No profiles yet
Recalled memories:
Tools used:
View details
tokens ( prompt + completion)
Memory Browser
View and manage memories
No memories found
Custom Invocations
Create shortcuts for common interactions or special responses
No custom invocations yet
Invocations are custom shortcuts that trigger specific response styles. Use them for greetings, check-ins, or any interaction you want to personalize.
Create your first invocation below to get started.
Create Custom Invocation
Settings
These settings are for:
Changes are automatically saved. Each model can have its own settings.
Providers
Configure your inference providers (Anthropic, OpenAI, Ollama, etc.) to use with your models.
Legacy Configuration Detected
You're using the legacy single-provider configuration. Migrate to the multi-provider system to use multiple inference providers and better organize your API connections.
No providers configured
Add a provider to connect to AI inference APIs like Anthropic, OpenAI, or local models via Ollama.
|
To use a provider with a model, set the provider_id in Profile model pools or model settings.
Unique identifier for this provider (lowercase, no spaces)
Use one of the options below, or leave both empty for local providers.
Will use $
No API key variables found. Set ANTHROPIC_API_KEY, OPENAI_API_KEY, etc. in your shell.
Custom Instructions
Define how the assistant should behave. This is your system prompt for .
Identity
Memory Types
Memory types define how information is structured and stored. Built-in types cover common use cases; you can also create custom types.
Hidden Types
Style Profile
Style profiles define voice, tone, and behavioral patterns.
Built-in styles: minimal, professional, creative
Style:
Semantic Search
Enable embeddings for meaning-based search across memories and conversations.
Requires sentence-transformers for local mode
Default: intfloat/e5-small-v2 (local) or text-embedding-3-small (OpenAI)
Coverage
Generate embeddings for memories and messages that don't have them yet.
Clear all embeddings (useful when switching models). You'll need to regenerate them after.
Model Configuration
Include memories in conversations
Gradually fade unused memories
Uses the default provider configuration
Per-Profile Memory Isolation
Each profile maintains its own isolated memory space. Shared memories are visible to all profiles.
Each profile gets its own memory space
New memories are shared across profiles by default
Memory Distribution
Active profile:
Memories created now will be scoped to this profile unless marked as shared.
Advanced Settings
Fade Unused Memories
How it works:
Gradually fades unused memories over time. Memories you access frequently stay strong, while rarely-used ones slowly become dormant.
When enabled, memories that haven't been accessed will gradually fade.
Save Configuration
Save your current settings to a file for persistence across sessions.
Statistics
Import Data
Import conversation history, profile configurations, or add text as memories.
Import Conversation History
Import your past conversations from Claude (conversations.json) or ChatGPT (conversations.json/.zip).
Importing to profile:
No profile selected - conversations will be shared across all profiles
Export from Claude settings:
- -
claude-conversations.zip - -
conversations.json
Export from ChatGPT settings:
- -
chatgpt-export.zip - -
conversations.json
Import successful!
- Source:
- Conversations:
- Messages:
- Projects:
- Scoped to profile:
- Shared across all profiles
Import failed
Import Memories
Add text files or paste content to create new memories. Separate each memory with a blank line.
Imported memories successfully.
Import Profile Configuration
Import a Threadlight profile exported from another instance. This creates a new profile with the imported settings.
Profile imported!
Profile "" has been created.
Import failed
Profiles
Active Now
Create Your First Profile
Profiles let you customize how the AI responds. Set different personalities, styles, and memory settings for different use cases - like work, creative writing, or casual chat.