@browserbasehq/orca 3.0.7-alpha-1 → 3.0.8-google-cua
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +6 -6
- package/dist/index.d.ts +105 -71
- package/dist/index.js +1057 -651
- package/package.json +1 -1
package/README.md
CHANGED
|
@@ -20,10 +20,10 @@
|
|
|
20
20
|
<img alt="MIT License" src="../../media/light_license.svg" />
|
|
21
21
|
</picture>
|
|
22
22
|
</a>
|
|
23
|
-
<a href="https://
|
|
23
|
+
<a href="https://stagehand.dev/discord">
|
|
24
24
|
<picture>
|
|
25
|
-
<source media="(prefers-color-scheme: dark)" srcset="../../media/
|
|
26
|
-
<img alt="
|
|
25
|
+
<source media="(prefers-color-scheme: dark)" srcset="../../media/dark_discord.svg" />
|
|
26
|
+
<img alt="Discord Community" src="../../media/light_discord.svg" />
|
|
27
27
|
</picture>
|
|
28
28
|
</a>
|
|
29
29
|
</p>
|
|
@@ -136,10 +136,10 @@ In your project's `package.json` set:
|
|
|
136
136
|
|
|
137
137
|
## Contributing
|
|
138
138
|
|
|
139
|
-
> [!NOTE]
|
|
140
|
-
> We highly value contributions to Stagehand! For questions or support, please join our [
|
|
139
|
+
> [!NOTE]
|
|
140
|
+
> We highly value contributions to Stagehand! For questions or support, please join our [Discord community](https://stagehand.dev/discord).
|
|
141
141
|
|
|
142
|
-
At a high level, we're focused on improving reliability, extensibility, speed, and cost in that order of priority. If you're interested in contributing, **bug fixes and small improvements are the best way to get started**. For more involved features, we strongly recommend reaching out to [Miguel Gonzalez](https://x.com/miguel_gonzf) or [Paul Klein](https://x.com/pk_iv) in our [
|
|
142
|
+
At a high level, we're focused on improving reliability, extensibility, speed, and cost in that order of priority. If you're interested in contributing, **bug fixes and small improvements are the best way to get started**. For more involved features, we strongly recommend reaching out to [Miguel Gonzalez](https://x.com/miguel_gonzf) or [Paul Klein](https://x.com/pk_iv) in our [Discord community](https://stagehand.dev/discord) before starting to ensure that your contribution aligns with our goals.
|
|
143
143
|
|
|
144
144
|
<!-- For more information, please see our [Contributing Guide](https://docs.stagehand.dev/examples/contributing). -->
|
|
145
145
|
|
package/dist/index.d.ts
CHANGED
|
@@ -8,7 +8,7 @@ import { ClientOptions as ClientOptions$1 } from 'openai';
|
|
|
8
8
|
import { Client, ClientOptions as ClientOptions$3 } from '@modelcontextprotocol/sdk/client/index.js';
|
|
9
9
|
import * as ai from 'ai';
|
|
10
10
|
import { ToolSet, ModelMessage, PrepareStepFunction, GenerateTextOnStepFinishCallback, StreamTextOnStepFinishCallback, StreamTextOnErrorCallback, StreamTextOnChunkCallback, StreamTextOnFinishCallback, StepResult, StreamTextResult, wrapLanguageModel, generateObject, generateText, streamText, streamObject, experimental_generateImage, embed, embedMany, experimental_transcribe, experimental_generateSpeech, InferUITools } from 'ai';
|
|
11
|
-
export { ModelMessage } from 'ai';
|
|
11
|
+
export { ModelMessage, Tool, tool } from 'ai';
|
|
12
12
|
import { Page as Page$1 } from 'playwright-core';
|
|
13
13
|
export { Page as PlaywrightPage } from 'playwright-core';
|
|
14
14
|
import { Page as Page$2 } from 'puppeteer-core';
|
|
@@ -1069,6 +1069,10 @@ declare const SessionStartResponseSchema: z$1.ZodObject<{
|
|
|
1069
1069
|
available: z$1.ZodBoolean;
|
|
1070
1070
|
}, z$1.core.$strip>;
|
|
1071
1071
|
}, z$1.core.$strip>;
|
|
1072
|
+
/** Session end request - empty JSON object (required). */
|
|
1073
|
+
declare const SessionEndRequestSchema: z$1.ZodObject<{
|
|
1074
|
+
_forceBody: z$1.ZodOptional<z$1.ZodUndefined>;
|
|
1075
|
+
}, z$1.core.$strict>;
|
|
1072
1076
|
declare const SessionEndResultSchema: z$1.ZodObject<{}, z$1.core.$strict>;
|
|
1073
1077
|
/** Session end response - just success flag, no data wrapper */
|
|
1074
1078
|
declare const SessionEndResponseSchema: z$1.ZodObject<{
|
|
@@ -1819,6 +1823,7 @@ type api_ReplayResponse = ReplayResponse;
|
|
|
1819
1823
|
declare const api_ReplayResponseSchema: typeof ReplayResponseSchema;
|
|
1820
1824
|
type api_ReplayResult = ReplayResult;
|
|
1821
1825
|
declare const api_ReplayResultSchema: typeof ReplayResultSchema;
|
|
1826
|
+
declare const api_SessionEndRequestSchema: typeof SessionEndRequestSchema;
|
|
1822
1827
|
type api_SessionEndResponse = SessionEndResponse;
|
|
1823
1828
|
declare const api_SessionEndResponseSchema: typeof SessionEndResponseSchema;
|
|
1824
1829
|
type api_SessionEndResult = SessionEndResult;
|
|
