browser-use 0.5.0 → 0.6.0

package/dist/llm/models.js

@@ -6,7 +6,9 @@ import { ChatCerebras } from './cerebras/chat.js';
 import { ChatDeepSeek } from './deepseek/chat.js';
 import { ChatGoogle } from './google/chat.js';
 import { ChatGroq } from './groq/chat.js';
+import { ChatLiteLLM } from './litellm/chat.js';
 import { ChatMistral } from './mistral/chat.js';
+import { ChatOCIRaw } from './oci-raw/chat.js';
 import { ChatOllama } from './ollama/chat.js';
 import { ChatOpenAI } from './openai/chat.js';
 import { ChatOpenRouter } from './openrouter/chat.js';
@@ -25,6 +27,8 @@ const AVAILABLE_PROVIDERS = [
     'mistral',
     'cerebras',
     'vercel',
+    'litellm',
+    'oci',
 ];
 const MISTRAL_ALIAS_MAP = {
     large: 'mistral-large-latest',
@@ -132,6 +136,12 @@ const inferProviderFromModel = (model) => {
     if (lower.startsWith('vercel:')) {
         return 'vercel';
     }
+    if (lower.startsWith('litellm:')) {
+        return 'litellm';
+    }
+    if (lower.startsWith('oci:')) {
+        return 'oci';
+    }
     if (lower.startsWith('mistral-') ||
         lower.startsWith('codestral') ||
         lower.startsWith('pixtral')) {
@@ -182,6 +192,12 @@ const normalizeModelForProvider = (provider, model) => {
     if (provider === 'vercel' && lower.startsWith('vercel:')) {
         return model.slice('vercel:'.length);
     }
+    if (provider === 'litellm' && lower.startsWith('litellm:')) {
+        return model.slice('litellm:'.length);
+    }
+    if (provider === 'oci' && lower.startsWith('oci:')) {
+        return model.slice('oci:'.length);
+    }
     if (provider === 'aws' && lower.startsWith('bedrock:')) {
         return model.slice('bedrock:'.length);
     }
@@ -248,6 +264,16 @@ const buildProviderModel = (provider, model) => {
                 apiKey: process.env.VERCEL_API_KEY,
                 baseURL: process.env.VERCEL_BASE_URL,
             });
+        case 'litellm':
+            return new ChatLiteLLM({
+                model,
+                apiKey: process.env.LITELLM_API_KEY,
+                baseURL: process.env.LITELLM_API_BASE ?? process.env.LITELLM_BASE_URL,
+            });
+        case 'oci':
+            return new ChatOCIRaw({
+                model,
+            });
         case 'aws':
             return new ChatBedrockConverse({
                 model,
@@ -294,13 +320,13 @@ export const getLlmByName = (modelName) => {
     const separator = normalizedName.indexOf('_');
     if (separator > 0) {
         const provider = normalizedName.slice(0, separator);
-        if (provider === 'oci') {
-            throw new Error('OCI models require manual configuration. Use ChatOCIRaw directly with your OCI credentials.');
-        }
         const modelPart = normalizedName.slice(separator + 1);
         if (provider === 'bu') {
             return buildProviderModel('browser-use', `bu-${modelPart.replace(/_/g, '-')}`);
         }
+        if (provider === 'oci') {
+            return buildProviderModel('oci', modelPart);
+        }
         if (provider === 'browser-use') {
             return buildProviderModel('browser-use', modelPart.replace(/_/g, '/'));
         }

