[vllm #23386]update recipe with container tool #43

morgendave · 2025-08-26T23:45:00Z

This adds descriptions for #23386

gemini-code-assist

Summary of Changes

Hello @morgendave, 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 updates the gpt-oss vLLM usage guide by introducing detailed documentation for the new container tool. This tool enables the execution of command-line operations within a stateful Docker container, enhancing the model's capabilities for tool use. The update also includes minor formatting adjustments for improved readability.

Highlights

New Tool Documentation: Documented the "container" tool for GPT OSS, which facilitates running command-line tools in a stateful Docker environment.
Container Tool Function Signature: Provided the expected signature for the "exec" function within the "container" tool, including its arguments like cmd, workdir, env, session_name, timeout, and user.
Tool Activation Instructions: Explained how to enable the "container" tool by setting the VLLM_ENABLE_CONTAINER_TOOL environment variable.
Documentation Formatting: Applied minor whitespace corrections across the document for consistency.

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 in your pull request via creating an issue comment (i.e. comment on the pull request page) using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands.

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 issue 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.

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 updates the gpt-oss recipe by removing trailing whitespaces and adding documentation for the container tool. The changes are good, but I've provided a few suggestions to improve the clarity and grammatical correctness of the newly added documentation section to make it easier for users to understand.

gemini-code-assist · 2025-08-26T23:47:00Z

OpenAI/GPT-OSS.md

-The URLs are expected to be MCP SSE servers that implement `instructions` in server info and well documented tools. The tools will be injected into the system prompt for the model to enable them. 
+The URLs are expected to be MCP SSE servers that implement `instructions` in server info and well documented tools. The tools will be injected into the system prompt for the model to enable them.
+
+GPT OSS also expects a builtin tool called container. It doesn't have exposed tool type in openai types.


For better readability and to follow standard English grammar, 'builtin' should be hyphenated to 'built-in' when used as an adjective.

Suggested change

GPT OSS also expects a builtin tool called container. It doesn't have exposed tool type in openai types.

GPT OSS also expects a built-in tool called container. It doesn't have exposed tool type in openai types.

gemini-code-assist · 2025-08-26T23:47:00Z

OpenAI/GPT-OSS.md

+GPT OSS also expects a builtin tool called container. It doesn't have exposed tool type in openai types.
+For reference the container tool is a stateful docker container that can be used to run command line tools.
+The enabled tool namespace is `container` and the tool name used the most is `exec`.
+MCP server need to implement the following functions to support container tool:


There's a minor grammatical error here. Since 'MCP server' is a singular subject, the verb should be 'needs' instead of 'need'.

Suggested change

MCP server need to implement the following functions to support container tool:

MCP server needs to implement the following functions to support container tool:

gemini-code-assist · 2025-08-26T23:47:01Z

OpenAI/GPT-OSS.md

+- for tool name: exec
+  - args:
+      {
+          "cmd":List[str] "command to execute",
+          "workdir":optional[str] "current working directory",
+          "env":optional[object/dict] "environment variables",
+          "session_name":optional[str] "session name",
+          "timeout":optional[int] "timeout in seconds",
+          "user":optional[str] "user name",
+      }
+Signature:
+async def exec(ctx: Context, rest_of_the_args) -> str
+expect ctx to contain a session id to identify the container session and make it stateful


The current format for describing the exec tool's arguments and signature is a bit difficult to read. The use of curly braces suggests a JSON object, but the content is not valid JSON, which can be confusing. I suggest refactoring this block to a simpler, more standard format for clarity. This will make it easier for developers to understand the required implementation.

Suggested change

- for tool name: exec

- args:

{

"cmd":List[str] "command to execute",

"workdir":optional[str] "current working directory",

"env":optional[object/dict] "environment variables",

"session_name":optional[str] "session name",

"timeout":optional[int] "timeout in seconds",

"user":optional[str] "user name",

}

Signature:

async def exec(ctx: Context, rest_of_the_args) -> str

expect ctx to contain a session id to identify the container session and make it stateful

Tool: exec

Arguments:

- cmd (List[str]): command to execute

- workdir (Optional[str]): current working directory

- env (Optional[Dict[str, str]]): environment variables

- session_name (Optional[str]): session name

- timeout (Optional[int]): timeout in seconds

- user (Optional[str]): user name

Signature:

async def exec(ctx: Context, **kwargs) -> str

# Note: `ctx` is expected to contain a session id to identify the container session and make it stateful.

heheda12345 · 2025-08-26T23:59:09Z

OpenAI/GPT-OSS.md

+async def exec(ctx: Context, rest_of_the_args) -> str
+expect ctx to contain a session id to identify the container session and make it stateful
+```
+Container tool runtime implementation can be referenced from https://github.com/SWE-agent/SWE-ReX