@@ -1848,7 +1853,7 @@ declare const api_TokenUsageSchema: typeof TokenUsageSchema;
|
|
|
1848
1853
|
declare const api_openApiLinks: typeof openApiLinks;
|
|
1849
1854
|
declare const api_openApiSecuritySchemes: typeof openApiSecuritySchemes;
|
|
1850
1855
|
declare namespace api {
|
|
1851
|
-
export { api_ActOptionsSchema as ActOptionsSchema, type api_ActRequest as ActRequest, api_ActRequestSchema as ActRequestSchema, type api_ActResponse as ActResponse, api_ActResponseSchema as ActResponseSchema, type ActResult$1 as ActResult, type api_ActResultData as ActResultData, api_ActResultDataSchema as ActResultDataSchema, api_ActResultSchema as ActResultSchema, type Action$1 as Action, api_ActionSchema as ActionSchema, type AgentAction$1 as AgentAction, api_AgentActionSchema as AgentActionSchema, api_AgentConfigSchema as AgentConfigSchema, api_AgentExecuteOptionsSchema as AgentExecuteOptionsSchema, type api_AgentExecuteRequest as AgentExecuteRequest, api_AgentExecuteRequestSchema as AgentExecuteRequestSchema, type api_AgentExecuteResponse as AgentExecuteResponse, api_AgentExecuteResponseSchema as AgentExecuteResponseSchema, type api_AgentExecuteResult as AgentExecuteResult, api_AgentExecuteResultSchema as AgentExecuteResultSchema, type api_AgentResultData as AgentResultData, api_AgentResultDataSchema as AgentResultDataSchema, type api_AgentUsage as AgentUsage, api_AgentUsageSchema as AgentUsageSchema, type api_BrowserConfig as BrowserConfig, api_BrowserConfigSchema as BrowserConfigSchema, type api_BrowserbaseBrowserSettings as BrowserbaseBrowserSettings, api_BrowserbaseBrowserSettingsSchema as BrowserbaseBrowserSettingsSchema, type api_BrowserbaseContext as BrowserbaseContext, api_BrowserbaseContextSchema as BrowserbaseContextSchema, type api_BrowserbaseFingerprint as BrowserbaseFingerprint, api_BrowserbaseFingerprintSchema as BrowserbaseFingerprintSchema, type api_BrowserbaseFingerprintScreen as BrowserbaseFingerprintScreen, api_BrowserbaseFingerprintScreenSchema as BrowserbaseFingerprintScreenSchema, type api_BrowserbaseProxyConfig as BrowserbaseProxyConfig, api_BrowserbaseProxyConfigSchema as BrowserbaseProxyConfigSchema, type api_BrowserbaseProxyGeolocation as BrowserbaseProxyGeolocation, api_BrowserbaseProxyGeolocationSchema as BrowserbaseProxyGeolocationSchema, type api_BrowserbaseSessionCreateParams as BrowserbaseSessionCreateParams, api_BrowserbaseSessionCreateParamsSchema as BrowserbaseSessionCreateParamsSchema, type api_BrowserbaseViewport as BrowserbaseViewport, api_BrowserbaseViewportSchema as BrowserbaseViewportSchema, api_ErrorResponseSchema as ErrorResponseSchema, type api_ExternalProxyConfig as ExternalProxyConfig, api_ExternalProxyConfigSchema as ExternalProxyConfigSchema, api_ExtractOptionsSchema as ExtractOptionsSchema, type api_ExtractRequest as ExtractRequest, api_ExtractRequestSchema as ExtractRequestSchema, type api_ExtractResponse as ExtractResponse, api_ExtractResponseSchema as ExtractResponseSchema, type ExtractResult$1 as ExtractResult, api_ExtractResultSchema as ExtractResultSchema, api_LocalBrowserLaunchOptionsSchema as LocalBrowserLaunchOptionsSchema, type api_ModelConfig as ModelConfig, api_ModelConfigObjectSchema as ModelConfigObjectSchema, api_ModelConfigSchema as ModelConfigSchema, api_ModelNameSchema as ModelNameSchema, api_NavigateOptionsSchema as NavigateOptionsSchema, type api_NavigateRequest as NavigateRequest, api_NavigateRequestSchema as NavigateRequestSchema, type api_NavigateResponse as NavigateResponse, api_NavigateResponseSchema as NavigateResponseSchema, type api_NavigateResult as NavigateResult, api_NavigateResultSchema as NavigateResultSchema, api_ObserveOptionsSchema as ObserveOptionsSchema, type api_ObserveRequest as ObserveRequest, api_ObserveRequestSchema as ObserveRequestSchema, type api_ObserveResponse as ObserveResponse, api_ObserveResponseSchema as ObserveResponseSchema, type api_ObserveResult as ObserveResult, api_ObserveResultSchema as ObserveResultSchema, api_Operations as Operations, api_ProxyConfigSchema as ProxyConfigSchema, type api_ReplayAction as ReplayAction, api_ReplayActionSchema as ReplayActionSchema, type api_ReplayPage as ReplayPage, api_ReplayPageSchema as ReplayPageSchema, type api_ReplayResponse as ReplayResponse, api_ReplayResponseSchema as ReplayResponseSchema, type api_ReplayResult as ReplayResult, api_ReplayResultSchema as ReplayResultSchema, type api_SessionEndResponse as SessionEndResponse, api_SessionEndResponseSchema as SessionEndResponseSchema, type api_SessionEndResult as SessionEndResult, api_SessionEndResultSchema as SessionEndResultSchema, type