package/dist/llm/oci-raw/chat.d.ts

@@ -0,0 +1,64 @@
+import type { BaseChatModel, ChatInvokeOptions } from '../base.js';
+import type { Message } from '../messages.js';
+import { ChatInvokeCompletion } from '../views.js';
+export interface ChatOCIRawOptions {
+    model?: string;
+    modelId?: string;
+    serviceEndpoint?: string;
+    compartmentId?: string;
+    provider?: string;
+    temperature?: number | null;
+    maxTokens?: number | null;
+    frequencyPenalty?: number | null;
+    presencePenalty?: number | null;
+    topP?: number | null;
+    topK?: number | null;
+    authType?: 'API_KEY' | 'INSTANCE_PRINCIPAL' | 'RESOURCE_PRINCIPAL' | string;
+    authProfile?: string;
+    configFilePath?: string;
+    tenancyId?: string;
+    userId?: string;
+    fingerprint?: string;
+    privateKey?: string;
+    passphrase?: string | null;
+    responseSchemaName?: string;
+}
+export declare class ChatOCIRaw implements BaseChatModel {
+    model: string;
+    provider: string;
+    private readonly serviceEndpoint;
+    private readonly compartmentId;
+    private readonly ociProvider;
+    private readonly temperature;
+    private readonly maxTokens;
+    private readonly frequencyPenalty;
+    private readonly presencePenalty;
+    private readonly topP;
+    private readonly topK;
+    private readonly authType;
+    private readonly authProfile;
+    private readonly configFilePath?;
+    private readonly tenancyId?;
+    private readonly userId?;
+    private readonly fingerprint?;
+    private readonly privateKey?;
+    private readonly passphrase;
+    private readonly responseSchemaName;
+    private clientPromise;
+    constructor(options?: ChatOCIRawOptions);
+    get name(): string;
+    get model_name(): string;
+    private usesCohereFormat;
+    private createAuthProvider;
+    private getClient;
+    private getUsage;
+    private buildGenericRequest;
+    private buildCohereRequest;
+    private buildChatRequest;
+    private extractText;
+    private mapError;
+    ainvoke(messages: Message[], outputFormat?: undefined, options?: ChatInvokeOptions): Promise<ChatInvokeCompletion<string>>;
+    ainvoke<T>(messages: Message[], outputFormat: {
+        parse: (input: unknown) => T;
+    } | undefined, options?: ChatInvokeOptions): Promise<ChatInvokeCompletion<T>>;
+}

