deepseek-reasoner does not support successive user or assistant messages

[freeferg]

Describe the bug

AI_APICallError: deepseek-reasoner does not support successive user or assistant messages (messages[2] and messages[3] in your input). You should interleave the user/assistant messages in the message sequence.

Link to the Bolt URL that caused the error

localhost

Steps to reproduce

1. Add Deepseek-reasoner as your LLM of choice
2. Send a prompt

Expected behavior

As bolt.diy now supports reasoning models and deepseek reasoner is already added to it by default, I would expect it to work from the start.

Screen Recording / Screenshot

No response

Platform

• OS: macOS
• Browser: Chrome Canary
• Version: Latest

Provider Used

Deepseek API

Model Used

deepseek-reasoner

Additional context

No response

[waylonkenning]

INFO stream-text Sending llm call to Deepseek with model deepseek-reasoner
DEBUG api.chat usage {"promptTokens":null,"completionTokens":null,"totalTokens":null}
DEBUG api.chat Total message length: 3436, words
INFO stream-text Sending llm call to Deepseek with model deepseek-reasoner
ERROR api.chat AI_APICallError: deepseek-reasoner does not support successive user or assistant messages (messages[9] and messages[10] in your input). You should interleave the user/assistant messages in the message sequence.
DEBUG api.chat Total message length: 3436, words

[Nojahhh]

Experience the same issue with deepseek-reasoner

[vavanv]

the same here

[thecodacus]

try using the context optimization option

[AuroraGeeker]

Modifying the value of the Base URL makes it work. When setting the API, set the base_url to 'https://api.deepseek.com' and don't add '/v1' after it.