api_SessionHeaders as SessionHeaders, api_SessionHeadersSchema as SessionHeadersSchema, type api_SessionIdParams as SessionIdParams, api_SessionIdParamsSchema as SessionIdParamsSchema, type api_SessionStartRequest as SessionStartRequest, api_SessionStartRequestSchema as SessionStartRequestSchema, type api_SessionStartResponse as SessionStartResponse, api_SessionStartResponseSchema as SessionStartResponseSchema, type api_SessionStartResult as SessionStartResult, api_SessionStartResultSchema as SessionStartResultSchema, type api_StreamEvent as StreamEvent, type api_StreamEventLogData as StreamEventLogData, api_StreamEventLogDataSchema as StreamEventLogDataSchema, api_StreamEventSchema as StreamEventSchema, type api_StreamEventStatus as StreamEventStatus, api_StreamEventStatusSchema as StreamEventStatusSchema, type api_StreamEventSystemData as StreamEventSystemData, api_StreamEventSystemDataSchema as StreamEventSystemDataSchema, type api_StreamEventType as StreamEventType, api_StreamEventTypeSchema as StreamEventTypeSchema, type api_TokenUsage as TokenUsage, api_TokenUsageSchema as TokenUsageSchema, api_openApiLinks as openApiLinks, api_openApiSecuritySchemes as openApiSecuritySchemes };
|
|
1856
|
+
export { api_ActOptionsSchema as ActOptionsSchema, type api_ActRequest as ActRequest, api_ActRequestSchema as ActRequestSchema, type api_ActResponse as ActResponse, api_ActResponseSchema as ActResponseSchema, type ActResult$1 as ActResult, type api_ActResultData as ActResultData, api_ActResultDataSchema as ActResultDataSchema, api_ActResultSchema as ActResultSchema, type Action$1 as Action, api_ActionSchema as ActionSchema, type AgentAction$1 as AgentAction, api_AgentActionSchema as AgentActionSchema, api_AgentConfigSchema as AgentConfigSchema, api_AgentExecuteOptionsSchema as AgentExecuteOptionsSchema, type api_AgentExecuteRequest as AgentExecuteRequest, api_AgentExecuteRequestSchema as AgentExecuteRequestSchema, type api_AgentExecuteResponse as AgentExecuteResponse, api_AgentExecuteResponseSchema as AgentExecuteResponseSchema, type api_AgentExecuteResult as AgentExecuteResult, api_AgentExecuteResultSchema as AgentExecuteResultSchema, type api_AgentResultData as AgentResultData, api_AgentResultDataSchema as AgentResultDataSchema, type api_AgentUsage as AgentUsage, api_AgentUsageSchema as AgentUsageSchema, type api_BrowserConfig as BrowserConfig, api_BrowserConfigSchema as BrowserConfigSchema, type api_BrowserbaseBrowserSettings as BrowserbaseBrowserSettings, api_BrowserbaseBrowserSettingsSchema as BrowserbaseBrowserSettingsSchema, type api_BrowserbaseContext as BrowserbaseContext, api_BrowserbaseContextSchema as BrowserbaseContextSchema, type api_BrowserbaseFingerprint as BrowserbaseFingerprint, api_BrowserbaseFingerprintSchema as BrowserbaseFingerprintSchema, type api_BrowserbaseFingerprintScreen as BrowserbaseFingerprintScreen, api_BrowserbaseFingerprintScreenSchema as BrowserbaseFingerprintScreenSchema, type api_BrowserbaseProxyConfig as BrowserbaseProxyConfig, api_BrowserbaseProxyConfigSchema as BrowserbaseProxyConfigSchema, type api_BrowserbaseProxyGeolocation as BrowserbaseProxyGeolocation, api_BrowserbaseProxyGeolocationSchema as BrowserbaseProxyGeolocationSchema, type api_BrowserbaseSessionCreateParams as BrowserbaseSessionCreateParams, api_BrowserbaseSessionCreateParamsSchema as BrowserbaseSessionCreateParamsSchema, type api_BrowserbaseViewport as BrowserbaseViewport, api_BrowserbaseViewportSchema as BrowserbaseViewportSchema, api_ErrorResponseSchema as ErrorResponseSchema, type api_ExternalProxyConfig as ExternalProxyConfig, api_ExternalProxyConfigSchema as ExternalProxyConfigSchema, api_ExtractOptionsSchema as ExtractOptionsSchema, type api_ExtractRequest as ExtractRequest, api_ExtractRequestSchema as ExtractRequestSchema, type api_ExtractResponse as ExtractResponse, api_ExtractResponseSchema as ExtractResponseSchema, type ExtractResult$1 as ExtractResult, api_ExtractResultSchema as ExtractResultSchema, api_LocalBrowserLaunchOptionsSchema as LocalBrowserLaunchOptionsSchema, type api_ModelConfig as ModelConfig, api_ModelConfigObjectSchema as ModelConfigObjectSchema, api_ModelConfigSchema as ModelConfigSchema, api_ModelNameSchema as ModelNameSchema, api_NavigateOptionsSchema as NavigateOptionsSchema, type api_NavigateRequest as NavigateRequest, api_NavigateRequestSchema as NavigateRequestSchema, type api_NavigateResponse as NavigateResponse, api_NavigateResponseSchema as NavigateResponseSchema, type api_NavigateResult as NavigateResult, api_NavigateResultSchema as NavigateResultSchema, api_ObserveOptionsSchema