langchain-ai/deepagents
1
0
Fork 0
mirror of https://github.com/langchain-ai/deepagents.git synced 2026-07-22 01:35:28 -04:00
Code Issues 169 Packages Projects Releases 16 Wiki Activity

Research_agent.py: Switching model provider from Claude => OpenAI (Azure) does not produce Subagents, responses, or tool calls. #18

New Issue
Closed
opened 2026-02-16 08:19:26 -05:00 by yindo · 7 comments
yindo commented 2026-02-16 08:19:26 -05:00
Owner
Copy Link
Copy Source

Originally created by @DerekKane on GitHub (Aug 16, 2025).

Hi LangChain team,

Great work as always. Very good work and Im a fan... I have been trying to launch the Researcher demo, keeping everything in tact, but switching out the model provider and model to Azure Open AI. My tests included: GPT4o, GPT4.1, GPT-5-Chat.

Inference is working with the swap but the deep agent is unable to create the sub agents, call the search tools, nor does it produce a response in the chat before invoking the write_file and write_todos. Feels like a prompting issue to me...

Question:

  1. Will this deep agent work out of the box by switching to Azure Open AI as the model provider?
  2. Are there fundamental issues which limits the usage of the agent to Claude Sonnet 4?

I ammended the research_agent.py example to include the following statements near the end:

######################################################################

import os
from langchain_openai import AzureChatOpenAI

def get_default_model():
"""
Returns an Azure OpenAI chat model via LangChain.
Requires these env vars:
- AZURE_OPENAI_API_KEY
- AZURE_OPENAI_ENDPOINT (e.g., https://.openai.azure.com)
- AZURE_OPENAI_DEPLOYMENT_NAME (your deployed model name, e.g. gpt-5-chat)
- AZURE_OPENAI_API_VERSION (e.g., 2025-01-01-preview)
"""
return AzureChatOpenAI(
azure_deployment=os.environ["AZURE_OPENAI_DEPLOYMENT_NAME"],
api_version=os.environ.get("AZURE_OPENAI_API_VERSION", "2025-01-01-preview"),
azure_endpoint=os.environ["AZURE_OPENAI_ENDPOINT"],
api_key=os.environ["AZURE_OPENAI_API_KEY"],
# Optional tuning:
max_tokens=16000, # Set a practical cap; Azure enforces per-model limits
timeout=60, # seconds
)

model = get_default_model()

######################################################################

Then I added the following statement into the create :

Create the agent

agent = create_deep_agent(
[internet_search],
research_instructions,
model=model,
subagents=[critique_sub_agent, research_sub_agent],
).with_config({"recursion_limit": 1000})

Originally created by @DerekKane on GitHub (Aug 16, 2025). Hi LangChain team, Great work as always. Very good work and Im a fan... I have been trying to launch the Researcher demo, keeping everything in tact, but switching out the model provider and model to Azure Open AI. My tests included: GPT4o, GPT4.1, GPT-5-Chat. Inference is working with the swap but the deep agent is unable to create the sub agents, call the search tools, nor does it produce a response in the chat before invoking the write_file and write_todos. Feels like a prompting issue to me... Question: 1. Will this deep agent work out of the box by switching to Azure Open AI as the model provider? 2. Are there fundamental issues which limits the usage of the agent to Claude Sonnet 4? I ammended the research_agent.py example to include the following statements near the end: ###################################################################### import os from langchain_openai import AzureChatOpenAI def get_default_model(): """ Returns an Azure OpenAI chat model via LangChain. Requires these env vars: - AZURE_OPENAI_API_KEY - AZURE_OPENAI_ENDPOINT (e.g., https://<resource>.openai.azure.com) - AZURE_OPENAI_DEPLOYMENT_NAME (your deployed model name, e.g. gpt-5-chat) - AZURE_OPENAI_API_VERSION (e.g., 2025-01-01-preview) """ return AzureChatOpenAI( azure_deployment=os.environ["AZURE_OPENAI_DEPLOYMENT_NAME"], api_version=os.environ.get("AZURE_OPENAI_API_VERSION", "2025-01-01-preview"), azure_endpoint=os.environ["AZURE_OPENAI_ENDPOINT"], api_key=os.environ["AZURE_OPENAI_API_KEY"], # Optional tuning: max_tokens=16000, # Set a practical cap; Azure enforces per-model limits timeout=60, # seconds ) model = get_default_model() ###################################################################### Then I added the following statement into the create : # Create the agent agent = create_deep_agent( [internet_search], research_instructions, model=model, subagents=[critique_sub_agent, research_sub_agent], ).with_config({"recursion_limit": 1000})
yindo added the bugneeds investigation labels 2026-02-16 08:19:26 -05:00
yindo closed this issue 2026-02-16 08:19:26 -05:00
yindo commented 2026-02-16 08:19:26 -05:00
Author
Owner
Copy Link
Copy Source

