Anirudh/add o3 (#452)

* add o3

* changeset

* fix

* reset example

* remove modelswithvision

* rm o3 from ci evals

Author: kamath

Diff for: .changeset/dull-buckets-cross.md

@@ -0,0 +1,5 @@
+---
+"@browserbasehq/stagehand": minor
+---
+
+add o3-mini to availablemodel

Diff for: examples/example.ts

@@ -9,8 +9,12 @@ import { Stagehand } from "@/dist";
 import StagehandConfig from "@/stagehand.config";
 
 async function example() {
-  const stagehand = new Stagehand(StagehandConfig);
+  const stagehand = new Stagehand({
+    ...StagehandConfig,
+    modelName: "o3-mini",
+  });
   await stagehand.init();
+  await stagehand.page.goto("https://www.google.com");
 }
 
 (async () => {

Diff for: lib/handlers/actHandler.ts

@@ -140,9 +140,8 @@ export class StagehandActHandler {
     // o1 is overkill for this task + this task uses a lot of tokens. So we switch it 4o
     let verifyLLmClient = llmClient;
     if (
-      llmClient.modelName === "o1-mini" ||
-      llmClient.modelName === "o1-preview" ||
-      llmClient.modelName.startsWith("o1-")
+      llmClient.modelName.startsWith("o1") ||
+      llmClient.modelName.startsWith("o3")
     ) {
       verifyLLmClient = this.llmProvider.getClient(
         "gpt-4o",

Diff for: lib/llm/LLMClient.ts

@@ -23,15 +23,6 @@ export interface ChatMessageTextContent {
   text: string;
 }
 
-export const modelsWithVision: AvailableModel[] = [
-  "gpt-4o",
-  "gpt-4o-mini",
-  "claude-3-5-sonnet-latest",
-  "claude-3-5-sonnet-20240620",
-  "claude-3-5-sonnet-20241022",
-  "gpt-4o-2024-08-06",
-];
-
 export const AnnotatedScreenshotText =
   "This is a screenshot of the current page state with the elements annotated on it. Each element id is annotated with a number to the top left of it. Duplicate annotations at the same location are under each other vertically.";
 
@@ -98,7 +89,6 @@ export abstract class LLMClient {
 
   constructor(modelName: AvailableModel, userProvidedInstructions?: string) {
     this.modelName = modelName;
-    this.hasVision = modelsWithVision.includes(modelName);
     this.userProvidedInstructions = userProvidedInstructions;
   }
 

Diff for: lib/llm/LLMProvider.ts

@@ -16,6 +16,7 @@ export class LLMProvider {
     "gpt-4o-2024-08-06": "openai",
     "o1-mini": "openai",
     "o1-preview": "openai",
+    "o3-mini": "openai",
     "claude-3-5-sonnet-latest": "anthropic",
     "claude-3-5-sonnet-20240620": "anthropic",
     "claude-3-5-sonnet-20241022": "anthropic",

Diff for: lib/llm/OpenAIClient.ts

@@ -59,7 +59,7 @@ export class OpenAIClient extends LLMClient {
     // O1 models do not support most of the options. So we override them.
     // For schema and tools, we add them as user messages.
     let isToolsOverridedForO1 = false;
-    if (this.modelName === "o1-mini" || this.modelName === "o1-preview") {
+    if (this.modelName.startsWith("o1") || this.modelName.startsWith("o3")) {
       /* eslint-disable */
       // Remove unsupported options
       let {
@@ -111,7 +111,7 @@ export class OpenAIClient extends LLMClient {
     }
     if (
       options.temperature &&
-      (this.modelName === "o1-mini" || this.modelName === "o1-preview")
+      (this.modelName.startsWith("o1") || this.modelName.startsWith("o3"))
     ) {
       throw new Error("Temperature is not supported for o1 models");
     }
@@ -207,7 +207,7 @@ export class OpenAIClient extends LLMClient {
     let responseFormat = undefined;
     if (options.response_model) {
       // For O1 models, we need to add the schema as a user message.
-      if (this.modelName === "o1-mini" || this.modelName === "o1-preview") {
+      if (this.modelName.startsWith("o1") || this.modelName.startsWith("o3")) {
         try {
           const parsedSchema = JSON.stringify(
             zodToJsonSchema(options.response_model.schema),

Diff for: types/model.ts

@@ -11,6 +11,7 @@ export const AvailableModelSchema = z.enum([
   "claude-3-5-sonnet-20240620",
   "o1-mini",
   "o1-preview",
+  "o3-mini",
 ]);
 
 export type AvailableModel = z.infer<typeof AvailableModelSchema>;