Feat/deepresearch for serving #50

zzhangpurdue · 2025-12-03T04:10:58Z

📝 PR Type

support deepresearch serving and webui.

📚 Description

[Please briefly describe the background, changes, and purpose of this PR. For example:

Added game_werewolves to demonstrate XYZ functionality in agentscope.
Fixed test failures in game_test.py caused by agentscope interface changes.
Updated dependency installation instructions in README.md of agentscope-samples.]

🧪 Testing Validation

[Please explain how to validate the changes:

How to run the added/modified test cases?
Is integration testing with agentscope required?
Has code been formatted (e.g., pre-commit)?]

✅ Checklist

Please complete the following checks before submitting the PR:

All sample code has been formatted with pre-commit run --all-files
All new/modified test cases have passed (run pytest tests/)
Test coverage has not decreased (if applicable)
Sample code follows agentscope best practices (e.g., config management, logging)
Related documentation in agentscope-samples has been updated (e.g., README.md)

cla-assistant · 2025-12-03T04:11:12Z

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.

Zhicheng Zhang seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

Copilot

Pull request overview

This PR adds support for deploying the Deep Research agent as a service with a web UI, integrating ModelStudio web search as the default search tool, and adding optional Aliyun OSS storage for generated reports.

Key changes:

Added deployment capability via deploy.py with AgentScope runtime integration
Switched default search from Tavily to ModelStudio web search while maintaining Tavily as optional
Added Aliyun OSS integration for uploading generated research reports with signed URLs

Reviewed changes

Copilot reviewed 10 out of 11 changed files in this pull request and generated 16 comments.

Show a summary per file

File	Description
`tests/agent_deep_research_test.py`	Updated import to use renamed `run` function instead of `main`
`deep_research/agent_deep_research/utils.py`	Added OSS upload utility function and reduced `TOOL_RESULTS_MAX_WORDS` from 5000 to 500
`deep_research/agent_deep_research/requirements.txt`	Added `agentscope-runtime` and `oss2` dependencies
`deep_research/agent_deep_research/main.py`	Renamed `main` function to `run` and changed example query to Chinese
`deep_research/agent_deep_research/deploy.py`	New file implementing agent deployment with session management and state persistence
`deep_research/agent_deep_research/deep_research_agent.py`	Integrated ModelStudio search, added OSS upload to report generation, modified message handling for list inputs, and uncommented debug print statements
`deep_research/agent_deep_research/built_in_prompt/*.md`	Added language sensitivity instructions to prompts to maintain user's input language
`deep_research/agent_deep_research/README.md`	Updated documentation for ModelStudio search setup and deployment instructions
`.gitignore`	Added `log/` directory to ignore list

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2025-12-10T09:37:09Z

deep_research/agent_deep_research/README.md

+Once run the deploy script, the deepresearch will run at `http://0.0.0.0:8090/process` with sync manner,
+and a webui will start at `http://localhost:5173/`, then user could test the service at the webui.
+



The word "sync" should be "asynchronous" or "async". The description "with sync manner" is grammatically awkward and technically inaccurate since the deploy script uses async functions. Consider rewording to: "the deepresearch will run at http://0.0.0.0:8090/process and a webui will start at http://localhost:5173/, then user could test the service at the webui."

Suggested change

Once run the deploy script, the deepresearch will run at `http://0.0.0.0:8090/process` with sync manner,

and a webui will start at `http://localhost:5173/`, then user could test the service at the webui.

Once you run the deploy script, the deepresearch service will be available at `http://0.0.0.0:8090/process`, and a webui will start at `http://localhost:5173/`. You can test the service using the webui.

