replace v1/completions with v1/chat/completions

[aseelert]

Currently the legacy v1 API calls are used, so there are a lot of limits in terms of modes and the TPM of them.
https://platform.openai.com/docs/api-reference/completions/create

there should be a check for endpoint https://api.openai.com/v1/completions or https://api.openai.com/v1/chat/completions

as currently the SDG fails with just using the newer type:

packages/instructlab/sdg/pipeline.py", line 203, in _generate_single
    raise PipelineBlockError(
instructlab.sdg.pipeline.PipelineBlockError: 
PipelineBlockError(<class 'instructlab.sdg.llmblock.LLMBlock'>/gen_contexts): Error code: 400 - 
{'error': {'message': 
"Missing required parameter: 'messages'.", 
'type': 'invalid_request_error', 
'param': 'messages',
'code': 'missing_required_parameter'}}

curl https://api.openai.com/v1/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  -d '{
    "model": "gpt-3.5-turbo-instruct",
    "prompt": "Say this is a test",
    "max_tokens": 7,
    "temperature": 0
  }'

vs new:

curl https://api.openai.com/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  -d '{
     "model": "gpt-4o-mini",
     "messages": [{"role": "user", "content": "Say this is a test!"}],
     "temperature": 0.7
   }'

[github-actions]

This issue has been automatically marked as stale because it has not had activity within 90 days. It will be automatically closed if no further activity occurs within 30 days.

[bbrowning]

We are working on adding a new Pipeline block that uses the chat completions API. However, I don't think there's a definite date to move all the existing Pipelines in data generation over to chat completions versus completions. I agree that ideally we'd only use the chat completions API, but an audit of our completions usage needs to be done to ensure that would work. Today we handling the prompt templates client-side for most of data generation, where the chat completions API handles that server-side via chat templates which doesn't let us directly change how we format prompts based on client-side needs per completion request.