Prompt Blocks¶

Prompt blocks are the building pieces of your preset. Each block is a section of text that gets assembled into the final prompt sent to the AI.

Block Properties¶

Every block has these settings:

Property	Description
Name	Display label for the block
Content	The text (can include macros)
Role	Message role: `system`, `user`, `assistant`, `user_append`, or `assistant_append`
Enabled	Whether this block is included
Position	`pre_history`, `post_history`, or `in_history`
Depth	For `in_history` position — how many messages from the end to insert
Marker	Structural marker name (for built-in blocks like char_description)
Color	UI color for visual organization
Locked	Prevent accidental editing
Injection Trigger	Which generation types activate this block
Group	Group membership for block grouping

Block Types¶

Structural Markers¶

These are "magic" blocks that expand to character/persona/chat data:

Marker	Expands To
`char_description`	Character's description field
`char_personality`	Character's personality field
`scenario`	Character's scenario field
`persona`	Your persona's description
`mes_examples`	Character's example messages
`system_prompt`	Character's system prompt
`post_history_instructions`	Character's post-history instructions
`chat_history`	All chat messages
`world_info_before`	World book entries (before position)
`world_info_after`	World book entries (after position)

Content Blocks¶

Custom blocks you write yourself. These can contain any text and macros:

Write vivid, sensory descriptions. Show, don't tell.
Keep responses between 2-4 paragraphs.
Always stay in character as {{char}}.

Category Markers¶

Visual dividers in the block list — they don't produce output, just help organize your blocks in the editor.

Block Ordering¶

The order of blocks in the list determines their order in the assembled prompt. Drag blocks to rearrange them.

A typical order looks like:

System prompt
Character description
Personality
Scenario
Persona description
World info (before)
Example messages
Chat history
World info (after)
Post-history instructions
Author's note (injected at depth)

But this is just a starting point — experiment to find what works best for your use case and model.

Roles¶

The role determines how the AI interprets the block:

Role	AI Interpretation
`system`	Background instructions (highest authority)
`user`	Appears as a user message
`assistant`	Appears as the AI's own prior output
`user_append`	Appended to the previous user message
`assistant_append`	Appended to the previous assistant message

Most instruction blocks should use system role. Use user or assistant roles sparingly for specific effects (like fake dialogue examples).

Injection Triggers¶

By default, a block is included in every generation. You can limit it to specific generation types:

Trigger	When Active
`normal`	Regular message sending
`continue`	When continuing a response
`regenerate`	When regenerating a response
`swipe`	When generating a swipe
`impersonate`	When the AI writes as the user
`quiet`	Background/silent generations

Leave the trigger list empty to include the block in all generation types.

Block Groups¶

Blocks can be assigned to groups for collective behavior. Groups support two modes:

Radio — Only one block in the group is active at a time (selecting one deselects the others)
Checkbox — Multiple blocks can be active simultaneously

Groups are useful for creating alternative instruction sets (e.g., different writing styles) where you want to pick one at a time.

Tips¶

Start simple

Begin with just a system prompt block and the structural markers. Add custom blocks as you learn what influences the AI's behavior.

Use the dry run

After configuring blocks, use Dry Run to see the assembled prompt. This shows you exactly what the AI receives, including resolved macros and injected world info.

Disable, don't delete

If a block isn't working well, disable it instead of deleting it. You can re-enable it later or reference it when building other blocks.