improve langchain.js support #165

manekinekko · 2026-01-07T12:43:41Z

No description provided.

…Server

…I_API_KEY and upgrade dependencies

Copilot

Pull request overview

This PR updates the LangChain.js integration to improve compatibility with newer versions of LangChain packages and the DeepAgents library. The changes include dependency version upgrades, refactoring of agent structures, updates to the event streaming API, improved error handling in the UI, and configuration adjustments for Azure OpenAI authentication.

Upgraded key dependencies including @langchain/core, @langchain/mcp-adapters, deepagents, and langchain to newer versions
Refactored agent configuration to use new DeepAgents API structure with direct properties instead of wrapped runnables
Updated event streaming API from v1 to v2 with destructured event handling
Added server error handling in the Angular UI service

Reviewed changes

Copilot reviewed 9 out of 10 changed files in this pull request and generated 10 comments.

Show a summary per file

File	Description
packages/ui-angular/src/app/services/api.service.ts	Added error handling for server errors in streaming responses
packages/mcp-servers/echo-ping/src/server.ts	Reordered imports to follow standard convention (external imports before internal) and added whitespace for readability
packages/api-langchain-js/src/tools/mcp-bridge.ts	Changed import to use type-only import for DynamicStructuredTool
packages/api-langchain-js/src/tools/index.ts	Simplified auth configuration by replacing requestInit headers with direct accessToken property
packages/api-langchain-js/src/server.ts	Removed debug console.log statement for cleaner output
packages/api-langchain-js/src/providers/azure-openai.ts	Added FORCE_AZURE_OPENAI_API_KEY environment variable support, changed default model fallback, and reordered configuration properties
packages/api-langchain-js/src/graph/index.ts	Updated streamEvents API to v2 with destructured event parameters and simplified event property access
packages/api-langchain-js/src/agents/index.ts	Refactored agent configuration to use DeepAgents SubAgent structure with direct model/tools/systemPrompt properties instead of runnable wrappers
packages/api-langchain-js/package.json	Updated dependency versions, added FORCE_AZURE_OPENAI_API_KEY to start script, and removed unused langchain packages

Copilot · 2026-01-07T12:48:24Z

packages/api-langchain-js/src/tools/index.ts

-              "Authorization": "Bearer " + process.env["MCP_ECHO_PING_ACCESS_TOKEN"],
-          }
-      },
+      accessToken: process.env["MCP_ECHO_PING_ACCESS_TOKEN"],


The authentication configuration has been changed from using 'requestInit.headers' with an Authorization header to using 'accessToken'. However, the 'mcp-bridge.ts' implementation still only extracts headers from 'requestInit.headers' (lines 16-19) and does not handle the new 'accessToken' field. This means the access token will not be sent with the requests, breaking authentication for the echo-ping service.

Suggested change

accessToken: process.env["MCP_ECHO_PING_ACCESS_TOKEN"],

requestInit: {

headers: {

Authorization: `Bearer ${process.env["MCP_ECHO_PING_ACCESS_TOKEN"]}`,

},

},

Copilot · 2026-01-07T12:48:24Z

packages/api-langchain-js/src/providers/azure-openai.ts

      baseURL: process.env.AZURE_OPENAI_ENDPOINT,
    },
-    modelName: process.env.AZURE_OPENAI_DEPLOYMENT_NAME ?? "gpt-5",
+    modelName: process.env.AZURE_OPENAI_DEPLOYMENT_NAME ?? "gpt-5-mini",


The default model name has been changed to "gpt-5-mini", but this model does not exist in Azure OpenAI. Valid GPT-4 model names include "gpt-4", "gpt-4-turbo", "gpt-4o", "gpt-4o-mini", or valid GPT-3.5 models like "gpt-35-turbo". The model name "gpt-5-mini" will cause runtime errors when Azure OpenAI API is called without a valid AZURE_OPENAI_DEPLOYMENT_NAME environment variable.

Copilot · 2026-01-07T12:48:25Z

packages/api-langchain-js/src/agents/index.ts

    const agent = {
      name: "EchoAgent",
-      description: "Echo back the received input. Do not respond with anything else. Always call the tools.",
-      runnable: createAgent({
+      description:
+        "Echo back the received input. Do not respond with anything else. Always call the tools.",
      model,
      tools: echoTools,
-      systemPrompt: "You are an Echo Agent that echoes back the received input. Do not respond with anything else. Always call the tools.",
-    })
+      systemPrompt:
+        "You are an Echo Agent that echoes back the received input. Do not respond with anything else. Always call the tools.",
    };


The agent structure has been changed to remove the 'runnable' property and instead spread the model, tools, and systemPrompt properties directly on the agent object. However, this change is incompatible with the 'SubAgent' type from 'deepagents' which expects an object with 'name', 'description', and 'runnable' properties. The new structure directly includes 'model', 'tools', and 'systemPrompt' which may not match the expected interface. This could cause type errors or runtime failures when creating the deep agent.

