Model Context Protocal with KerasHub Models #2166

laxmareddyp · 2025-08-17T05:33:57Z

Refere working colab gist

gemini-code-assist

Summary of Changes

Hello @laxmareddyp, 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 introduces a new guide and associated code demonstrating the Model Context Protocol (MCP) using KerasHub models. The primary goal is to illustrate how AI models can move beyond simple text generation to actively interact with external tools and perform real-world actions. The guide provides a modular architecture for building such systems, including components for tool definition, registration, and client-side orchestration, exemplified through practical use cases like fetching weather, performing calculations, and conducting searches.

Highlights

Introduction of Model Context Protocol (MCP) Guide: A new comprehensive guide has been added, demonstrating how to build a Model Context Protocol (MCP) system using KerasHub models. This guide illustrates how AI models can be empowered to interact with external tools and perform actions.
Implementation of Core MCP Components: The PR introduces the core components of an MCP system: MCPTool (to define individual tools), MCPToolRegistry (to manage and register available tools), and MCPClient (to orchestrate the interaction between the AI model and the tools).
Inclusion of Example Tools (Weather, Calculator, Search): Three practical example tools are provided: a weather_tool (demonstrating external data access), a calculator_tool (for mathematical computations), and a search_tool (for information retrieval). These examples showcase diverse capabilities that an MCP system can integrate.
Integration with KerasHub Models for Intelligent Tool Calling: The guide demonstrates the integration of a KerasHub model, specifically the Gemma3 Instruct 1B, to understand user requests and intelligently determine when and how to call the registered tools.

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 introduces a new guide on using KerasHub models with a Model Context Protocol (MCP) for tool calling. The guide is well-structured and provides a good walkthrough with examples for weather, calculator, and search tools.

My review focuses on improving the security and correctness of the provided code examples. I've identified a critical security vulnerability in the calculator_tool due to the use of eval(), and a bug in the tool call parsing logic for the same tool. I've provided suggestions to address these issues in both the Python script and the Markdown guide.

Also, there's a small typo in the pull request title: "Protocal" should be "Protocol".

Overall, this is a great addition and with these changes, it will be an excellent and safe guide for users.

guides/keras_hub/mcp_with_keras_hub.py

guides/md/keras_hub/mcp_with_keras_hub.md

laxmareddyp · 2025-08-17T06:41:29Z

/gemini review

gemini-code-assist

Code Review

This pull request introduces a new guide on the Model Context Protocol (MCP) with KerasHub, provided in Jupyter Notebook, Python script, and Markdown formats. The guide is well-structured and the code is generally clear. My review includes suggestions to improve code robustness and maintainability, particularly by refactoring a hardcoded tool-parsing function to be more generic. I've also pointed out opportunities for code style consistency and clarity improvements in the documentation's example outputs.

guides/keras_hub/mcp_with_keras_hub.py

guides/md/keras_hub/mcp_with_keras_hub.md

scripts/hub_master.py

danowar2 · 2025-08-17T07:04:38Z

mda@danowar2

divyashreepathihalli · 2025-08-27T21:05:35Z

guides/keras_hub/mcp_with_keras_hub.py

+
+        # Look for TOOL_CALL blocks with strict JSON parsing
+        pattern = r"TOOL_CALL:\s*\n(\{[^}]*\})"
+        matches = re.findall(pattern, response, re.DOTALL)


pattern matching here seems to be fragile right?
using Json might be better and force the model output to adhere to a specific json format

divyashreepathihalli · 2025-08-27T21:06:44Z

guides/keras_hub/mcp_with_keras_hub.py

+        tool_calls = self._extract_tool_calls(response)
+
+        if tool_calls:
+            # Safety check: if multiple tool calls found, execute only the first one


is this for easy demo purpose? if so we should mention why we are only executing one call

divyashreepathihalli · 2025-08-27T21:08:24Z

guides/keras_hub/mcp_with_keras_hub.py

+        Returns:
+            A formatted prompt for the AI model
+        """
+        tools_list = self.tool_registry.get_tools_list()


this can include more information
Clear instructions on when to use tools.
Examples of how to call tools (few-shot prompting).
Instructions on how to respond if no tool is needed.
A more explicit format for tool calls that the parser expects.
Error handling instructions (e.g., "If a tool returns an error, try to use another tool or inform the user").