package/dist/llm/oci-raw/chat.js

@@ -0,0 +1,350 @@
+import { ConfigFileAuthenticationDetailsProvider, InstancePrincipalsAuthenticationDetailsProviderBuilder, ResourcePrincipalAuthenticationDetailsProvider, SimpleAuthenticationDetailsProvider, } from 'oci-common';
+import { GenerativeAiInferenceClient } from 'oci-generativeaiinference';
+import { ModelProviderError, ModelRateLimitError } from '../exceptions.js';
+import { SchemaOptimizer, zodSchemaToJsonSchema } from '../schema.js';
+import { ChatInvokeCompletion } from '../views.js';
+import { OCIRawMessageSerializer } from './serializer.js';
+const parseStructuredJson = (text) => {
+    let jsonText = String(text ?? '').trim();
+    if (jsonText.startsWith('```json') && jsonText.endsWith('```')) {
+        jsonText = jsonText.slice(7, -3).trim();
+    }
+    else if (jsonText.startsWith('```') && jsonText.endsWith('```')) {
+        jsonText = jsonText.slice(3, -3).trim();
+    }
+    return JSON.parse(jsonText);
+};
+const extractZodSchema = (outputFormat) => {
+    const direct = outputFormat;
+    if (direct &&
+        typeof direct === 'object' &&
+        typeof direct.safeParse === 'function' &&
+        typeof direct.parse === 'function') {
+        return direct;
+    }
+    const nested = outputFormat?.schema;
+    if (nested &&
+        typeof nested === 'object' &&
+        typeof nested.safeParse === 'function' &&
+        typeof nested.parse === 'function') {
+        return nested;
+    }
+    return null;
+};
+const parseOutput = (outputFormat, payload) => {
+    const maybeWrapped = outputFormat;
+    if (maybeWrapped &&
+        typeof maybeWrapped === 'object' &&
+        maybeWrapped.schema &&
+        typeof maybeWrapped.schema.parse === 'function') {
+        return maybeWrapped.schema.parse(payload);
+    }
+    return outputFormat.parse(payload);
+};
+const appendJsonInstruction = (messages, schema) => {
+    const instruction = '\n\nIMPORTANT: You must respond with ONLY a valid JSON object ' +
+        '(no markdown, no code blocks, no explanations)' +
+        (schema
+            ? ` that exactly matches this schema:\n${JSON.stringify(schema, null, 2)}`
+            : '.');
+    const target = [...messages]
+        .reverse()
+        .find((message) => message.role === 'USER' || message.role === 'SYSTEM') ?? null;
+    if (target) {
+        const content = Array.isArray(target.content)
+            ? [...target.content]
+            : [];
+        content.push({
+            type: 'TEXT',
+            text: instruction,
+        });
+        target.content = content;
+        return messages;
+    }
+    return [
+        {
+            role: 'SYSTEM',
+            content: [
+                {
+                    type: 'TEXT',
+                    text: instruction,
+                },
+            ],
+        },
+        ...messages,
+    ];
+};
+export class ChatOCIRaw {
+    model;
+    provider = 'oci-raw';
+    serviceEndpoint;
+    compartmentId;
+    ociProvider;
+    temperature;
+    maxTokens;
+    frequencyPenalty;
+    presencePenalty;
+    topP;
+    topK;
+    authType;
+    authProfile;
+    configFilePath;
+    tenancyId;
+    userId;
+    fingerprint;
+    privateKey;
+    passphrase;
+    responseSchemaName;
+    clientPromise = null;
+    constructor(options = {}) {
+        this.model =
+            options.model ??
+                options.modelId ??
+                process.env.OCI_MODEL_ID ??
+                (() => {
+                    throw new Error('ChatOCIRaw requires model or OCI_MODEL_ID');
+                })();
+        this.serviceEndpoint =
+            options.serviceEndpoint ??
+                process.env.OCI_SERVICE_ENDPOINT ??
+                (() => {
+                    throw new Error('ChatOCIRaw requires serviceEndpoint or OCI_SERVICE_ENDPOINT');
+                })();
+        this.compartmentId =
+            options.compartmentId ??
+                process.env.OCI_COMPARTMENT_ID ??
+                (() => {
+                    throw new Error('ChatOCIRaw requires compartmentId or OCI_COMPARTMENT_ID');
+                })();
+        this.ociProvider = options.provider ?? process.env.OCI_PROVIDER ?? 'meta';
+        this.temperature = options.temperature ?? 1.0;
+        this.maxTokens = options.maxTokens ?? 600;
+        this.frequencyPenalty = options.frequencyPenalty ?? 0.0;
+        this.presencePenalty = options.presencePenalty ?? 0.0;
+        this.topP = options.topP ?? 0.75;
+        this.topK = options.topK ?? 0;
+        this.authType = options.authType ?? process.env.OCI_AUTH_TYPE ?? 'API_KEY';