@abdimussa87 commented on GitHub (Aug 20, 2025):

Seeing the same thing from my end as well. Tried both gpt-4o and gpt-4.1

Image
@abdimussa87 commented on GitHub (Aug 20, 2025): Seeing the same thing from my end as well. Tried both gpt-4o and gpt-4.1 <img width="1505" height="764" alt="Image" src="https://github.com/user-attachments/assets/7245d9f2-212b-43ed-a3a5-b972db235856" />
yindo commented 2026-02-16 08:19:27 -05:00
Author
Owner
Copy Link
Copy Source

@malotho commented on GitHub (Aug 20, 2025):

Stuck on the same situation also

@malotho commented on GitHub (Aug 20, 2025): Stuck on the same situation also
yindo commented 2026-02-16 08:19:27 -05:00
Author
Owner
Copy Link
Copy Source

@57hemanth commented on GitHub (Aug 21, 2025):

I tried switching the model to "Gemini 2.5 Pro", but I’m still running into the same problem. I don’t get any response back.

@57hemanth commented on GitHub (Aug 21, 2025): I tried switching the model to "Gemini 2.5 Pro", but I’m still running into the same problem. I don’t get any response back.
yindo commented 2026-02-16 08:19:27 -05:00
Author
Owner
Copy Link
Copy Source

@LukeDevs commented on GitHub (Aug 21, 2025):

There is nothing fundamentally wrong with the deep agents implementation when providing an Azure Open AI model. I have this working with Azure Open AI GPT-4.1 and the GPT-5 variants.

Your input prompt (query) isn't triggering a tool call in the GPT models, so the GPT models are either exiting early or more likely only leveraging their latent knowledge.

Try rewriting your prompt to be more explicit on the use of internet search.

For example (@abdimussa87) try: "Produce a report that thoroughly compares Lionel Messi and Cristiano Ronaldo's career achievements. ALWAYS use internet search."

This is likely just a case of the System message having been tuned and tested on only the Claude models. The plumbing of the application works.

Prompts will always be unexpectadly brittle when moving between models!

@LukeDevs commented on GitHub (Aug 21, 2025): There is nothing fundamentally wrong with the deep agents implementation when providing an Azure Open AI model. I have this working with Azure Open AI GPT-4.1 and the GPT-5 variants. Your input prompt (query) isn't triggering a tool call in the GPT models, so the GPT models are either exiting early or more likely only leveraging their latent knowledge. Try rewriting your prompt to be more explicit on the use of internet search. For example (@abdimussa87) try: "Produce a report that thoroughly compares Lionel Messi and Cristiano Ronaldo's career achievements. ALWAYS use internet search." This is likely just a case of the System message having been tuned and tested on only the Claude models. The plumbing of the application works. Prompts will always be unexpectadly brittle when moving between models!
yindo commented 2026-02-16 08:19:27 -05:00
Author
Owner
Copy Link
Copy Source

@DerekKane commented on GitHub (Aug 21, 2025):

I see what you are saying and this makes sense. However, the user should not have to specify the terms "ALWAYS use internet search" and the agent should just handle it accordingly under the hood.

@LukeDevs - How would you ammend the prompts to achieve this behavior for all the tool calls and subagents to make it compatible with OpenAI?

Here is the heart of the stock example:

sub_research_prompt = """You are a dedicated researcher. Your job is to conduct research based on the users questions.