as ObserveOptionsSchema, type api_ObserveRequest as ObserveRequest, api_ObserveRequestSchema as ObserveRequestSchema, type api_ObserveResponse as ObserveResponse, api_ObserveResponseSchema as ObserveResponseSchema, type api_ObserveResult as ObserveResult, api_ObserveResultSchema as ObserveResultSchema, api_Operations as Operations, api_ProxyConfigSchema as ProxyConfigSchema, type api_ReplayAction as ReplayAction, api_ReplayActionSchema as ReplayActionSchema, type api_ReplayPage as ReplayPage, api_ReplayPageSchema as ReplayPageSchema, type api_ReplayResponse as ReplayResponse, api_ReplayResponseSchema as ReplayResponseSchema, type api_ReplayResult as ReplayResult, api_ReplayResultSchema as ReplayResultSchema, api_SessionEndRequestSchema as SessionEndRequestSchema, type api_SessionEndResponse as SessionEndResponse, api_SessionEndResponseSchema as SessionEndResponseSchema, type api_SessionEndResult as SessionEndResult, api_SessionEndResultSchema as SessionEndResultSchema, type api_SessionHeaders as SessionHeaders, api_SessionHeadersSchema as SessionHeadersSchema, type api_SessionIdParams as SessionIdParams, api_SessionIdParamsSchema as SessionIdParamsSchema, type api_SessionStartRequest as SessionStartRequest, api_SessionStartRequestSchema as SessionStartRequestSchema, type api_SessionStartResponse as SessionStartResponse, api_SessionStartResponseSchema as SessionStartResponseSchema, type api_SessionStartResult as SessionStartResult, api_SessionStartResultSchema as SessionStartResultSchema, type api_StreamEvent as StreamEvent, type api_StreamEventLogData as StreamEventLogData, api_StreamEventLogDataSchema as StreamEventLogDataSchema, api_StreamEventSchema as StreamEventSchema, type api_StreamEventStatus as StreamEventStatus, api_StreamEventStatusSchema as StreamEventStatusSchema, type api_StreamEventSystemData as StreamEventSystemData, api_StreamEventSystemDataSchema as StreamEventSystemDataSchema, type api_StreamEventType as StreamEventType, api_StreamEventTypeSchema as StreamEventTypeSchema, type api_TokenUsage as TokenUsage, api_TokenUsageSchema as TokenUsageSchema, api_openApiLinks as openApiLinks, api_openApiSecuritySchemes as openApiSecuritySchemes };
|
|
1852
1857
|
}
|
|
1853
1858
|
|
|
1854
1859
|
declare class StagehandAPIError extends Error {
|
|
@@ -2800,6 +2805,12 @@ declare class Page {
|
|
|
2800
2805
|
* Mirrors Playwright's API signatures.
|
|
2801
2806
|
*/
|
|
2802
2807
|
waitForLoadState(state: LoadState, timeoutMs?: number): Promise<void>;
|
|
2808
|
+
/**
|
|
2809
|
+
* Wait for a specified amount of time.
|
|
2810
|
+
*
|
|
2811
|
+
* @param ms The number of milliseconds to wait.
|
|
2812
|
+
*/
|
|
2813
|
+
waitForTimeout(ms: number): Promise<void>;
|
|
2803
2814
|
/**
|
|
2804
2815
|
* Evaluate a function or expression in the current main frame's main world.
|
|
2805
2816
|
* - If a string is provided, it is treated as a JS expression.
|
|
@@ -2944,6 +2955,12 @@ interface AgentResult {
|
|
|
2944
2955
|
* @experimental
|
|
2945
2956
|
*/
|
|
2946
2957
|
messages?: ModelMessage[];
|
|
2958
|
+
/**
|
|
2959
|
+
* Custom output data extracted based on the `output` schema provided in execute options.
|
|
2960
|
+
* Only populated if an `output` schema was provided.
|
|
2961
|
+
* @experimental
|
|
2962
|
+
*/
|
|
2963
|
+
output?: Record<string, unknown>;
|
|
2947
2964
|
}
|
|
2948
2965
|
type AgentStreamResult = StreamTextResult<ToolSet, never> & {
|
|
2949
2966
|
result: Promise<AgentResult>;
|
|
@@ -3166,6 +3183,26 @@ interface AgentExecuteOptionsBase {
|
|
|
3166
3183
|
* ```
|
|
3167
3184
|
*/
|
|
3168
3185
|
excludeTools?: string[];
|
|
3186
|
+
/**
|
|
3187
|
+
* A Zod schema defining custom output data to return when the task completes.
|
|
3188
|
+
* The agent will populate this data in the final close tool call.
|
|
3189
|
+
*
|
|
3190
|
+
* @experimental
|
|
3191
|
+
* @example
|
|
3192
|
+
* ```typescript
|
|
3193
|
+
* const result = await agent.execute({
|
|
3194
|
+
* instruction: "Find the cheapest flight from NYC to LA",
|
|
3195
|
+
* output: z.object({
|
|
3196
|
+
* price: z.string().describe("The price of the flight"),
|
|
3197
|
+
* airline: z.string().describe("The airline name"),
|
|
3198
|
+
* departureTime: z.string().describe("Departure time"),
|
|
3199
|
+
* }),
|
|
3200
|
+
* });
|
|
3201
|
+
*
|
|
3202
|
+
* console.log(result.output); // { price: "$199", airline: "Delta", departureTime: "8:00 AM" }
|
|
3203
|
+
* ```
|
|
3204
|
+
*/
|
|
3205
|
+
output?: StagehandZodObject;
|
|
3169
3206
|
}
|
|
3170
3207
|
/**
|
|
3171
3208
|
* Options for non-streaming agent execution.
