cohere[minor]: Add support for tool calling cohere

docs/core_docs/.gitignore

@@ -176,4 +176,4 @@ docs/how_to/assign.mdx
 docs/how_to/agent_executor.md
 docs/how_to/agent_executor.mdx
 docs/integrations/llms/mistral.md
-docs/integrations/llms/mistral.mdx
+docs/integrations/llms/mistral.mdx

docs/core_docs/docs/integrations/chat/cohere.mdx

@@ -62,6 +62,14 @@ import StatefulChatExample from "@examples/models/chat/cohere/stateful_conversat
 You can see the LangSmith traces from this example [here](https://smith.langchain.com/public/8e67b05a-4e63-414e-ac91-a91acf21b262/r) and [here](https://smith.langchain.com/public/50fabc25-46fe-4727-a59c-7e4eb0de8e70/r)
 :::
 
+### Tools
+
+The Cohere API supports tool calling, along with multi-hop-tool calling. The following example demonstrates how to call tools:
+
+import CohereToolCaling from "@examples/models/chat/cohere/tool_calling.ts";
+
+<CodeBlock language="typescript">{CohereToolCaling}</CodeBlock>
+
 ### RAG
 
 Cohere also comes out of the box with RAG support.

examples/src/models/chat/cohere/chat_cohere.ts

@@ -3,7 +3,6 @@ import { ChatPromptTemplate } from "@langchain/core/prompts";
 
 const model = new ChatCohere({
   apiKey: process.env.COHERE_API_KEY, // Default
-  model: "command", // Default
 });
 const prompt = ChatPromptTemplate.fromMessages([
   ["ai", "You are a helpful assistant"],

examples/src/models/chat/cohere/chat_stream_cohere.ts

@@ -4,7 +4,6 @@ import { StringOutputParser } from "@langchain/core/output_parsers";
 
 const model = new ChatCohere({
   apiKey: process.env.COHERE_API_KEY, // Default
-  model: "command", // Default
 });
 const prompt = ChatPromptTemplate.fromMessages([
   ["ai", "You are a helpful assistant"],

examples/src/models/chat/cohere/connectors.ts

@@ -3,7 +3,6 @@ import { HumanMessage } from "@langchain/core/messages";
 
 const model = new ChatCohere({
   apiKey: process.env.COHERE_API_KEY, // Default
-  model: "command", // Default
 });
 
 const response = await model.invoke(

examples/src/models/chat/cohere/rag.ts

@@ -3,7 +3,6 @@ import { HumanMessage } from "@langchain/core/messages";
 
 const model = new ChatCohere({
   apiKey: process.env.COHERE_API_KEY, // Default
-  model: "command", // Default
 });
 
 const documents = [

examples/src/models/chat/cohere/stateful_conversation.ts

@@ -3,7 +3,6 @@ import { HumanMessage } from "@langchain/core/messages";
 
 const model = new ChatCohere({
   apiKey: process.env.COHERE_API_KEY, // Default
-  model: "command", // Default
 });
 
 const conversationId = `demo_test_id-${Math.random()}`;

examples/src/models/chat/cohere/tool_calling.ts

@@ -0,0 +1,79 @@
+import { ChatCohere } from "@langchain/cohere";
+import { HumanMessage, ToolMessage } from "@langchain/core/messages";
+import { convertToCohereTool } from "@langchain/core/utils/function_calling";
+import { z } from "zod";
+import { DynamicStructuredTool } from "@langchain/core/tools";
+
+const model = new ChatCohere({
+  apiKey: process.env.COHERE_API_KEY, // Default
+});
+
+const magicFunctionTool = new DynamicStructuredTool({
+    name: "magic_function",
+    description: "Apply a magic function to the input number",
+    schema: z.object({
+    num: z.number().describe("The number to apply the magic function for"),
+    }),
+    func: async ({ num }) => {
+    return `The magic function of ${num} is ${num + 5}`;
+    },
+});
+
+  const tools = [magicFunctionTool];
+  const modelWithTools = model.bind({
+    tools: tools.map(convertToCohereTool),
+  });
+
+  let messages = [new HumanMessage("What is the magic function of number 5?")];
+const response = await modelWithTools.invoke(
+  messages,
+);
+/**
+response:  AIMessage {
+  lc_serializable: true,
+  lc_kwargs: {
+    content: 'I will use the magic_function tool to answer this question.',
+    additional_kwargs: {
+      response_id: 'd0b189e5-3dbf-493c-93f8-99ed4b01d96d',
+      generationId: '8982a68f-c64c-48f8-bf12-0b4bea0018b6',
+      chatHistory: [Array],
+      finishReason: 'COMPLETE',
+      meta: [Object],
+      toolCalls: [Array]
+    },
+    tool_calls: [ [Object] ],
+    usage_metadata: { input_tokens: 920, output_tokens: 54, total_tokens: 974 },
+    invalid_tool_calls: [],
+    response_metadata: {}
+  },
+  lc_namespace: [ 'langchain_core', 'messages' ],
+  content: 'I will use the magic_function tool to answer this question.',
+  name: undefined,
+  additional_kwargs: {
+    response_id: 'd0b189e5-3dbf-493c-93f8-99ed4b01d96d',
+    generationId: '8982a68f-c64c-48f8-bf12-0b4bea0018b6',
+    chatHistory: [ [Object], [Object] ],
+    finishReason: 'COMPLETE',
+    meta: { apiVersion: [Object], billedUnits: [Object], tokens: [Object] },
+    toolCalls: [ [Object] ]
+  },
+  response_metadata: {
+    estimatedTokenUsage: { completionTokens: 54, promptTokens: 920, totalTokens: 974 },
+    response_id: 'd0b189e5-3dbf-493c-93f8-99ed4b01d96d',
+    generationId: '8982a68f-c64c-48f8-bf12-0b4bea0018b6',
+    chatHistory: [ [Object], [Object] ],
+    finishReason: 'COMPLETE',
+    meta: { apiVersion: [Object], billedUnits: [Object], tokens: [Object] },
+    toolCalls: [ [Object] ]
+  },
+  tool_calls: [
+    {
+      name: 'magic_function',
+      args: [Object],
+      id: '4ec98550-ba9a-4043-adfe-566230e5'
+    }
+  ],
+  invalid_tool_calls: [],
+  usage_metadata: { input_tokens: 920, output_tokens: 54, total_tokens: 974 }
+}
+ */