+        this.authProfile =
+            options.authProfile ??
+                process.env.OCI_AUTH_PROFILE ??
+                process.env.OCI_CONFIG_PROFILE ??
+                'DEFAULT';
+        this.configFilePath = options.configFilePath ?? process.env.OCI_CONFIG_FILE;
+        this.tenancyId = options.tenancyId ?? process.env.OCI_TENANCY_ID;
+        this.userId = options.userId ?? process.env.OCI_USER_ID;
+        this.fingerprint = options.fingerprint ?? process.env.OCI_FINGERPRINT;
+        this.privateKey = options.privateKey ?? process.env.OCI_PRIVATE_KEY;
+        this.passphrase =
+            options.passphrase ?? process.env.OCI_PRIVATE_KEY_PASSPHRASE ?? null;
+        this.responseSchemaName =
+            options.responseSchemaName ?? 'browser_use_response';
+    }
+    get name() {
+        if (this.model.length <= 90) {
+            return this.model;
+        }
+        const parts = this.model.split('.');
+        if (parts.length >= 4) {
+            return `oci-${this.ociProvider}-${parts[3]}`;
+        }
+        return `oci-${this.ociProvider}-model`;
+    }
+    get model_name() {
+        return this.name;
+    }
+    usesCohereFormat() {
+        return this.ociProvider.toLowerCase() === 'cohere';
+    }
+    async createAuthProvider() {
+        const authType = this.authType.toUpperCase();
+        if (authType === 'INSTANCE_PRINCIPAL') {
+            return new InstancePrincipalsAuthenticationDetailsProviderBuilder().build();
+        }
+        if (authType === 'RESOURCE_PRINCIPAL') {
+            return ResourcePrincipalAuthenticationDetailsProvider.builder();
+        }
+        if (this.tenancyId && this.userId && this.fingerprint && this.privateKey) {
+            return new SimpleAuthenticationDetailsProvider(this.tenancyId, this.userId, this.fingerprint, this.privateKey, this.passphrase);
+        }
+        return new ConfigFileAuthenticationDetailsProvider(this.configFilePath, this.authProfile);
+    }
+    async getClient() {
+        if (!this.clientPromise) {
+            this.clientPromise = (async () => {
+                const authenticationDetailsProvider = await this.createAuthProvider();
+                const client = new GenerativeAiInferenceClient({
+                    authenticationDetailsProvider,
+                });
+                client.endpoint = this.serviceEndpoint;
+                return client;
+            })();
+        }
+        return this.clientPromise;
+    }
+    getUsage(payload) {
+        const usage = payload?.usage;
+        if (!usage) {
+            return null;
+        }
+        const reasoningTokens = usage.completionTokensDetails?.reasoningTokens ?? 0;
+        const completionTokens = (usage.completionTokens ?? 0) + reasoningTokens;
+        return {
+            prompt_tokens: usage.promptTokens ?? 0,
+            prompt_cached_tokens: usage.promptTokensDetails?.cachedTokens ?? null,
+            prompt_cache_creation_tokens: null,
+            prompt_image_tokens: null,
+            completion_tokens: completionTokens,
+            total_tokens: usage.totalTokens ?? (usage.promptTokens ?? 0) + completionTokens,
+        };
+    }
+    buildGenericRequest(messages, outputFormat) {
+        const zodSchema = extractZodSchema(outputFormat);
+        const serializedMessages = OCIRawMessageSerializer.serializeMessages(messages);
+        const optimizedSchema = zodSchema
+            ? SchemaOptimizer.createOptimizedJsonSchema(zodSchemaToJsonSchema(zodSchema))
+            : null;
+        const requestMessages = outputFormat && !zodSchema
+            ? appendJsonInstruction(serializedMessages, null)
+            : serializedMessages;
+        const request = {
+            apiFormat: 'GENERIC',
+            messages: requestMessages,
+        };
+        if (this.temperature !== null) {
+            request.temperature = this.temperature;
+        }
+        if (this.maxTokens !== null) {
+            request.maxTokens = this.maxTokens;
+        }
+        if (this.frequencyPenalty !== null) {
+            request.frequencyPenalty = this.frequencyPenalty;
+        }
+        if (this.presencePenalty !== null) {
+            request.presencePenalty = this.presencePenalty;
+        }
+        if (this.topP !== null) {
+            request.topP = this.topP;
+        }
+        if (this.topK !== null) {
+            request.topK = this.topK;
+        }
+        if (optimizedSchema) {
+            request.responseFormat = {
+                type: 'JSON_SCHEMA',
+                jsonSchema: {