|
|
@@ -3190,7 +3227,7 @@ interface AgentStreamExecuteOptions extends AgentExecuteOptionsBase {
|
|
|
3190
3227
|
callbacks?: AgentStreamCallbacks;
|
|
3191
3228
|
}
|
|
3192
3229
|
type AgentType = "openai" | "anthropic" | "google" | "microsoft";
|
|
3193
|
-
declare const AVAILABLE_CUA_MODELS: readonly ["openai/computer-use-preview", "openai/computer-use-preview-2025-03-11", "anthropic/claude-3-7-sonnet-latest", "anthropic/claude-opus-4-5-20251101", "anthropic/claude-haiku-4-5-20251001", "anthropic/claude-sonnet-4-20250514", "anthropic/claude-sonnet-4-5-20250929", "google/gemini-2.5-computer-use-preview-10-2025", "microsoft/fara-7b"];
|
|
3230
|
+
declare const AVAILABLE_CUA_MODELS: readonly ["openai/computer-use-preview", "openai/computer-use-preview-2025-03-11", "anthropic/claude-3-7-sonnet-latest", "anthropic/claude-opus-4-5-20251101", "anthropic/claude-haiku-4-5-20251001", "anthropic/claude-sonnet-4-20250514", "anthropic/claude-sonnet-4-5-20250929", "google/gemini-2.5-computer-use-preview-10-2025", "google/gemini-3-flash-computer-use", "microsoft/fara-7b"];
|
|
3194
3231
|
type AvailableCuaModel = (typeof AVAILABLE_CUA_MODELS)[number];
|
|
3195
3232
|
interface AgentExecutionOptions<TOptions extends AgentExecuteOptions = AgentExecuteOptions> {
|
|
3196
3233
|
options: TOptions;
|
|
@@ -3397,6 +3434,63 @@ interface StreamingAgentInstance {
|
|
|
3397
3434
|
interface NonStreamingAgentInstance {
|
|
3398
3435
|
execute: (instructionOrOptions: string | AgentExecuteOptions) => Promise<AgentResult>;
|
|
3399
3436
|
}
|
|
3437
|
+
/**
|
|
3438
|
+
* Content item type for toModelOutput return values.
|
|
3439
|
+
* Used in tool definitions to return text and/or media to the model.
|
|
3440
|
+
*/
|
|
3441
|
+
type ModelOutputContentItem = {
|
|
3442
|
+
type: "text";
|
|
3443
|
+
text: string;
|
|
3444
|
+
} | {
|
|
3445
|
+
type: "media";
|
|
3446
|
+
mediaType: string;
|
|
3447
|
+
data: string;
|
|
3448
|
+
};
|
|
3449
|
+
interface ClickToolResult {
|
|
3450
|
+
success: boolean;
|
|
3451
|
+
describe?: string;
|
|
3452
|
+
coordinates?: number[];
|
|
3453
|
+
error?: string;
|
|
3454
|
+
screenshotBase64?: string;
|
|
3455
|
+
}
|
|
3456
|
+
interface TypeToolResult {
|
|
3457
|
+
success: boolean;
|
|
3458
|
+
describe?: string;
|
|
3459
|
+
text?: string;
|
|
3460
|
+
error?: string;
|
|
3461
|
+
screenshotBase64?: string;
|
|
3462
|
+
}
|
|
3463
|
+
interface DragAndDropToolResult {
|
|
3464
|
+
success: boolean;
|
|
3465
|
+
describe?: string;
|
|
3466
|
+
error?: string;
|
|
3467
|
+
screenshotBase64?: string;
|
|
3468
|
+
}
|
|
3469
|
+
interface FillFormField {
|
|
3470
|
+
action: string;
|
|
3471
|
+
value: string;
|
|
3472
|
+
coordinates: {
|
|
3473
|
+
x: number;
|
|
3474
|
+
y: number;
|
|
3475
|
+
};
|
|
3476
|
+
}
|
|
3477
|
+
interface FillFormVisionToolResult {
|
|
3478
|
+
success: boolean;
|
|
3479
|
+
playwrightArguments?: FillFormField[];
|
|
3480
|
+
error?: string;
|
|
3481
|
+
screenshotBase64?: string;
|
|
3482
|
+
}
|
|
3483
|
+
interface ScrollVisionToolResult {
|
|
3484
|
+
success: boolean;
|
|
3485
|
+
message: string;
|
|
3486
|
+
scrolledPixels: number;
|
|
3487
|
+
screenshotBase64?: string;
|
|
3488
|
+
}
|
|
3489
|
+
interface WaitToolResult {
|
|
3490
|
+
success: boolean;
|
|
3491
|
+
waited: number;
|
|
3492
|
+
screenshotBase64?: string;
|
|
3493
|
+
}
|
|
3400
3494
|
|
|
3401
3495
|
type OpenAIClientOptions = Pick<ClientOptions$1, "baseURL" | "apiKey">;
|
|
3402
3496
|
type AnthropicClientOptions = Pick<ClientOptions$2, "baseURL" | "apiKey">;
|
|
@@ -3809,12 +3903,9 @@ declare const screenshotTool: (v3: V3) => ai.Tool<Record<string, never>, {
|
|
|
3809
3903
|
pageUrl: string;
|
|
3810
3904
|
}>;
|
|
3811
3905
|
|
|
3812
|
-
declare const waitTool: (v3: V3) => ai.Tool<{
|
|
3906
|
+
declare const waitTool: (v3: V3, mode?: AgentToolMode) => ai.Tool<{
|
|
3813
3907
|
timeMs: number;
|
|
3814
|
-
},
|
|
3815
|
-
success: boolean;
|
|
3816
|
-
waited: number;
|
|
3817
|
-
}>;
|
|
3908
|
+
}, WaitToolResult>;
|
|
3818
3909
|
|
|
3819
3910
|
