feat(openai): Alias system messages as developer messages for o1, add…

… reasoning_effort param (#7398)

libs/langchain-openai/package.json

@@ -36,7 +36,7 @@
   "license": "MIT",
   "dependencies": {
     "js-tiktoken": "^1.0.12",
-    "openai": "^4.71.0",
+    "openai": "^4.77.0",
     "zod": "^3.22.4",
     "zod-to-json-schema": "^3.22.3"
   },

libs/langchain-openai/src/chat_models.ts

@@ -95,7 +95,13 @@ interface OpenAILLMOutput {
 }
 
 // TODO import from SDK when available
-type OpenAIRoleEnum = "system" | "assistant" | "user" | "function" | "tool";
+type OpenAIRoleEnum =
+  | "system"
+  | "developer"
+  | "assistant"
+  | "user"
+  | "function"
+  | "tool";
 
 type OpenAICompletionParam =
   OpenAIClient.Chat.Completions.ChatCompletionMessageParam;
@@ -105,6 +111,7 @@ type OpenAIFnCallOption = OpenAIClient.Chat.ChatCompletionFunctionCallOption;
 function extractGenericMessageCustomRole(message: ChatMessage) {
   if (
     message.role !== "system" &&
+    message.role !== "developer" &&
     message.role !== "assistant" &&
     message.role !== "user" &&
     message.role !== "function" &&
@@ -249,6 +256,14 @@ function _convertDeltaToMessageChunk(
     });
   } else if (role === "system") {
     return new SystemMessageChunk({ content, response_metadata });
+  } else if (role === "developer") {
+    return new SystemMessageChunk({
+      content,
+      response_metadata,
+      additional_kwargs: {
+        __openai_role__: "developer",
+      },
+    });
   } else if (role === "function") {
     return new FunctionMessageChunk({
       content,
@@ -270,13 +285,18 @@ function _convertDeltaToMessageChunk(
 
 // Used in LangSmith, export is important here
 export function _convertMessagesToOpenAIParams(
-  messages: BaseMessage[]
+  messages: BaseMessage[],
+  model?: string
 ): OpenAICompletionParam[] {
   // TODO: Function messages do not support array content, fix cast
   return messages.flatMap((message) => {
+    let role = messageToOpenAIRole(message);
+    if (role === "system" && model?.startsWith("o1")) {
+      role = "developer";
+    }
     // eslint-disable-next-line @typescript-eslint/no-explicit-any
     const completionParam: Record<string, any> = {
-      role: messageToOpenAIRole(message),
+      role,
       content: message.content,
     };
     if (message.name != null) {
@@ -428,6 +448,12 @@ export interface ChatOpenAICallOptions
    * [Learn more](https://platform.openai.com/docs/guides/latency-optimization#use-predicted-outputs).
    */
   prediction?: OpenAIClient.ChatCompletionPredictionContent;
+
+  /**
+   * Constrains effort on reasoning for reasoning models. Currently supported values are low, medium, and high.
+   * Reducing reasoning effort can result in faster responses and fewer tokens used on reasoning in a response.
+   */
+  reasoning_effort?: OpenAIClient.Chat.ChatCompletionReasoningEffort;
 }
 
 export interface ChatOpenAIFields
@@ -994,6 +1020,7 @@ export class ChatOpenAI<
       "promptIndex",
       "response_format",
       "seed",
+      "reasoning_effort",
     ];
   }
 
@@ -1092,6 +1119,8 @@ export class ChatOpenAI<
 
   modalities?: Array<OpenAIClient.Chat.ChatCompletionModality>;
 
+  reasoningEffort?: OpenAIClient.Chat.ChatCompletionReasoningEffort;
+
   constructor(
     fields?: ChatOpenAIFields,
     /** @deprecated */
@@ -1162,6 +1191,7 @@ export class ChatOpenAI<
     this.__includeRawResponse = fields?.__includeRawResponse;
     this.audio = fields?.audio;
     this.modalities = fields?.modalities;
+    this.reasoningEffort = fields?.reasoningEffort;
 
     if (this.azureOpenAIApiKey || this.azureADTokenProvider) {
       if (
@@ -1337,6 +1367,10 @@ export class ChatOpenAI<
     if (options?.prediction !== undefined) {
       params.prediction = options.prediction;
     }
+    const reasoningEffort = options?.reasoning_effort ?? this.reasoningEffort;
+    if (reasoningEffort !== undefined) {
+      params.reasoning_effort = reasoningEffort;
+    }
     return params;
   }
 
