Conversation and Context Management

Split conversations by task

Don't put every question in one thread.

• Thread A: Product planning
• Thread B: Code implementation
• Thread C: Copywriting

Splitting reduces context contamination.

Use projects to organize conversations

Create projects and group related conversations under them. Projects support descriptions and instructions that are inherited by conversations inside them.

Group chats

Create a group chat and add multiple bots (each can use a different model). Send a message and each bot responds independently. Useful for comparing model outputs or having different-role bots collaborate.

• Each bot decides independently whether to reply
• Per-member prompts are supported
• Supports quoted replies

Favorites

Favorite any AI reply from its more-actions menu for easy access later.

Conversation branches

Create a branch from any message to explore different directions without affecting the original thread.

Tips for long conversations

1. One goal per message — don't pack too many requests into one prompt.
2. Repeat important constraints, like "must output JSON".
3. When context gets too long, start a new conversation and paste a brief summary in the first message.

Prompt structure

This order tends to work well:

1. Background (what you're doing)
2. Target output (what you want)
3. Constraints (format, length, style)
4. Example (optional)

Local data and privacy

BotHub stores your data locally by default. When signed in, conversations and settings (including API keys) sync to the cloud. When you call an online model, message content is sent to that provider. Check provider policies before sending sensitive data.