declare const navBackTool: (v3: V3) => ai.Tool<{
|
|
3820
3911
|
reasoningText: string;
|
|
@@ -3822,15 +3913,6 @@ declare const navBackTool: (v3: V3) => ai.Tool<{
|
|
|
3822
3913
|
success: boolean;
|
|
3823
3914
|
}>;
|
|
3824
3915
|
|
|
3825
|
-
declare const closeTool: () => ai.Tool<{
|
|
3826
|
-
reasoning: string;
|
|
3827
|
-
taskComplete: boolean;
|
|
3828
|
-
}, {
|
|
3829
|
-
success: boolean;
|
|
3830
|
-
reasoning: string;
|
|
3831
|
-
taskComplete: boolean;
|
|
3832
|
-
}>;
|
|
3833
|
-
|
|
3834
3916
|
declare const ariaTreeTool: (v3: V3) => ai.Tool<Record<string, never>, {
|
|
3835
3917
|
content: string;
|
|
3836
3918
|
pageUrl: string;
|
|
@@ -3867,11 +3949,7 @@ declare const scrollVisionTool: (v3: V3, provider?: string) => ai.Tool<{
|
|
|
3867
3949
|
direction: "up" | "down";
|
|
3868
3950
|
coordinates?: number[];
|
|
3869
3951
|
percentage?: number;
|
|
3870
|
-
},
|
|
3871
|
-
success: boolean;
|
|
3872
|
-
message: string;
|
|
3873
|
-
scrolledPixels: number;
|
|
3874
|
-
}>;
|
|
3952
|
+
}, ScrollVisionToolResult>;
|
|
3875
3953
|
|
|
3876
3954
|
declare const extractTool: (v3: V3, executionModel?: string, logger?: (message: LogLine) => void) => ai.Tool<{
|
|
3877
3955
|
instruction: string;
|
|
@@ -3889,47 +3967,19 @@ declare const extractTool: (v3: V3, executionModel?: string, logger?: (message:
|
|
|
3889
3967
|
declare const clickTool: (v3: V3, provider?: string) => ai.Tool<{
|
|
3890
3968
|
describe: string;
|
|
3891
3969
|
coordinates: number[];
|
|
3892
|
-
},
|
|
3893
|
-
success: boolean;
|
|
3894
|
-
describe: string;
|
|
3895
|
-
coordinates: number[];
|
|
3896
|
-
error?: undefined;
|
|
3897
|
-
} | {
|
|
3898
|
-
success: boolean;
|
|
3899
|
-
error: string;
|
|
3900
|
-
describe?: undefined;
|
|
3901
|
-
coordinates?: undefined;
|
|
3902
|
-
}>;
|
|
3970
|
+
}, ClickToolResult>;
|
|
3903
3971
|
|
|
3904
3972
|
declare const typeTool: (v3: V3, provider?: string) => ai.Tool<{
|
|
3905
3973
|
describe: string;
|
|
3906
3974
|
text: string;
|
|
3907
3975
|
coordinates: number[];
|
|
3908
|
-
},
|
|
3909
|
-
success: boolean;
|
|
3910
|
-
describe: string;
|
|
3911
|
-
text: string;
|
|
3912
|
-
error?: undefined;
|
|
3913
|
-
} | {
|
|
3914
|
-
success: boolean;
|
|
3915
|
-
error: string;
|
|
3916
|
-
describe?: undefined;
|
|
3917
|
-
text?: undefined;
|
|
3918
|
-
}>;
|
|
3976
|
+
}, TypeToolResult>;
|
|
3919
3977
|
|
|
3920
3978
|
declare const dragAndDropTool: (v3: V3, provider?: string) => ai.Tool<{
|
|
3921
3979
|
describe: string;
|
|
3922
3980
|
startCoordinates: number[];
|
|
3923
3981
|
endCoordinates: number[];
|
|
3924
|
-
},
|
|
3925
|
-
success: boolean;
|
|
3926
|
-
describe: string;
|
|
3927
|
-
error?: undefined;
|
|
3928
|
-
} | {
|
|
3929
|
-
success: boolean;
|
|
3930
|
-
error: string;
|
|
3931
|
-
describe?: undefined;
|
|
3932
|
-
}>;
|
|
3982
|
+
}, DragAndDropToolResult>;
|
|
3933
3983
|
|
|
3934
3984
|
declare const clickAndHoldTool: (v3: V3, provider?: string) => ai.Tool<{
|
|
3935
3985
|
describe: string;
|
|
@@ -3978,22 +4028,7 @@ declare const fillFormVisionTool: (v3: V3, provider?: string) => ai.Tool<{
|
|
|
3978
4028
|
y: number;
|
|
3979
4029
|
};
|
|
3980
4030
|
}[];
|
|
3981
|
-
},
|
|
3982
|
-
success: boolean;
|
|
3983
|
-
playwrightArguments: {
|
|
3984
|
-
coordinates: {
|
|
3985
|
-
x: number;
|
|
3986
|
-
y: number;
|
|
3987
|
-
};
|
|
3988
|
-
action: string;
|
|
3989
|
-
value: string;
|
|
3990
|
-
}[];
|
|
3991
|
-
error?: undefined;
|
|
3992
|
-
} | {
|
|
3993
|
-
success: boolean;
|
|
3994
|
-
error: string;
|
|
3995
|
-
playwrightArguments?: undefined;
|
|
3996
|
-
}>;
|
|
4031
|
+
}, FillFormVisionToolResult>;
|
|
3997
4032
|
|
|
3998
4033
|
declare const thinkTool: () => ai.Tool<{
|
|
3999
4034
|
reasoning: string;
|
|
@@ -4047,7 +4082,6 @@ type AgentToolTypesMap = {
|
|
|
4047
4082
|
ariaTree: ReturnType<typeof ariaTreeTool>;
|
|
4048
4083
|
click: ReturnType<typeof clickTool>;
|
|
4049
4084
|
clickAndHold: ReturnType<typeof clickAndHoldTool>;
|
|
4050
|
-
close: ReturnType<typeof closeTool>;
|
|
4051
4085
|