Conduct thorough research and then reply to the user with a detailed answer to their question

only your FINAL answer will be passed on to the user. They will have NO knowledge of anything except your final message, so your final report should be your final message!"""

research_sub_agent = {
"name": "research-agent",
"description": "Used to research more in depth questions. Only give this researcher one topic at a time. Do not pass multiple sub questions to this researcher. Instead, you should break down a large topic into the necessary components, and then call multiple research agents in parallel, one for each sub question.",
"prompt": sub_research_prompt,
"tools": ["internet_search"]
}

sub_critique_prompt = """You are a dedicated editor. You are being tasked to critique a report.

You can find the report at final_report.md.

You can find the question/topic for this report at question.txt.

The user may ask for specific areas to critique the report in. Respond to the user with a detailed critique of the report. Things that could be improved.

You can use the search tool to search for information, if that will help you critique the report

Do not write to the final_report.md yourself.

Things to check:

  • Check that each section is appropriately named
  • Check that the report is written as you would find in an essay or a textbook - it should be text heavy, do not let it just be a list of bullet points!
  • Check that the report is comprehensive. If any paragraphs or sections are short, or missing important details, point it out.
  • Check that the article covers key areas of the industry, ensures overall understanding, and does not omit important parts.
  • Check that the article deeply analyzes causes, impacts, and trends, providing valuable insights
  • Check that the article closely follows the research topic and directly answers questions
  • Check that the article has a clear structure, fluent language, and is easy to understand.
    """

critique_sub_agent = {
"name": "critique-agent",
"description": "Used to critique the final report. Give this agent some infomration about how you want it to critique the report.",
"prompt": sub_critique_prompt,
}

Prompt prefix to steer the agent to be an expert researcher

research_instructions = """You are an expert researcher. Your job is to conduct thorough research, and then write a polished report.

The first thing you should do is to write the original user question to question.txt so you have a record of it.

Use the research-agent to conduct deep research. It will respond to your questions/topics with a detailed answer.

When you think you enough information to write a final report, write it to final_report.md

You can call the critique-agent to get a critique of the final report. After that (if needed) you can do more research and edit the final_report.md
You can do this however many times you want until are you satisfied with the result.

Only edit the file once at a time (if you call this tool in parallel, there may be conflicts).

Here are instructions for writing the final report:

<report_instructions>

CRITICAL: Make sure the answer is written in the same language as the human messages! If you make a todo plan - you should note in the plan what language the report should be in so you dont forget!
Note: the language the report should be in is the language the QUESTION is in, not the language/country that the question is ABOUT.

Please create a detailed answer to the overall research brief that:

  1. Is well-organized with proper headings (# for title, ## for sections, ### for subsections)
  2. Includes specific facts and insights from the research
  3. References relevant sources using Title format
  4. Provides a balanced, thorough analysis. Be as comprehensive as possible, and include all information that is relevant to the overall research question. People are using you for deep research and will expect detailed, comprehensive answers.
  5. Includes a "Sources" section at the end with all referenced links

You can structure your report in a number of different ways. Here are some examples:

To answer a question that asks you to compare two things, you might structure your report like this:
1/ intro
2/ overview of topic A
3/ overview of topic B
4/ comparison between A and B
5/ conclusion

To answer a question that asks you to return a list of things, you might only need a single section which is the entire list.
1/ list of things or table of things
Or, you could choose to make each item in the list a separate section in the report. When asked for lists, you don't need an introduction or conclusion.
1/ item 1
2/ item 2
3/ item 3

To answer a question that asks you to summarize a topic, give a report, or give an overview, you might structure your report like this:
1/ overview of topic
2/ concept 1
3/ concept 2
4/ concept 3
5/ conclusion

If you think you can answer the question with a single section, you can do that too!
1/ answer

REMEMBER: Section is a VERY fluid and loose concept. You can structure your report however you think is best, including in ways that are not listed above!
Make sure that your sections are cohesive, and make sense for the reader.