Copilot · 2026-01-07T12:48:25Z

packages/api-langchain-js/src/agents/index.ts

    const agent = {
      name: "CustomerQueryAgent",
-      description: "Assists employees in better understanding customer needs, facilitating more accurate and personalized service. This agent is particularly useful for handling nuanced queries, such as specific travel preferences or budget constraints, which are common in travel agency interactions.",
-      runnable: createAgent({
-        model,
-        name: "CustomerQueryAgent",
-        tools: customerTools,
-        systemPrompt: "You are a Customer Query Agent that assists employees in better understanding customer needs, facilitating more accurate and personalized service. This agent is particularly useful for handling nuanced queries, such as specific travel preferences or budget constraints, which are common in travel agency interactions.",
-      }),
+      description:
+        "Assists employees in better understanding customer needs, facilitating more accurate and personalized service. This agent is particularly useful for handling nuanced queries, such as specific travel preferences or budget constraints, which are common in travel agency interactions.",
+      model,
+      tools: customerTools,
+      systemPrompt:
+        "You are a Customer Query Agent that assists employees in better understanding customer needs, facilitating more accurate and personalized service. This agent is particularly useful for handling nuanced queries, such as specific travel preferences or budget constraints, which are common in travel agency interactions.",
    };


The agent structure has been changed to remove the 'runnable' property and instead spread the model, tools, and systemPrompt properties directly on the agent object. However, this change is incompatible with the 'SubAgent' type from 'deepagents' which expects an object with 'name', 'description', and 'runnable' properties. The new structure directly includes 'model', 'tools', and 'systemPrompt' which may not match the expected interface. This could cause type errors or runtime failures when creating the deep agent.

Copilot · 2026-01-07T12:48:25Z

packages/api-langchain-js/package.json

  "types": "dist/index.d.ts",
  "scripts": {
-    "start": "tsx --watch src/server.ts",
+    "start": "FORCE_AZURE_OPENAI_API_KEY=true tsx --watch src/server.ts",


The start script sets FORCE_AZURE_OPENAI_API_KEY=true using Unix/Linux shell syntax, which will not work on Windows systems. This makes the project less portable. Consider using cross-platform tools like 'cross-env' (e.g., "cross-env FORCE_AZURE_OPENAI_API_KEY=true tsx --watch src/server.ts") to ensure the script works across all operating systems.

Copilot · 2026-01-07T12:48:25Z

packages/ui-angular/src/app/services/api.service.ts

+              if (parsedData.data.error) {
+                this.handleApiError(
+                  "Server error: " + parsedData.data.error,
+                  0
+                );
+                return null;
+              }


The error handling logic accesses 'parsedData.data.error' without first verifying that 'parsedData.data' exists. If parsedData.data is null or undefined, this will throw a TypeError. Consider adding a null/undefined check for parsedData.data before accessing its error property.

Copilot · 2026-01-07T12:48:26Z

packages/api-langchain-js/src/providers/azure-openai.ts

        baseURL: process.env.AZURE_OPENAI_ENDPOINT,
      },
-      modelName: process.env.AZURE_OPENAI_DEPLOYMENT_NAME ?? "gpt-5",
+      modelName: process.env.AZURE_OPENAI_DEPLOYMENT_NAME ?? "gpt-5-mini",


The default model name has been changed to "gpt-5-mini", but this model does not exist in Azure OpenAI. Valid GPT-4 model names include "gpt-4", "gpt-4-turbo", "gpt-4o", "gpt-4o-mini", or valid GPT-3.5 models like "gpt-35-turbo". The model name "gpt-5-mini" will cause runtime errors when Azure OpenAI API is called without a valid AZURE_OPENAI_DEPLOYMENT_NAME environment variable.