dragAndDrop: ReturnType<typeof dragAndDropTool>;
|
|
4052
4086
|
extract: ReturnType<typeof extractTool>;
|
|
4053
4087
|
fillForm: ReturnType<typeof fillFormTool>;
|
|
@@ -4181,4 +4215,4 @@ declare class V3Evaluator {
|
|
|
4181
4215
|
private _evaluateWithMultipleScreenshots;
|
|
4182
4216
|
}
|
|
4183
4217
|
|
|
4184
|
-
export { type AISDKCustomProvider, type AISDKProvider, AISdkClient, AVAILABLE_CUA_MODELS, type ActOptions, type ActResult, ActTimeoutError, type Action, type ActionExecutionResult, AgentAbortError, type AgentAction, type AgentCallbacks, type AgentConfig, type AgentContext, type AgentExecuteCallbacks, type AgentExecuteOptions, type AgentExecuteOptionsBase, type AgentExecutionOptions, type AgentHandlerOptions, type AgentInstance, type AgentModelConfig, AgentProvider, type AgentProviderType, type AgentResult, AgentScreenshotProviderError, type AgentState, type AgentStreamCallbacks, type AgentStreamExecuteOptions, type AgentStreamResult, type AgentToolCall, type AgentToolMode, type AgentToolResult, type AgentToolTypesMap, type AgentTools, type AgentType, type AgentUITools, AnnotatedScreenshotText, type AnthropicClientOptions, type AnthropicContentBlock, type AnthropicJsonSchemaObject, type AnthropicMessage, type AnthropicTextBlock, type AnthropicToolResult, type AnyPage, api as Api, type AvailableCuaModel, type AvailableModel, BrowserbaseSessionNotFoundError, CaptchaTimeoutError, type ChatCompletionOptions, type ChatMessage, type ChatMessageContent, type ChatMessageImageContent, type ChatMessageTextContent, type ClientOptions, type ComputerCallItem, ConnectionTimeoutError, type ConsoleListener, ConsoleMessage, ContentFrameNotFoundError, type CreateChatCompletionOptions, CreateChatCompletionResponseError, CuaModelRequiredError, ElementNotVisibleError, ExperimentalApiConflictError, ExperimentalNotConfiguredError, type ExtractOptions, type ExtractResult, ExtractTimeoutError, type FunctionCallItem, type GoogleServiceAccountCredentials, type GoogleVertexProviderSettings, HandlerNotInitializedError, type HistoryEntry, type InferStagehandSchema, InvalidAISDKModelFormatError, type JsonSchema, type JsonSchemaDocument, type JsonSchemaProperty, LLMClient, type LLMParsedResponse, type LLMResponse, LLMResponseError, type LLMTool, type LLMUsage, LOG_LEVEL_NAMES, type LoadState, type LocalBrowserLaunchOptions, type LogLevel, type LogLine, type Logger, MCPConnectionError, MissingEnvironmentVariableError, MissingLLMConfigurationError, type ModelConfiguration, type ModelProvider, type NonStreamingAgentInstance, type ObserveOptions, ObserveTimeoutError, type OpenAIClientOptions, Page, PageNotFoundError, Response$1 as Response, ResponseBodyError, type ResponseInputItem, type ResponseItem, ResponseParseError, type SafetyCheck, type SafetyConfirmationHandler, type SafetyConfirmationResponse, V3 as Stagehand, StagehandAPIError, StagehandAPIUnauthorizedError, StagehandClickError, StagehandClosedError, StagehandDefaultError, StagehandDomProcessError, StagehandElementNotFoundError, StagehandEnvironmentError, StagehandError, StagehandEvalError, StagehandHttpError, StagehandIframeError, StagehandInitError, StagehandInvalidArgumentError, type StagehandMetrics, StagehandMissingArgumentError, StagehandNotInitializedError, StagehandResponseBodyError, StagehandResponseParseError, StagehandServerError, StagehandShadowRootMissingError, StagehandShadowSegmentEmptyError, StagehandShadowSegmentNotFoundError, type StagehandZodObject, type StagehandZodSchema, type StreamingAgentInstance, StreamingCallbacksInNonStreamingModeError, TimeoutError, type ToolUseItem, UnsupportedAISDKModelProviderError, UnsupportedModelError, UnsupportedModelProviderError, V3, type V3Env, V3Evaluator, V3FunctionName, type V3Options, XPathResolutionError, ZodSchemaValidationError, connectToMCPServer, defaultExtractSchema, getZodType, injectUrls, isRunningInBun, isZod3Schema, isZod4Schema, jsonSchemaToZod, loadApiKeyFromEnv, localBrowserLaunchOptionsSchema, modelToAgentProviderMap, pageTextSchema, providerEnvVarMap, toGeminiSchema, toJsonSchema, transformSchema, trimTrailingTextNode, validateZodSchema };
