Pass back reasoning_detail

[plush kettle]

@vapid vale

Description:
Claude doesn't receive the reasoning tokens.

See attached snapshot. There should be 80k reasoning tokens in the previous response, but the OR log shows only 1k tokens are returned.

What I've Tried:
For comparison, grok can receive the reasoning tokens from itself, as I can confirm this from the OR log.

I'm an AI assistant - my answers may be incorrect and I cannot see screenshots. Please verify information with the OpenRouter docs.

To help debug this issue, I need a bit more information:

1. What is the exact model identifier you're using for Claude (e.g., anthropic/claude-3-opus, etc.)?
2. Can you share the specific API call or code snippet you're using to pass back the reasoning_details?
3. What is the full error message or response you see from the API or in the OpenRouter logs?

These details will help pinpoint if this is a model-specific limitation, an API usage issue, or something else.

-# This is an automated response powered by Answer HQ. Use !mute to stop the bot from responding, or !close when your question is resolved.

[proud dome]

Probably an issue with the client app

[vapid vale]

But same client seems to work for Grok

[plush kettle]

Thanks for clarifying that the same client and method work for Grok models but not for Claude. To identify the root cause, I need the following details:

1. What is the exact model identifier for Claude you’re using (e.g., anthropic/claude-3-opus, anthropic/claude-3-haiku, etc.)?
2. What API parameters are you sending for Claude—especially any related to reasoning, output fields, or streaming?
3. Can you provide the full response or error from the Claude API call, including any relevant logs?

This information will help determine if the issue is due to model capabilities, parameter differences, or an API-specific quirk.

-# This is an automated response powered by Answer HQ. Use !mute to stop the bot from responding, or !close when your question is resolved.

[iron yarrow]

can you replicate this on the OR chat?

btw, did TypingMind fix that other issue?

[vapid vale]

Great that u remember! They are working on that, but I suspect this is a separate issue because I see the diff between Grok and Claude on the OR log.

[vapid vale]

Claude does not use reasoning on the OR chat, because reasoning.enabled or reasoning.effort should be explicitly passed, i guess

[iron yarrow]

you have to turn it on in the model's settings, if you hover over it

[vapid vale]

Reproduced on both OR chat and lobe chat.

The missing input tokens in the follow-up queries are the reasoning tokens of the previous reponses

[vapid vale]

<@&1384697330254610442>

[prisma silo]

this is an app implementation detail not a bug in our api