fix(ops): add streaming metrics and LLM span for agent-chat traces

[minimAluminiumalism]

Important

1. Make sure you have read our contribution guidelines
2. Ensure there is an associated issue and you have been assigned to it
3. Use the correct syntax to link this PR: Fixes #<issue number>.

Summary

Fix #28319

This PR contains the following changes.

• Track first/last token times in EasyUITaskState
• Persist streaming metrics (TTFT, TTG) to message metadata
• Add dedicated LLM child span with model attributes
• Include provider latency and tokens in Message serialization

Screenshots

Checklist

• 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

[gemini-code-assist]

Summary of Changes

Hello @minimAluminiumalism, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request significantly enhances the observability and tracing capabilities for agent-chat interactions, particularly for streaming LLM responses. It introduces mechanisms to track and persist critical streaming performance metrics like time-to-first-token and time-to-generate, while also providing more granular LLM-specific tracing data. These improvements will allow for better performance analysis and debugging of LLM-powered agent conversations.

Highlights

• Streaming Metrics Tracking: Introduced "first_token_time", "last_token_time", and "is_streaming_response" fields to "EasyUITaskState" to accurately capture timing for streaming LLM responses.
• Persistent Streaming Metrics: Calculated and persisted "time_to_first_token" (TTFT) and "time_to_generate" (TTG) into the message metadata's "usage" field, providing valuable performance insights.
• Dedicated LLM Span: Added a new "build_message_llm_span" function and integrated it to create a dedicated child span for LLM interactions within agent-chat traces, enriching observability with model-specific attributes.
• Enhanced Message Serialization: Updated the "Message" model's serialization to include "model_provider", "message_tokens", "answer_tokens", and "provider_response_latency", ensuring comprehensive data capture.

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point by creating a comment using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

Feature	Command	Description
Code Review	/gemini review	Performs a code review for the current pull request in its current state.
Pull Request Summary	/gemini summary	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in pull request comments and review comments.
Help	/gemini help	Displays a list of available commands.

Customization

To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here.

You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.

Footnotes

1. Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

[gemini-code-assist]

Code Review

This pull request effectively adds streaming metrics and a dedicated LLM span for agent-chat traces, which will improve observability. The changes are well-implemented across the different files. I've provided a couple of suggestions to improve code readability and simplify logic. Overall, great work on this enhancement.

[gemini-code-assist]

The condition self._task_state.is_streaming_response and self._task_state.first_token_time is redundant. Based on the logic in _process_stream_response, is_streaming_response is always True when first_token_time is set. You can simplify this condition to make the code more concise.

Suggested change

	if self._task_state.is_streaming_response and self._task_state.first_token_time:
	if self._task_state.first_token_time:

[minimAluminiumalism]

I prefer to keep the explicit check because the dual condition provides better protection.

[gemini-code-assist]

The logic for extracting model_provider and model_name is a bit dense and could be refactored for better readability. Also, there's a small typo in the comment on line 234 (metadata`` should be metadata`). Consider breaking down the logic into more explicit steps.

Suggested change

	model_provider = trace_metadata.get("ls_provider") or (
	message_data.get("model_provider", "") if isinstance(message_data, dict) else ""
	)
	model_name = trace_metadata.get("ls_model_name") or (
	message_data.get("model_id", "") if isinstance(message_data, dict) else ""
	)
	model_provider = trace_metadata.get("ls_provider", "")
	if not model_provider and isinstance(message_data, dict):
	model_provider = message_data.get("model_provider", "")
	model_name = trace_metadata.get("ls_model_name", "")
	if not model_name and isinstance(message_data, dict):
	model_name = message_data.get("model_id", "")

[minimAluminiumalism]

Since this is purely a style preference I'd prefer to keep it as-is unless there's a functional issue.