Is it possible to use pre-fill content for claude?

Questions
1

Is is possible to use prefill for claude ai for better output control?
Currently the structure output control is very limited and somewhat random and buggy.

2

It looks like your topic is missing some important information. Could you provide the following if applicable.

  • n8n version:
  • Database (default: SQLite):
  • n8n EXECUTIONS_PROCESS setting (default: own, main):
  • Running n8n via (Docker, npm, n8n cloud, desktop app):
  • Operating system:
3

Is this similar to OpenAI’s Predicted Output?

If yes, then n8n currently doesn’t support these features natively.

But please add your vote and comment on the feature request, so we can get their attention:

If my reply answers your question, please remember to mark it as a solution.

4

I recommend using an OpenAI assistant with JSON schema, this allows you to configure your desired output structure.

1 Like
5

Thanks i will try that out.

6

No problem. Just make sure you have the output set to the desired JSON, and that it saves, not some models are not compatible with the JSON output.

If this helps you out, feel free to mark as solved!

7

No its actually not solved it would be solved when we can control the output of the model with prefill. Is anybody here a contributor to point me in the right direction to implement as a contribution to the project?

8

Ok well even with claude, from the link you sent

To prefill, include the desired initial text in the Assistant message (Claude’s response will continue from where the Assistant message leaves off):

This can easily be accomplished with an HTTP request.

9

This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.