|
|
4218
|
+
export { type AISDKCustomProvider, type AISDKProvider, AISdkClient, AVAILABLE_CUA_MODELS, type ActOptions, type ActResult, ActTimeoutError, type Action, type ActionExecutionResult, AgentAbortError, type AgentAction, type AgentCallbacks, type AgentConfig, type AgentContext, type AgentExecuteCallbacks, type AgentExecuteOptions, type AgentExecuteOptionsBase, type AgentExecutionOptions, type AgentHandlerOptions, type AgentInstance, type AgentModelConfig, AgentProvider, type AgentProviderType, type AgentResult, AgentScreenshotProviderError, type AgentState, type AgentStreamCallbacks, type AgentStreamExecuteOptions, type AgentStreamResult, type AgentToolCall, type AgentToolMode, type AgentToolResult, type AgentToolTypesMap, type AgentTools, type AgentType, type AgentUITools, AnnotatedScreenshotText, type AnthropicClientOptions, type AnthropicContentBlock, type AnthropicJsonSchemaObject, type AnthropicMessage, type AnthropicTextBlock, type AnthropicToolResult, type AnyPage, api as Api, type AvailableCuaModel, type AvailableModel, BrowserbaseSessionNotFoundError, CaptchaTimeoutError, type ChatCompletionOptions, type ChatMessage, type ChatMessageContent, type ChatMessageImageContent, type ChatMessageTextContent, type ClickToolResult, type ClientOptions, type ComputerCallItem, ConnectionTimeoutError, type ConsoleListener, ConsoleMessage, ContentFrameNotFoundError, type CreateChatCompletionOptions, CreateChatCompletionResponseError, CuaModelRequiredError, type DragAndDropToolResult, ElementNotVisibleError, ExperimentalApiConflictError, ExperimentalNotConfiguredError, type ExtractOptions, type ExtractResult, ExtractTimeoutError, type FillFormField, type FillFormVisionToolResult, type FunctionCallItem, type GoogleServiceAccountCredentials, type GoogleVertexProviderSettings, HandlerNotInitializedError, type HistoryEntry, type InferStagehandSchema, InvalidAISDKModelFormatError, type JsonSchema, type JsonSchemaDocument, type JsonSchemaProperty, LLMClient, type LLMParsedResponse, type LLMResponse, LLMResponseError, type LLMTool, type LLMUsage, LOG_LEVEL_NAMES, type LoadState, type LocalBrowserLaunchOptions, type LogLevel, type LogLine, type Logger, MCPConnectionError, MissingEnvironmentVariableError, MissingLLMConfigurationError, type ModelConfiguration, type ModelOutputContentItem, type ModelProvider, type NonStreamingAgentInstance, type ObserveOptions, ObserveTimeoutError, type OpenAIClientOptions, Page, PageNotFoundError, Response$1 as Response, ResponseBodyError, type ResponseInputItem, type ResponseItem, ResponseParseError, type SafetyCheck, type SafetyConfirmationHandler, type SafetyConfirmationResponse, type ScrollVisionToolResult, V3 as Stagehand, StagehandAPIError, StagehandAPIUnauthorizedError, StagehandClickError, StagehandClosedError, StagehandDefaultError, StagehandDomProcessError, StagehandElementNotFoundError, StagehandEnvironmentError, StagehandError, StagehandEvalError, StagehandHttpError, StagehandIframeError, StagehandInitError, StagehandInvalidArgumentError, type StagehandMetrics, StagehandMissingArgumentError, StagehandNotInitializedError, StagehandResponseBodyError, StagehandResponseParseError, StagehandServerError, StagehandShadowRootMissingError, StagehandShadowSegmentEmptyError, StagehandShadowSegmentNotFoundError, type StagehandZodObject, type StagehandZodSchema, type StreamingAgentInstance, StreamingCallbacksInNonStreamingModeError, TimeoutError, type ToolUseItem, type TypeToolResult, UnsupportedAISDKModelProviderError, UnsupportedModelError, UnsupportedModelProviderError, V3, type V3Env, V3Evaluator, V3FunctionName, type V3Options, type WaitToolResult, XPathResolutionError, ZodSchemaValidationError, connectToMCPServer, defaultExtractSchema, getZodType, injectUrls, isRunningInBun, isZod3Schema, isZod4Schema, jsonSchemaToZod, loadApiKeyFromEnv, localBrowserLaunchOptionsSchema, modelToAgentProviderMap, pageTextSchema, providerEnvVarMap, toGeminiSchema, toJsonSchema, transformSchema, trimTrailingTextNode, validateZodSchema };
|