langchain-core/src/utils/function_calling.ts

@@ -18,6 +18,41 @@ export function convertToOpenAIFunction(
   };
 }
 
+/**
+ * Formats a `StructuredTool` instance into a format that is compatible
+ * with OpenAI function calling. It uses the `zodToJsonSchema`
+ * function to convert the schema of the `StructuredTool` into a JSON
+ * schema, which is then used as the parameters for the Cohere function.
+ */
+
+/** @ignore */
+export function convertToCohereTool(
+  tool: StructuredToolInterface
+): any {
+  const parameterDefinitionsFromZod = zodToJsonSchema(tool.schema);
+  const parameterDefinitionsProperties = "properties" in parameterDefinitionsFromZod ? parameterDefinitionsFromZod.properties : {};
+  const parameterDefinitionsRequired = "required" in parameterDefinitionsFromZod ? parameterDefinitionsFromZod.required : [];
+
+  const parameterDefinitionsFinal: Record<string, any> = {};
+
+  // Iterate through all properties
+  Object.keys(parameterDefinitionsProperties).forEach(propertyName => {
+    // Create the property in the new object
+    parameterDefinitionsFinal[propertyName] = parameterDefinitionsProperties[propertyName];
+    // Set the required property based on the 'required' array
+    if (parameterDefinitionsRequired === undefined) {
+      parameterDefinitionsRequired = [];
+    }
+    parameterDefinitionsFinal[propertyName].required = parameterDefinitionsRequired.includes(propertyName);
+  });
+
+  return {
+    name: tool.name,
+    description: tool.description,
+    parameterDefinitions: parameterDefinitionsFinal,
+  };
+}
+
 /**
  * Formats a `StructuredTool` instance into a format that is compatible
  * with OpenAI tool calling. It uses the `zodToJsonSchema`

langchain-core/tsconfig.json

@@ -9,7 +9,7 @@
       "ES2022.Object",
       "DOM"
     ],
-    "module": "ES2020",
+    "module": "NodeNext",
     "moduleResolution": "nodenext",
     "esModuleInterop": true,
     "declaration": true,

langchain/tsconfig.json

@@ -9,7 +9,7 @@
       "ES2022.Object",
       "DOM"
     ],
-    "module": "ES2020",
+    "module": "NodeNext",
     "moduleResolution": "nodenext",
     "esModuleInterop": true,
     "declaration": true,

libs/langchain-anthropic/tsconfig.json

@@ -9,7 +9,7 @@
       "ES2022.Object",
       "DOM"
     ],
-    "module": "ES2020",
+    "module": "NodeNext",
     "moduleResolution": "nodenext",
     "esModuleInterop": true,
     "declaration": true,

libs/langchain-azure-openai/tsconfig.json

@@ -9,7 +9,7 @@
       "ES2022.Object",
       "DOM"
     ],
-    "module": "ES2020",
+    "module": "NodeNext",
     "moduleResolution": "nodenext",
     "esModuleInterop": true,
     "declaration": true,

libs/langchain-cohere/package.json

@@ -25,7 +25,7 @@
     "test": "NODE_OPTIONS=--experimental-vm-modules jest --testPathIgnorePatterns=\\.int\\.test.ts --testTimeout 30000 --maxWorkers=50%",
     "test:watch": "NODE_OPTIONS=--experimental-vm-modules jest --watch --testPathIgnorePatterns=\\.int\\.test.ts",
     "test:single": "NODE_OPTIONS=--experimental-vm-modules yarn run jest --config jest.config.cjs --testTimeout 100000",
-    "test:int": "NODE_OPTIONS=--experimental-vm-modules jest --testPathPattern=\\.int\\.test.ts --testTimeout 100000 --maxWorkers=50%",
+    "test:int": "NODE_OPTIONS=--experimental-vm-modules jest --testPathPattern=\\chat_models.int\\.test.ts --testTimeout 100000 --maxWorkers=50%",
     "test:standard:unit": "NODE_OPTIONS=--experimental-vm-modules jest --testPathPattern=\\.standard\\.test.ts --testTimeout 100000 --maxWorkers=50%",
     "test:standard:int": "NODE_OPTIONS=--experimental-vm-modules jest --testPathPattern=\\.standard\\.int\\.test.ts --testTimeout 100000 --maxWorkers=50%",
     "test:standard": "yarn test:standard:unit && yarn test:standard:int",
@@ -36,7 +36,8 @@
   "license": "MIT",
   "dependencies": {
     "@langchain/core": ">=0.2.5 <0.3.0",
-    "cohere-ai": "^7.10.5"
+    "cohere-ai": "^7.10.5",
+    "zod": "^3.23.8"
   },
   "devDependencies": {
     "@jest/globals": "^29.5.0",