Replies: 3 comments 1 reply
-
Claude recommends sending the reasoning, although in the format the sdk seems to place it it breaks and throws a zod validation error. |
Beta Was this translation helpful? Give feedback.
0 replies
-
Only claude 3.7 supports this (and it's optional). |
Beta Was this translation helpful? Give feedback.
0 replies
-
How about for Deepseek R1 and its various distilled thinking models. Do you
append the reasoning back for subsequent messages?
…On Tue, Mar 4, 2025 at 1:40 AM Lars Grammel ***@***.***> wrote:
Only claude 3.7 supports this (and it's optional).
—
Reply to this email directly, view it on GitHub
<#5005 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AOW4AKJRKLYNDQDXEABGJ4D2SVYKPAVCNFSM6AAAAABX7MD7LOVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTEMZYGYYTMNQ>
.
You are receiving this because you authored the thread.Message ID:
***@***.***>
|
Beta Was this translation helpful? Give feedback.
1 reply
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
I'm asking because sending reasoning text seems to be optional as you can opt out. Also different models encode the reasoning differently and there doesn't seem to be an agreed format to append it back. So my intuition sends no, but I want to make sure.
Beta Was this translation helpful? Give feedback.
All reactions