Copilot · 2026-01-07T12:48:26Z

packages/api-langchain-js/src/agents/index.ts

    const agent = {
      name: "ItineraryPlanningAgent",
-      description: "Creates a travel itinerary based on user preferences and requirements.",
-      runnable: createAgent({
-        model,
-        tools: itineraryTools,
-        systemPrompt: "You are an Itinerary Planning Agent that creates travel itineraries based on user preferences and requirements.",
-      }),
+      description:
+        "Creates a travel itinerary based on user preferences and requirements.",
+      model,
+      tools: itineraryTools,
+      systemPrompt:
+        "You are an Itinerary Planning Agent that creates travel itineraries based on user preferences and requirements.",
    };


The agent structure has been changed to remove the 'runnable' property and instead spread the model, tools, and systemPrompt properties directly on the agent object. However, this change is incompatible with the 'SubAgent' type from 'deepagents' which expects an object with 'name', 'description', and 'runnable' properties. The new structure directly includes 'model', 'tools', and 'systemPrompt' which may not match the expected interface. This could cause type errors or runtime failures when creating the deep agent.

Copilot · 2026-01-07T12:48:26Z

packages/api-langchain-js/src/agents/index.ts

    const agent = {
      name: "DestinationRecommendationAgent",
-      description: "Suggests destinations based on customer preferences and requirements.",
-      runnable: createAgent({
-        model,
-        tools: destinationTools,
-        systemPrompt: "You are a Destination Recommendation Agent that suggests destinations based on customer preferences and requirements.",
-      })
+      description:
+        "Suggests destinations based on customer preferences and requirements.",
+      model,
+      tools: destinationTools,
+      systemPrompt:
+        "You are a Destination Recommendation Agent that suggests destinations based on customer preferences and requirements.",
    };


The agent structure has been changed to remove the 'runnable' property and instead spread the model, tools, and systemPrompt properties directly on the agent object. However, this change is incompatible with the 'SubAgent' type from 'deepagents' which expects an object with 'name', 'description', and 'runnable' properties. The new structure directly includes 'model', 'tools', and 'systemPrompt' which may not match the expected interface. This could cause type errors or runtime failures when creating the deep agent.

Copilot · 2026-01-07T12:48:26Z

packages/api-langchain-js/src/graph/index.ts

-        } else if (event.event === "on_tool_start") {
-          // Tool execution started
+        } else if (event === "on_tool_start") {
+          //   // Tool execution started


The comment has an extra leading space and appears to be commented out or incomplete ("// // Tool execution started"). This should either be removed entirely or properly completed as a single-line comment without the extra spaces and duplicate comment markers.

Suggested change

// // Tool execution started

// Tool execution started

manekinekko added 8 commits January 7, 2026 10:17

fix: handle server errors in API response parsing

9e87af2

fix: remove duplicate imports and improve request handling in EchoMCP…

a66119b

…Server

fix: improve logging and event handling in TravelAgentsWorkflow

e187eb7

fix: update API key handling and model name in Azure OpenAI integration

9bb3683

fix: simplify request initialization for Echo Ping configuration

4398884

fix: update import statement to use type for DynamicStructuredTool

5b3d0da

fix: remove unnecessary logging in chat event handling

b9f2177

Update package.json: Modify start script to include FORCE_AZURE_OPENA…

27e9762

…I_API_KEY and upgrade dependencies

manekinekko requested a review from Copilot January 7, 2026 12:43

Copilot started reviewing on behalf of manekinekko January 7, 2026 12:44 View session

Copilot AI reviewed Jan 7, 2026

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

improve langchain.js support #165

improve langchain.js support #165

Uh oh!

manekinekko commented Jan 7, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Jan 7, 2026

Uh oh!

Copilot AI Jan 7, 2026

Uh oh!

Copilot AI Jan 7, 2026

Uh oh!

Copilot AI Jan 7, 2026

Uh oh!

Copilot AI Jan 7, 2026

Uh oh!

Copilot AI Jan 7, 2026

Uh oh!

Copilot AI Jan 7, 2026

Uh oh!

Copilot AI Jan 7, 2026

Uh oh!

Copilot AI Jan 7, 2026

Uh oh!

Copilot AI Jan 7, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

-      accessToken: process.env["MCP_ECHO_PING_ACCESS_TOKEN"],
+      requestInit: {
+        headers: {
+          Authorization: `Bearer ${process.env["MCP_ECHO_PING_ACCESS_TOKEN"]}`,
+        },
+      },

improve langchain.js support #165

Are you sure you want to change the base?

improve langchain.js support #165

Uh oh!

Conversation

manekinekko commented Jan 7, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Jan 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 7, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants