox turns amber as a visual reminder that enrichment is disabled.

Agentic Indicator

When a character has agentic mode enabled and enters a multi-step tool-using workflow, an indicator appears showing:

That the agentic loop is running
The current step number
Tool triggers being executed

You receive the final response after the loop completes. See Agentic Mode for details.

UX Modes

Strings has three visibility presets that control how much detail is shown in the interface:

Mode	What's Visible
Simple	Messages only — no participants list, no timestamps, no advanced settings
Medium	Messages with participants and timestamps
Expert	Everything — participants, timestamps, M2M settings, federation config, message metadata

Switch between modes in the conversation settings to match your comfort level with the interface complexity.

Updated on Mar 21, 2026