+                    name: this.responseSchemaName,
+                    schema: optimizedSchema,
+                    isStrict: true,
+                },
+            };
+        }
+        return request;
+    }
+    buildCohereRequest(messages, outputFormat) {
+        const zodSchema = extractZodSchema(outputFormat);
+        const optimizedSchema = zodSchema
+            ? SchemaOptimizer.createOptimizedJsonSchema(zodSchemaToJsonSchema(zodSchema))
+            : null;
+        let conversation = OCIRawMessageSerializer.serializeMessagesForCohere(messages);
+        if (outputFormat) {
+            conversation +=
+                '\n\nIMPORTANT: You must respond with ONLY a valid JSON object (no markdown, no code blocks, no explanations)' +
+                    (optimizedSchema
+                        ? ` that exactly matches this schema:\n${JSON.stringify(optimizedSchema, null, 2)}`
+                        : '.');
+        }
+        const request = {
+            apiFormat: 'COHERE',
+            message: conversation,
+        };
+        if (this.temperature !== null) {
+            request.temperature = this.temperature;
+        }
+        if (this.maxTokens !== null) {
+            request.maxTokens = this.maxTokens;
+        }
+        if (this.frequencyPenalty !== null) {
+            request.frequencyPenalty = this.frequencyPenalty;
+        }
+        if (this.presencePenalty !== null) {
+            request.presencePenalty = this.presencePenalty;
+        }
+        if (this.topP !== null) {
+            request.topP = this.topP;
+        }
+        if (this.topK !== null) {
+            request.topK = this.topK;
+        }
+        return request;
+    }
+    buildChatRequest(messages, outputFormat) {
+        const chatRequest = this.usesCohereFormat()
+            ? this.buildCohereRequest(messages, outputFormat)
+            : this.buildGenericRequest(messages, outputFormat);
+        return {
+            chatDetails: {
+                compartmentId: this.compartmentId,
+                servingMode: {
+                    servingType: 'ON_DEMAND',
+                    modelId: this.model,
+                },
+                chatRequest,
+            },
+        };
+    }
+    extractText(payload) {
+        if (!payload) {
+            throw new ModelProviderError('OCI response did not include chatResponse payload', 502, this.name);
+        }
+        if (payload.apiFormat === 'COHERE') {
+            const coherePayload = payload;
+            return {
+                text: coherePayload.text ?? '',
+                thinking: null,
+                stopReason: coherePayload.finishReason ?? null,
+            };
+        }
+        const genericPayload = payload;
+        const choice = genericPayload.choices?.[0];
+        const text = choice?.message?.content
+            ?.filter((part) => part.type === 'TEXT')
+            .map((part) => part.text ?? '')
+            .join('\n')
+            .trim();
+        return {
+            text: text ?? '',
+            thinking: choice?.message?.reasoningContent ?? null,
+            stopReason: choice?.finishReason ?? null,
+        };
+    }
+    mapError(error) {
+        const statusCode = typeof error?.statusCode === 'number'
+            ? error.statusCode
+            : 502;
+        const message = String(error?.message ?? error ?? 'OCI request failed');
+        if (statusCode === 429) {
+            throw new ModelRateLimitError(message, statusCode, this.name);
+        }
+        throw new ModelProviderError(message, statusCode, this.name);
+    }
+    async ainvoke(messages, outputFormat, _options = {}) {
+        try {
+            const client = await this.getClient();
+            const response = (await client.chat(this.buildChatRequest(messages, outputFormat)));
+            if (!response || typeof response !== 'object') {
+                throw new ModelProviderError('OCI chat request returned an empty response', 502, this.name);
+            }
+            const payload = response.chatResult?.chatResponse;
+            const usage = this.getUsage(payload);
+            const { text, thinking, stopReason } = this.extractText(payload);
+            if (!outputFormat) {
+                return new ChatInvokeCompletion(text, usage, thinking, null, stopReason);
+            }
+            const parsed = parseOutput(outputFormat, parseStructuredJson(text));
+            return new ChatInvokeCompletion(parsed, usage, thinking, null, stopReason);
+        }
+        catch (error) {
+            this.mapError(error);
+        }
+    }
+}

