-
Notifications
You must be signed in to change notification settings - Fork 233
Add toolset to switch model #1539
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
base: main
Are you sure you want to change the base?
Conversation
Signed-off-by: David Gageot <david.gageot@docker.com>
|
|
||
| ### When to Consider Switching Models | ||
|
|
||
| **Switch to a faster/cheaper model when:** |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
just one thing here.. how is a given model supposed to know what is faster/cheaper or better at some task? should we just rely on the users giving their model definitions reasonable name?
| Use switch_model strategically: | ||
| - For simple tasks (formatting, basic Q&A, summaries), use the faster/cheaper model | ||
| - For complex tasks (code generation, analysis, reasoning), use the more powerful model | ||
| - The switch_model tool description shows available models and which one is current | ||
| - After completing a specialized task, consider switching back to the default model | ||
|
|
||
| Example workflow: | ||
| 1. User asks a simple question -> switch to 'fast' for efficiency | ||
| 2. User asks for complex code -> switch to 'powerful' for quality | ||
| 3. Task complete -> switch back to 'fast' |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
do we expect the example work decently even without this extra prompting, given all the instructions included in the tool? if not maybe we can reduce the amount of instructions, to avoid potential conflicts between user prompts and built-in instructions the users don't see?
mostly thinking about context pollution with smaller/local models and conflicting prompts
No description provided.