divyashreepathihalli · 2025-08-27T21:10:09Z

guides/keras_hub/mcp_with_keras_hub.py

+
+if __name__ == "__main__":
+    main()
+


the simulated data is okay - not sure if this is possible but is there a real tool API we can actually show?

divyashreepathihalli · 2025-08-27T21:11:40Z

guides/keras_hub/mcp_with_keras_hub.py

+        self.model = model
+        self.tool_registry = tool_registry
+
+    def _build_prompt(self, user_input: str) -> str:


also, should we save conversation history to provide more stateful reposnse?

divyashreepathihalli · 2025-08-27T21:13:03Z

guides/keras_hub/mcp_with_keras_hub.py

+        # Look for TOOL_CALL blocks with strict JSON parsing
+        pattern = r"TOOL_CALL:\s*\n(\{[^}]*\})"
+        matches = re.findall(pattern, response, re.DOTALL)
+        for match in matches:


a more formal schema (like OpenAI's function calling) could provide clearer guidance, especially for argument types and validation.

divyashreepathihalli · 2025-08-27T21:13:38Z

guides/keras_hub/mcp_with_keras_hub.py

+        """
+        tools_list = self.tool_registry.get_tools_list()
+
+        # Ultra-simple prompt - just the essentials


lets remove llm vibes like "ultra" etc here and everywhere

laxmareddyp added 5 commits August 8, 2025 22:03

Generated the RAG guide files

a48b7ae

Add guide to keras-io

51c6802

Merge branch 'keras-team:master' into master

8e809cb

Merge branch 'keras-team:master' into master

b69ac01

MCP guide with kerashub models

3a36aad

gemini-code-assist bot reviewed Aug 17, 2025

View reviewed changes

github-actions bot assigned sachinprasadhs Aug 17, 2025

gemini-code-assist bot reviewed Aug 17, 2025

View reviewed changes

laxmareddyp added 2 commits August 17, 2025 05:40

Add path to guides

1a5367a

Fix text alignment and address gemini comments

6d8f702

gemini-code-assist bot reviewed Aug 17, 2025

View reviewed changes

guides/keras_hub/mcp_with_keras_hub.py Show resolved Hide resolved

guides/keras_hub/mcp_with_keras_hub.py Outdated Show resolved Hide resolved

guides/md/keras_hub/mcp_with_keras_hub.md Show resolved Hide resolved

scripts/hub_master.py Outdated Show resolved Hide resolved

danowar2 approved these changes Aug 17, 2025

View reviewed changes

Address gemin comments

41517d9

laxmareddyp marked this pull request as ready for review August 17, 2025 07:42

laxmareddyp requested review from fchollet and MarkDaoust as code owners August 17, 2025 07:42

divyashreepathihalli reviewed Aug 27, 2025

View reviewed changes

Merge branch 'keras-team:master' into master

cc74ec9

Model Context Protocal with KerasHub Models #2166

Are you sure you want to change the base?

Model Context Protocal with KerasHub Models #2166

Uh oh!

Conversation

laxmareddyp commented Aug 17, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

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!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

laxmareddyp commented Aug 17, 2025

Uh oh!

gemini-code-assist bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

danowar2 commented Aug 17, 2025

Uh oh!

divyashreepathihalli Aug 27, 2025

Choose a reason for hiding this comment

Uh oh!

divyashreepathihalli Aug 27, 2025

Choose a reason for hiding this comment

Uh oh!

divyashreepathihalli Aug 27, 2025

Choose a reason for hiding this comment

Uh oh!

divyashreepathihalli Aug 27, 2025

Choose a reason for hiding this comment

Uh oh!

divyashreepathihalli Aug 27, 2025

Choose a reason for hiding this comment

Uh oh!

divyashreepathihalli Aug 27, 2025

Choose a reason for hiding this comment

Uh oh!

divyashreepathihalli Aug 27, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

laxmareddyp commented Aug 17, 2025 •

edited

Loading