package/dist/llm/oci-raw/index.d.ts

@@ -0,0 +1,2 @@
+export * from './chat.js';
+export * from './serializer.js';

package/dist/llm/oci-raw/index.js

@@ -0,0 +1,2 @@
+export * from './chat.js';
+export * from './serializer.js';

package/dist/llm/oci-raw/serializer.d.ts

@@ -0,0 +1,12 @@
+import { type Message } from '../messages.js';
+type OciChatContent = Record<string, unknown>;
+type OciMessage = {
+    role: string;
+    name?: string;
+    content: OciChatContent[];
+};
+export declare class OCIRawMessageSerializer {
+    static serializeMessages(messages: Message[]): OciMessage[];
+    static serializeMessagesForCohere(messages: Message[]): string;
+}
+export {};

package/dist/llm/oci-raw/serializer.js

@@ -0,0 +1,128 @@
+import { AssistantMessage, ContentPartImageParam, ContentPartRefusalParam, ContentPartTextParam, SystemMessage, UserMessage, } from '../messages.js';
+const textContent = (text) => ({
+    type: 'TEXT',
+    text,
+});
+const imageContent = (url) => ({
+    type: 'IMAGE',
+    imageUrl: {
+        url,
+    },
+});
+const contentPartsToOci = (content) => {
+    if (typeof content === 'string') {
+        return [textContent(content)];
+    }
+    if (!Array.isArray(content)) {
+        return [];
+    }
+    const parts = [];
+    for (const part of content) {
+        if (part instanceof ContentPartTextParam) {
+            parts.push(textContent(part.text));
+            continue;
+        }
+        if (part instanceof ContentPartImageParam) {
+            parts.push(imageContent(part.image_url.url));
+            continue;
+        }
+        if (part instanceof ContentPartRefusalParam) {
+            parts.push(textContent(`[Refusal] ${part.refusal}`));
+        }
+    }
+    return parts;
+};
+const serializeRole = (message) => {
+    if (message instanceof SystemMessage) {
+        return 'SYSTEM';
+    }
+    if (message instanceof AssistantMessage) {
+        return 'ASSISTANT';
+    }
+    return 'USER';
+};
+const serializeName = (message) => {
+    if (message instanceof UserMessage || message instanceof SystemMessage) {
+        return message.name ?? undefined;
+    }
+    return undefined;
+};
+export class OCIRawMessageSerializer {
+    static serializeMessages(messages) {
+        const serialized = [];
+        for (const message of messages) {
+            const content = message instanceof AssistantMessage
+                ? contentPartsToOci(message.content)
+                : contentPartsToOci(message.content);
+            if (content.length === 0) {
+                continue;
+            }
+            serialized.push({
+                role: serializeRole(message),
+                name: serializeName(message),
+                content,
+            });
+        }
+        return serialized;
+    }
+    static serializeMessagesForCohere(messages) {
+        const conversationParts = [];
+        for (const message of messages) {
+            let text = '';
+            if (message instanceof UserMessage || message instanceof SystemMessage) {
+                const content = message.content;
+                if (typeof content === 'string') {
+                    text = content;
+                }
+                else {
+                    text = content
+                        .map((part) => {
+                        if (part instanceof ContentPartTextParam) {
+                            return part.text;
+                        }
+                        if (part instanceof ContentPartImageParam) {
+                            return part.image_url.url.startsWith('data:image/')
+                                ? '[Image: base64_data]'
+                                : '[Image: external_url]';
+                        }
+                        return '';
+                    })
+                        .filter(Boolean)
+                        .join(' ');
+                }
+            }
+            else if (message instanceof AssistantMessage) {
+                if (typeof message.content === 'string') {
+                    text = message.content;
+                }
+                else if (Array.isArray(message.content)) {
+                    text = message.content
+                        .map((part) => {
+                        if (part instanceof ContentPartTextParam) {
+                            return part.text;
+                        }
+                        if (part instanceof ContentPartRefusalParam) {
+                            return `[Refusal] ${part.refusal}`;
+                        }
+                        return '';
+                    })
+                        .filter(Boolean)
+                        .join(' ');
+                }
+                else if (message.refusal) {
+                    text = `[Refusal] ${message.refusal}`;
+                }
+            }
+            if (!text) {
+                continue;
+            }
+            const prefix = message instanceof SystemMessage
+                ? 'System'
+                : message instanceof AssistantMessage
+                    ? 'Assistant'
+                    : 'User';
+            conversationParts.push(`${prefix}: ${text}`);
+        }
+        return conversationParts.join('\n\n');
+    }
+}

package/dist/mcp/server.d.ts

@@ -75,6 +75,7 @@ export declare class MCPServer {
     private cleanupExpiredSessions;
     private startSessionCleanupLoop;
     private stopSessionCleanupLoop;
+    private formatToolResult;
     private setupHandlers;
     private ensureController;
     private ensureBrowserSession;