For each section of the report, do the following:

  • Use simple, clear language
  • Use ## for section title (Markdown format) for each section of the report
  • Do NOT ever refer to yourself as the writer of the report. This should be a professional report without any self-referential language.
  • Do not say what you are doing in the report. Just write the report without any commentary from yourself.
  • Each section should be as long as necessary to deeply answer the question with the information you have gathered. It is expected that sections will be fairly long and verbose. You are writing a deep research report, and users will expect a thorough answer.
  • Use bullet points to list out information when appropriate, but by default, write in paragraph form.

REMEMBER:
The brief and research may be in English, but you need to translate this information to the right language when writing the final answer.
Make sure the final answer report is in the SAME language as the human messages in the message history.

Format the report in clear markdown with proper structure and include source references where appropriate.

- Assign each unique URL a single citation number in your text - End with ### Sources that lists each source with corresponding numbers - IMPORTANT: Number sources sequentially without gaps (1,2,3,4...) in the final list regardless of which sources you choose - Each source should be a separate line item in a list, so that in markdown it is rendered as a list. - Example format: [1] Source Title: URL [2] Source Title: URL - Citations are extremely important. Make sure to include these, and pay a lot of attention to getting these right. Users will often use these citations to look into more information.

You have access to a few tools.

internet_search

Use this to run an internet search for a given query. You can specify the number of results, the topic, and whether raw content should be included.

@DerekKane commented on GitHub (Aug 21, 2025): I see what you are saying and this makes sense. However, the user should not have to specify the terms "ALWAYS use internet search" and the agent should just handle it accordingly under the hood. @LukeDevs - How would you ammend the prompts to achieve this behavior for all the tool calls and subagents to make it compatible with OpenAI? Here is the heart of the stock example: sub_research_prompt = """You are a dedicated researcher. Your job is to conduct research based on the users questions. Conduct thorough research and then reply to the user with a detailed answer to their question only your FINAL answer will be passed on to the user. They will have NO knowledge of anything except your final message, so your final report should be your final message!""" research_sub_agent = { "name": "research-agent", "description": "Used to research more in depth questions. Only give this researcher one topic at a time. Do not pass multiple sub questions to this researcher. Instead, you should break down a large topic into the necessary components, and then call multiple research agents in parallel, one for each sub question.", "prompt": sub_research_prompt, "tools": ["internet_search"] } sub_critique_prompt = """You are a dedicated editor. You are being tasked to critique a report. You can find the report at `final_report.md`. You can find the question/topic for this report at `question.txt`. The user may ask for specific areas to critique the report in. Respond to the user with a detailed critique of the report. Things that could be improved. You can use the search tool to search for information, if that will help you critique the report Do not write to the `final_report.md` yourself. Things to check: - Check that each section is appropriately named - Check that the report is written as you would find in an essay or a textbook - it should be text heavy, do not let it just be a list of bullet points! - Check that the report is comprehensive. If any paragraphs or sections are short, or missing important details, point it out. - Check that the article covers key areas of the industry, ensures overall understanding, and does not omit important parts. - Check that the article deeply analyzes causes, impacts, and trends, providing valuable insights - Check that the article closely follows the research topic and directly answers questions - Check that the article has a clear structure, fluent language, and is easy to understand. """ critique_sub_agent = { "name": "critique-agent", "description": "Used to critique the final report. Give this agent some infomration about how you want it to critique the report.", "prompt": sub_critique_prompt, } # Prompt prefix to steer the agent to be an expert researcher research_instructions = """You are an expert researcher. Your job is to conduct thorough research, and then write a polished report. The first thing you should do is to write the original user question to `question.txt` so you have a record of it. Use the research-agent to conduct deep research. It will respond to your questions/topics with a detailed answer. When you think you enough information to write a final report, write it to `final_report.md` You can call the critique-agent to get a critique of the final report. After that (if needed) you can do more research and edit the `final_report.md` You can do this however many times you want until are you satisfied with the result. Only edit the file once at a time (if you call this tool in parallel, there may be conflicts). Here are instructions for writing the final report: <report_instructions> CRITICAL: Make sure the answer is written in the same language as the human messages! If you make a todo plan - you should note in the plan what language the report should be in so you dont forget! Note: the language the report should be in is the language the QUESTION is in, not the language/country that the question is ABOUT. Please create a detailed answer to the overall research brief that: 1. Is well-organized with proper headings (# for title, ## for sections, ### for subsections) 2. Includes specific facts and insights from the research 3. References relevant sources using [Title](URL) format 4. Provides a balanced, thorough analysis. Be as comprehensive as possible, and include all information that is relevant to the overall research question. People are using you for deep research and will expect detailed, comprehensive answers. 5. Includes a "Sources" section at the end with all referenced links You can structure your report in a number of different ways. Here are some examples: To answer a question that asks you to compare two things, you might structure your report like this: 1/ intro 2/ overview of topic A 3/ overview of topic B 4/ comparison between A and B 5/ conclusion To answer a question that asks you to return a list of things, you might only need a single section which is the entire list. 1/ list of things or table of things Or, you could choose to make each item in the list a separate section in the report. When asked for lists, you don't need an introduction or conclusion. 1/ item 1 2/ item 2 3/ item 3 To answer a question that asks you to summarize a topic, give a report, or give an overview, you might structure your report like this: 1/ overview of topic 2/ concept 1 3/ concept 2 4/ concept 3 5/ conclusion If you think you can answer the question with a single section, you can do that too! 1/ answer REMEMBER: Section is a VERY fluid and loose concept. You can structure your report however you think is best, including in ways that are not listed above! Make sure that your sections are cohesive, and make sense for the reader. For each section of the report, do the following: - Use simple, clear language - Use ## for section title (Markdown format) for each section of the report - Do NOT ever refer to yourself as the writer of the report. This should be a professional report without any self-referential language. - Do not say what you are doing in the report. Just write the report without any commentary from yourself. - Each section should be as long as necessary to deeply answer the question with the information you have gathered. It is expected that sections will be fairly long and verbose. You are writing a deep research report, and users will expect a thorough answer. - Use bullet points to list out information when appropriate, but by default, write in paragraph form. REMEMBER: The brief and research may be in English, but you need to translate this information to the right language when writing the final answer. Make sure the final answer report is in the SAME language as the human messages in the message history. Format the report in clear markdown with proper structure and include source references where appropriate. <Citation Rules> - Assign each unique URL a single citation number in your text - End with ### Sources that lists each source with corresponding numbers - IMPORTANT: Number sources sequentially without gaps (1,2,3,4...) in the final list regardless of which sources you choose - Each source should be a separate line item in a list, so that in markdown it is rendered as a list. - Example format: [1] Source Title: URL [2] Source Title: URL - Citations are extremely important. Make sure to include these, and pay a lot of attention to getting these right. Users will often use these citations to look into more information. </Citation Rules> </report_instructions> You have access to a few tools. ## `internet_search` Use this to run an internet search for a given query. You can specify the number of results, the topic, and whether raw content should be included.
yindo commented 2026-02-16 08:19:27 -05:00
Author
Owner
Copy Link
Copy Source

@DerekKane commented on GitHub (Aug 21, 2025):

@LukeDevs Are you getting subagents to spin up via OpenAI as well or only tool calls?

@DerekKane commented on GitHub (Aug 21, 2025): @LukeDevs Are you getting subagents to spin up via OpenAI as well or only tool calls?
yindo commented 2026-02-16 08:19:27 -05:00
Author
Owner
Copy Link
Copy Source

@sydney-runkle commented on GitHub (Jan 16, 2026):

Closing this as stale for now, but definitely feel free to open a new issue if this is still a problem for you. We've been doing some prompt optimizations that will hopefully help w/ this.

@sydney-runkle commented on GitHub (Jan 16, 2026): Closing this as stale for now, but definitely feel free to open a new issue if this is still a problem for you. We've been doing some prompt optimizations that will hopefully help w/ this.
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
MCP
backends
bug
cli
config
customization
deepagents
documentation
external
feature
fix
good first issue
help wanted
integration
internal
invalid
middleware
model
needs investigation
p0
p1
p2
p3
performance
pull-request
Mirrored from GitHub Pull Request
 question
sdk
subagents
unable to reproduce
windows
No Label bug needs investigation
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
yindo
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: langchain-ai/deepagents#18
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?