@@ -1360,7 +1394,7 @@ export class ChatOpenAI<
     runManager?: CallbackManagerForLLMRun
   ): AsyncGenerator<ChatGenerationChunk> {
     const messagesMapped: OpenAICompletionParam[] =
-      _convertMessagesToOpenAIParams(messages);
+      _convertMessagesToOpenAIParams(messages, this.model);
     const params = {
       ...this.invocationParams(options, {
         streaming: true,
@@ -1489,7 +1523,7 @@ export class ChatOpenAI<
     const usageMetadata = {} as UsageMetadata;
     const params = this.invocationParams(options);
     const messagesMapped: OpenAICompletionParam[] =
-      _convertMessagesToOpenAIParams(messages);
+      _convertMessagesToOpenAIParams(messages, this.model);
 
     if (params.stream) {
       const stream = this._streamResponseChunks(messages, options, runManager);

libs/langchain-openai/src/tests/chat_models.int.test.ts

@@ -1157,8 +1157,6 @@ describe("Audio output", () => {
         content: [userInput],
       }),
     ]);
-    // console.log("userInputRes.content", userInputRes.content);
-    // console.log("userInputRes.additional_kwargs.audio", userInputRes.additional_kwargs.audio);
     expect(userInputRes.additional_kwargs.audio).toBeTruthy();
     expect(
       (userInputRes.additional_kwargs.audio as Record<string, any>).transcript
@@ -1191,6 +1189,23 @@ test("Can stream o1 requests", async () => {
     expect(finalMsg.content.length).toBeGreaterThanOrEqual(1);
   }
 
-  // A
   expect(numChunks).toBeGreaterThan(3);
 });
+
+test("Allows developer messages with o1", async () => {
+  const model = new ChatOpenAI({
+    model: "o1",
+    reasoningEffort: "low",
+  });
+  const res = await model.invoke([
+    {
+      role: "developer",
+      content: `Always respond only with the word "testing"`,
+    },
+    {
+      role: "user",
+      content: "hi",
+    },
+  ]);
+  expect(res.content).toEqual("testing");
+});

libs/langchain-openai/src/tests/chat_models_structured_output.int.test.ts

@@ -36,6 +36,35 @@ test("withStructuredOutput zod schema function calling", async () => {
   expect("number2" in result).toBe(true);
 });
 
+test("withStructuredOutput with o1", async () => {
+  const model = new ChatOpenAI({
+    model: "o1",
+  });
+
+  const calculatorSchema = z.object({
+    operation: z.enum(["add", "subtract", "multiply", "divide"]),
+    number1: z.number(),
+    number2: z.number(),
+  });
+  const modelWithStructuredOutput = model.withStructuredOutput(
+    calculatorSchema,
+    {
+      name: "calculator",
+    }
+  );
+
+  const prompt = ChatPromptTemplate.fromMessages([
+    ["developer", "You are VERY bad at math and must always use a calculator."],
+    ["human", "Please help me!! What is 2 + 2?"],
+  ]);
+  const chain = prompt.pipe(modelWithStructuredOutput);
+  const result = await chain.invoke({});
+  // console.log(result);
+  expect("operation" in result).toBe(true);
+  expect("number1" in result).toBe(true);
+  expect("number2" in result).toBe(true);
+});
+
 test("withStructuredOutput zod schema streaming", async () => {
   const model = new ChatOpenAI({
     temperature: 0,

libs/langchain-openai/src/types.ts

@@ -188,6 +188,12 @@ export interface OpenAIChatInput extends OpenAIBaseInput {
    * [Learn more](https://platform.openai.com/docs/guides/audio).
    */
   audio?: OpenAIClient.Chat.ChatCompletionAudioParam;
+
+  /**
+   * Constrains effort on reasoning for reasoning models. Currently supported values are low, medium, and high.
+   * Reducing reasoning effort can result in faster responses and fewer tokens used on reasoning in a response.
+   */
+  reasoningEffort?: OpenAIClient.ChatCompletionReasoningEffort;
 }
 
 export declare interface AzureOpenAIInput {

yarn.lock

@@ -12904,7 +12904,7 @@ __metadata:
     jest: ^29.5.0
     jest-environment-node: ^29.6.4
     js-tiktoken: ^1.0.12
-    openai: ^4.71.0
+    openai: ^4.77.0
     prettier: ^2.8.3
     release-it: ^17.6.0
     rimraf: ^5.0.1
@@ -36216,9 +36216,9 @@ __metadata:
   languageName: node
   linkType: hard
 
-"openai@npm:^4.71.0":
-  version: 4.71.0
-  resolution: "openai@npm:4.71.0"
+"openai@npm:^4.77.0":
+  version: 4.77.0
+  resolution: "openai@npm:4.77.0"
   dependencies:
     "@types/node": ^18.11.18
     "@types/node-fetch": ^2.6.4
@@ -36234,7 +36234,7 @@ __metadata:
       optional: true
   bin:
     openai: bin/cli
-  checksum: ba4b3772e806c59b1ea1235a40486392c797906e45dd97914f2cd819b4be2996e207c7b7c67d43236692300354f4e9ffa8ebfca6e97d3555655ebf0f3f01e3f2
+  checksum: e311130e3b35a7dc924e7125cca3246b7ac958b3c451f2f4ef2cae72144c82429f9c6db7bf67059fc9e10f3911087ebf8a9a4b2919ac915235ed3c324897b146
   languageName: node
   linkType: hard