Settings

Platform configuration and data overview

Import Settings

Videos per channel limit

Cap the number of videos imported per channel. Useful for testing. Set to 0 for no limit.

video_settings

Chunking Model

Used to split video transcripts into topic chunks during import. Cheapest model recommended — quality difference is minimal for this task.

Chat Model

Language Model

Used for all chat responses via OpenRouter

smart_toy

Current model

Embedding Model

Used for semantic search via OpenAI

text-embedding-3-small

Chunking Model

Used for splitting transcripts into topic chunks

Claude Haiku 4.5

Retrieval Strategy

70% semantic, 30% keyword — top 12 chunks

Hybrid (Vector + BM25)

Knowledge Base

Chat Defaults

Default Mode

Can be changed per session in the chat interface

Learn

Max History

Prior messages included for conversation context

20 messages

Streaming

Responses stream token by token

Enabled

Data Transfer

Export all personas, sources, and knowledge chunks as a JSON file. Import on another instance to sync data.

Danger Zone

Rebuild All Knowledge

Deletes all knowledge chunks and re-processes every persona from their existing transcripts.