Copilot · 2025-12-10T09:37:09Z

deep_research/agent_deep_research/README.md

 )
 await tavily_search_client.connect()
 ```
+and make sure disable modelstudio_web_search, and enable tavily-search, in `deep_research_agent.py` at line 170 and line 177, respectively.


Grammar error: "and make sure disable" should be "and make sure to disable". Also, the instruction could be clearer about the specific changes needed.

Suggested change

and make sure disable modelstudio_web_search, and enable tavily-search, in `deep_research_agent.py` at line 170 and line 177, respectively.

and make sure to disable `modelstudio_web_search` and to enable `tavily-search` in `deep_research_agent.py` at line 170 and line 177, respectively.

Copilot · 2025-12-10T09:37:10Z

deep_research/agent_deep_research/deep_research_agent.py

+        return write_report_tool_res_msg, detailed_report_path, oss_url

    async def _summarizing(self) -> Msg:
        """Generate a report based on the exsisting findings when the


Typo in the docstring: "exsisting" should be "existing".

Suggested change

"""Generate a report based on the exsisting findings when the

"""Generate a report based on the existing findings when the

Copilot · 2025-12-10T09:37:10Z

deep_research/agent_deep_research/deep_research_agent.py

+            await self.toolkit.register_mcp_client(search_mcp_client)
+            # using modelstudio web search
+            self.toolkit.tools.pop("tavily-search")


When search_mcp_client is None (which is now allowed by the Optional parameter), calling await self.toolkit.register_mcp_client(search_mcp_client) will likely fail. The code should check if search_mcp_client is not None before attempting to register it.

Suggested change

await self.toolkit.register_mcp_client(search_mcp_client)

# using modelstudio web search

self.toolkit.tools.pop("tavily-search")

if search_mcp_client is not None:

await self.toolkit.register_mcp_client(search_mcp_client)

# using modelstudio web search

self.toolkit.tools.pop("tavily-search")

Copilot · 2025-12-10T09:37:10Z

deep_research/agent_deep_research/deep_research_agent.py

+        loop = asyncio.get_running_loop()
+
+        async def _init_mcp_client():
+            await self.toolkit.register_mcp_client(search_mcp_client)
+            # using modelstudio web search
+            self.toolkit.tools.pop("tavily-search")
+
+        loop.create_task(_init_mcp_client())


The task to initialize the MCP client is created but not awaited, which could lead to race conditions. The code immediately sets self.search_function = "modelstudio_web_search" on line 187, but attempts to pop "tavily-search" from self.toolkit.tools inside the unawaited task. If the agent is used before this task completes, "tavily-search" might still be present in the toolkit, or the registration might not be complete.

Copilot · 2025-12-10T09:37:13Z

deep_research/agent_deep_research/README.md

+and make sure disable modelstudio_web_search, and enable tavily-search, in `deep_research_agent.py` at line 170 and line 177, respectively.
+



The line number references (line 170 and line 177) are now outdated after the code changes. With the new code structure, line 170 is where self.toolkit.register_tool_function(self.generate_response) is, and line 177 is within the modelstudio tool registration. The instructions should be updated to provide accurate line numbers or be more descriptive (e.g., "comment out the modelstudio search registration and uncomment the tavily-search line").

Suggested change

and make sure disable modelstudio_web_search, and enable tavily-search, in `deep_research_agent.py` at line 170 and line 177, respectively.

In `deep_research_agent.py`, make sure to disable the modelstudio web search registration and enable the tavily-search registration.

Specifically, comment out the line that registers the modelstudio search tool, and uncomment the line that registers the tavily-search tool.

Copilot · 2025-12-10T09:37:13Z

deep_research/agent_deep_research/README.md

+    make sure  `export DASHSCOPE_API_KEY="your_dashscope_api_key_here" have enable the web search from 
+    [here](https://bailian.console.aliyun.com/?tab=app#/mcp-market/detail/WebSearch)
+
+3. **Optional:Test Tavily MCP Server**:


Missing colon after "Optional". The correct formatting should be "3. Optional: Test Tavily MCP Server:"

Suggested change

3. **Optional:Test Tavily MCP Server**:

3. **Optional: Test Tavily MCP Server**:

Copilot · 2025-12-10T09:37:13Z

deep_research/agent_deep_research/deploy.py

+    request: AgentRequest = None,
+    **kwargs,
+):
+    assert kwargs is not None, "kwargs is Required for query_func"


The assert statement is checking kwargs which is always truthy (it's a dict, even if empty). This assertion will never fail. If you want to check that kwargs is not empty, use assert kwargs or assert len(kwargs) > 0. However, the comment suggests kwargs is required but the code doesn't actually use kwargs, so this assertion may not be needed at all.

Suggested change

assert kwargs is not None, "kwargs is Required for query_func"

Copilot · 2025-12-10T09:37:14Z

