feat: add langfuse llm node input and output

[ZhouhaoJiang]

Summary

Fix #14272
Fix #13951

Tip

Close issue syntax: Fixes #<issue number> or Resolves #<issue number>, see documentation for more details.

Screenshots

Now

Before

Checklist

Important

Please review the checklist below before submitting your pull request.

• This change requires a documentation update, included: Dify Document
• I understand that this PR may be closed in case there was no previous discussion or issues. (This doesn't apply to typos!)
• I've added a test for each change that was introduced, and I tried as much as possible to make a single atomic change.
• I've updated the documentation accordingly.
• I ran dev/reformat(backend) and cd web && npx lint-staged(frontend) to appease the lint gods

[Copilot]

Pull Request Overview

This PR implements language model tracing with enhanced logging and token usage tracking by integrating Langfuse for llm node input and output.

• Update exception logging to include error details in ops_trace_task.py.
• Retrieve token usage from Message records and correctly assign input and output tokens in langfuse_trace.py.

Reviewed Changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
api/tasks/ops_trace_task.py	Augmented exception handling with detailed logging.
api/core/ops/langfuse_trace/langfuse_trace.py	Added token usage tracking by querying Message data; potential inversion in token assignment.

Comments suppressed due to low confidence (1)

api/tasks/ops_trace_task.py:52

• Consider using logging.error instead of logging.info when logging errors to more clearly indicate failure events.

logging.info(f"Processing trace tasks failed, app_id: {app_id}, error: {e}")

[DavideDelbianco]

@ZhouhaoJiang
From your screenshot of the after, the total token sum does not match token-in + token-out

Only one can be correct

I think the fix you wrote, repeat the TOTAL workflow consumption in every single node.

In fact if you sum the total token of the LLM + total token of the Extraction it adds up to 1827

@crazywoola
If you approve PR, make sure the PR is a fix and not a worse Bug... You don't have to understand the code, but the image is clear, cannot be the same values in every node....

This is a product that subscribers pay for... c'mon...

DON'T Release this PR it is going to break analytical data for thousand of users....