Can you give a series of commands to setup the SWE MCP server and run vllm with this MCP server?

we don't have a standalone MCP server but it should be the same as any MCP server to run standalone

I mean some instructions similar to the python & browser mcp server:

use (https://github.com/openai/gpt-oss/tree/main/gpt-oss-mcp-server and run

mcp run -t sse browser_server.py:mcp mcp run -t sse python_server.py:mcp vllm serve ... --tool-server ip-1:port-1,ip-2:port-2

heheda12345 · 2025-08-27T00:00:26Z

Can you also explain VLLM_HARMONY_SYSTEM_INSTRUCTIONS?

heheda12345 · 2025-09-02T21:24:26Z

OpenAI/sample_container_mcp.md

+```
+from mcp.server.fastmcp import fastmcp
+# dummy showing how to import container tool
+from swe_rex import SweRexManager


How to use https://github.com/SWE-agent/SWE-ReX to build this SweRexManager?

simon-mo · 2025-09-04T20:47:11Z

OpenAI/sample_container_mcp.md

feel free to just drop this as python script here and link it with gh link

gemini-code-assist bot reviewed Aug 26, 2025

View reviewed changes

heheda12345 reviewed Aug 26, 2025

View reviewed changes

zhiweiz added 3 commits August 28, 2025 15:06

update recipe with container tool

c6fd13e

add more docs for envs

633b153

add container examples

8ba65aa

morgendave force-pushed the container-tool-update branch from ec21634 to 8ba65aa Compare August 28, 2025 22:06

heheda12345 reviewed Sep 2, 2025

View reviewed changes

zhiweiz added 2 commits September 2, 2025 15:48

add some AI generated container manager example

09a483c

change env string

189523f

simon-mo reviewed Sep 4, 2025

View reviewed changes

OpenAI/sample_container_mcp.md

Copy link

Contributor

simon-mo Sep 4, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

feel free to just drop this as python script here and link it with gh link

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

[vllm #23386]update recipe with container tool #43

[vllm #23386]update recipe with container tool #43

morgendave commented Aug 26, 2025

Uh oh!

gemini-code-assist bot left a comment

Uh oh!

gemini-code-assist bot left a comment

Uh oh!

gemini-code-assist bot Aug 26, 2025

Uh oh!

gemini-code-assist bot Aug 26, 2025

Uh oh!

gemini-code-assist bot Aug 26, 2025

Uh oh!

heheda12345 Aug 26, 2025

Uh oh!

morgendave Aug 27, 2025

Uh oh!

heheda12345 Aug 27, 2025

Uh oh!

heheda12345 commented Aug 27, 2025

Uh oh!

heheda12345 Sep 2, 2025

Uh oh!

simon-mo Sep 4, 2025

Uh oh!

Uh oh!

	GPT OSS also expects a builtin tool called container. It doesn't have exposed tool type in openai types.
	GPT OSS also expects a built-in tool called container. It doesn't have exposed tool type in openai types.

	MCP server need to implement the following functions to support container tool:
	MCP server needs to implement the following functions to support container tool:

[vllm #23386]update recipe with container tool #43

Are you sure you want to change the base?

[vllm #23386]update recipe with container tool #43

Conversation

morgendave commented Aug 26, 2025

Uh oh!

gemini-code-assist bot left a comment

Choose a reason for hiding this comment

Summary of Changes

Highlights

Footnotes

Uh oh!

gemini-code-assist bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

gemini-code-assist bot Aug 26, 2025

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist bot Aug 26, 2025

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist bot Aug 26, 2025

Choose a reason for hiding this comment

Uh oh!

heheda12345 Aug 26, 2025

Choose a reason for hiding this comment

Uh oh!

morgendave Aug 27, 2025

Choose a reason for hiding this comment

Uh oh!

heheda12345 Aug 27, 2025

Choose a reason for hiding this comment

Uh oh!

heheda12345 commented Aug 27, 2025

Uh oh!

heheda12345 Sep 2, 2025

Choose a reason for hiding this comment

Uh oh!

simon-mo Sep 4, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!