Support configurable thinking output formats

[sestinj]

Validations

• I believe this is a way to improve. I'll try to join the Continue Discord for questions
• I'm not able to find an open issue that requests the same enhancement

Problem

Some LLM providers like vLLM let you change the output format of thinking. Right now Continue expects only the standard format from OpenAI, Anthropic, etc. but doesn't have a way of supporting arbitrary thinking tag formats.

For reference: https://docs.vllm.ai/en/latest/features/reasoning_outputs.html#streaming-chat-completions

Solution

No response

[Iamshankhadeep]

@sestinj I will be working on this issue next.

[vickywane]

@Iamshankhadeep did you get to work on this? -- I'm interested in giving it a go

[TheNha]

+1