deep_research/agent_deep_research/deep_research_agent.py

+                # if (
+                #     tool_call["name"] != self.finish_function_name
+                #     or (tool_call["name"] == self.finish_function_name
+                #     and not chunk.metadata.get("success"))
+                # ):
+                await self.print(tool_res_msg, chunk.is_last)


The commented-out code should be removed rather than left in place. If this conditional printing logic is no longer needed, remove it. If it might be needed later, document why it was removed in a comment or commit message. The unconditional await self.print(tool_res_msg, chunk.is_last) on line 352 now prints all tool results including the finish function, which may not be the desired behavior based on the original logic.

Copilot · 2025-12-10T09:37:14Z

deep_research/agent_deep_research/deep_research_agent.py

+                # ):
+                await self.print(tool_res_msg, chunk.is_last)
+
+                logger.info(f"chunk: {chunk}")


This debug logging statement should be removed or changed to use the appropriate log level (debug instead of info). Logging "chunk" for every iteration could produce excessive logs in production.

Suggested change

logger.info(f"chunk: {chunk}")

logger.debug(f"chunk: {chunk}")

Zhicheng Zhang added 5 commits December 2, 2025 17:13

add agentapp run function

e4f6ab1

update generate response

01e0d5b

update requirements

815a53c

update requirements

99adf99

fix lint

9b84aea

Zhicheng Zhang added 3 commits December 3, 2025 12:11

remove useless ignore

9778e17

update readme

ee1291c

update readme for modelstudio web search

bc7885b

zzhangpurdue requested review from a team and Copilot December 10, 2025 09:27

Copilot started reviewing on behalf of zzhangpurdue December 10, 2025 09:28 View session

update readme for modelstudio web search

e53342a

Copilot AI reviewed Dec 10, 2025

View reviewed changes

update readme for tavily extract

35da22c

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Feat/deepresearch for serving #50

Feat/deepresearch for serving #50

zzhangpurdue commented Dec 3, 2025

Uh oh!

cla-assistant bot commented Dec 3, 2025

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Dec 10, 2025

Uh oh!

Copilot AI Dec 10, 2025

Uh oh!

Copilot AI Dec 10, 2025

Uh oh!

Copilot AI Dec 10, 2025

Uh oh!

Copilot AI Dec 10, 2025

Uh oh!

Copilot AI Dec 10, 2025

Uh oh!

Copilot AI Dec 10, 2025

Uh oh!

Copilot AI Dec 10, 2025

Uh oh!

Copilot AI Dec 10, 2025

Uh oh!

Copilot AI Dec 10, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

		Once run the deploy script, the deepresearch will run at `http://0.0.0.0:8090/process` with sync manner,
		and a webui will start at `http://localhost:5173/`, then user could test the service at the webui.

	Once run the deploy script, the deepresearch will run at `http://0.0.0.0:8090/process` with sync manner,
	and a webui will start at `http://localhost:5173/`, then user could test the service at the webui.
	Once you run the deploy script, the deepresearch service will be available at `http://0.0.0.0:8090/process`, and a webui will start at `http://localhost:5173/`. You can test the service using the webui.

	and make sure disable modelstudio_web_search, and enable tavily-search, in `deep_research_agent.py` at line 170 and line 177, respectively.
	and make sure to disable `modelstudio_web_search` and to enable `tavily-search` in `deep_research_agent.py` at line 170 and line 177, respectively.

	"""Generate a report based on the exsisting findings when the
	"""Generate a report based on the existing findings when the

	and make sure disable modelstudio_web_search, and enable tavily-search, in `deep_research_agent.py` at line 170 and line 177, respectively.
	In `deep_research_agent.py`, make sure to disable the modelstudio web search registration and enable the tavily-search registration.
	Specifically, comment out the line that registers the modelstudio search tool, and uncomment the line that registers the tavily-search tool.

	3. Optional:Test Tavily MCP Server:
	3. Optional: Test Tavily MCP Server:

	logger.info(f"chunk: {chunk}")
	logger.debug(f"chunk: {chunk}")

Feat/deepresearch for serving #50

Are you sure you want to change the base?

Feat/deepresearch for serving #50

Conversation

zzhangpurdue commented Dec 3, 2025

📝 PR Type

support deepresearch serving and webui.

📚 Description

🧪 Testing Validation

✅ Checklist

Uh oh!

cla-assistant bot commented Dec 3, 2025

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Dec 10, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Dec 10, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Dec 10, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Dec 10, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Dec 10, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Dec 10, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Dec 10, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Dec 10, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Dec 10, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Dec 10, 2025

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant