libretto 0.3.1 → 0.4.0

package/dist/shared/llm/client.d.ts

@@ -1,6 +1,13 @@
 import { LLMClient } from './types.js';
 import 'zod';
 
+type Provider = "google" | "vertex" | "anthropic" | "openai";
+declare function parseModel(model: string): {
+    provider: Provider;
+    modelId: string;
+};
+declare function hasProviderCredentials(provider: Provider, env?: NodeJS.ProcessEnv): boolean;
+declare function missingProviderCredentialsMessage(provider: Provider): string;
 declare function createLLMClient(model: string): LLMClient;
 
-export { createLLMClient };
+export { type Provider, createLLMClient, hasProviderCredentials, missingProviderCredentialsMessage, parseModel };

package/dist/shared/llm/client.js

@@ -1,32 +1,87 @@
-import { createVertex } from "@ai-sdk/google-vertex";
-import { createAnthropic } from "@ai-sdk/anthropic";
-import { createOpenAI } from "@ai-sdk/openai";
 import { generateObject } from "ai";
+const GEMINI_API_KEY_ENV_VARS = [
+  "GEMINI_API_KEY",
+  "GOOGLE_GENERATIVE_AI_API_KEY"
+];
+const VERTEX_PROJECT_ENV_VARS = [
+  "GOOGLE_CLOUD_PROJECT",
+  "GCLOUD_PROJECT"
+];
+const SUPPORTED_PROVIDER_ALIASES = {
+  google: "google",
+  gemini: "google",
+  vertex: "vertex",
+  anthropic: "anthropic",
+  codex: "openai",
+  openai: "openai"
+};
+function readFirstEnvValue(env, names) {
+  for (const name of names) {
+    const value = env[name]?.trim();
+    if (value) return value;
+  }
+  return null;
+}
 function parseModel(model) {
   const slashIndex = model.indexOf("/");
   if (slashIndex === -1) {
     throw new Error(
-      `Invalid model string "${model}". Expected format: "provider/model-id" (e.g. "google/gemini-3-flash-preview").`
+      `Invalid model string "${model}". Expected format: "provider/model-id" (for example "openai/gpt-5.4", "anthropic/claude-sonnet-4-6", "google/gemini-2.5-pro", or "vertex/gemini-2.5-pro").`
     );
   }
-  const provider = model.slice(0, slashIndex);
+  const providerInput = model.slice(0, slashIndex).toLowerCase();
+  const provider = SUPPORTED_PROVIDER_ALIASES[providerInput];
   const modelId = model.slice(slashIndex + 1);
-  if (!["google", "anthropic", "openai"].includes(provider)) {
+  if (!provider) {
     throw new Error(
-      `Unsupported provider "${provider}". Supported providers: google, anthropic, openai.`
+      `Unsupported provider "${providerInput}". Supported providers: openai/codex, anthropic, google (Gemini API), and vertex.`
     );
   }
   return { provider, modelId };
 }
-function getProviderModel(provider, modelId) {
+function hasProviderCredentials(provider, env = process.env) {
+  switch (provider) {
+    case "google":
+      return readFirstEnvValue(env, GEMINI_API_KEY_ENV_VARS) !== null;
+    case "vertex":
+      return readFirstEnvValue(env, VERTEX_PROJECT_ENV_VARS) !== null;
+    case "anthropic":
+      return Boolean(env.ANTHROPIC_API_KEY?.trim());
+    case "openai":
+      return Boolean(env.OPENAI_API_KEY?.trim());
+  }
+}
+function missingProviderCredentialsMessage(provider) {
+  switch (provider) {
+    case "google":
+      return "Missing Gemini API key. Set GEMINI_API_KEY or GOOGLE_GENERATIVE_AI_API_KEY.";
+    case "vertex":
+      return "Missing Vertex AI project. Set GOOGLE_CLOUD_PROJECT (or GCLOUD_PROJECT) and ensure application default credentials are configured.";
+    case "anthropic": {
+      return "Missing Anthropic API key. Set ANTHROPIC_API_KEY.";
+    }
+    case "openai": {
+      return "Missing OpenAI API key. Set OPENAI_API_KEY.";
+    }
+  }
+}
+async function getProviderModel(provider, modelId) {
   switch (provider) {
     case "google": {
-      const project = process.env.GOOGLE_CLOUD_PROJECT || process.env.GCLOUD_PROJECT;
+      const apiKey = readFirstEnvValue(process.env, GEMINI_API_KEY_ENV_VARS);
+      if (!apiKey) {
+        throw new Error(missingProviderCredentialsMessage(provider));
+      }
+      const { createGoogleGenerativeAI } = await import("@ai-sdk/google");
+      const google = createGoogleGenerativeAI({ apiKey });
+      return google(modelId);
+    }
+    case "vertex": {
+      const project = readFirstEnvValue(process.env, VERTEX_PROJECT_ENV_VARS);
       if (!project) {
-        throw new Error(
-          "Missing GCP project for Vertex AI. Set GOOGLE_CLOUD_PROJECT environment variable and ensure application default credentials are configured (gcloud auth application-default login)."
-        );
+        throw new Error(missingProviderCredentialsMessage(provider));
       }
+      const { createVertex } = await import("@ai-sdk/google-vertex");
       const vertex = createVertex({
         project,
         location: process.env.GOOGLE_CLOUD_LOCATION || "global"
@@ -34,22 +89,20 @@ function getProviderModel(provider, modelId) {
       return vertex(modelId);
     }
     case "anthropic": {
-      const apiKey = process.env.ANTHROPIC_API_KEY;
+      const apiKey = process.env.ANTHROPIC_API_KEY?.trim();
       if (!apiKey) {
-        throw new Error(
-          "Missing API key for Anthropic. Set ANTHROPIC_API_KEY environment variable."
-        );
+        throw new Error(missingProviderCredentialsMessage(provider));
       }
+      const { createAnthropic } = await import("@ai-sdk/anthropic");
       const anthropic = createAnthropic({ apiKey });
       return anthropic(modelId);
     }
     case "openai": {
-      const apiKey = process.env.OPENAI_API_KEY;
+      const apiKey = process.env.OPENAI_API_KEY?.trim();
       if (!apiKey) {
-        throw new Error(
-          "Missing API key for OpenAI. Set OPENAI_API_KEY environment variable."
-        );
+        throw new Error(missingProviderCredentialsMessage(provider));
       }
+      const { createOpenAI } = await import("@ai-sdk/openai");
       const openai = createOpenAI({ apiKey });
       return openai(modelId);
     }
@@ -60,7 +113,11 @@ function convertUserContentParts(parts) {
     if (part.type === "text") {
       return { type: "text", text: part.text };
     }
-    return { type: "image", image: part.image };
+    return {
+      type: "image",
+      image: part.image,
+      ...part.mediaType ? { mediaType: part.mediaType } : {}
+    };
   });
 }
 function convertAssistantContentParts(parts) {
@@ -88,9 +145,14 @@ function convertMessages(messages) {
 }
 function createLLMClient(model) {
   const { provider, modelId } = parseModel(model);
-  const aiModel = getProviderModel(provider, modelId);
+  let modelPromise = null;
+  const getModel = () => {
+    modelPromise ??= getProviderModel(provider, modelId);
+    return modelPromise;
+  };
   return {
     async generateObject(opts) {
+      const aiModel = await getModel();
       const result = await generateObject({
         model: aiModel,
         prompt: opts.prompt,
@@ -100,6 +162,7 @@ function createLLMClient(model) {
       return result.object;
     },
     async generateObjectFromMessages(opts) {
+      const aiModel = await getModel();
       const result = await generateObject({
         model: aiModel,
         messages: convertMessages(opts.messages),
@@ -111,5 +174,8 @@ function createLLMClient(model) {
   };
 }
 export {
-  createLLMClient
+  createLLMClient,
+  hasProviderCredentials,
+  missingProviderCredentialsMessage,
+  parseModel
 };

package/dist/shared/llm/types.d.cts

@@ -5,7 +5,8 @@ type MessageContentPart = {
     text: string;
 } | {
     type: "image";
-    image: string;
+    image: string | Uint8Array;
+    mediaType?: string;
 };
 type Message = {
     role: "user" | "assistant";
@@ -42,7 +43,7 @@ interface LLMClient {
         prompt: string;
         schema: T;
         temperature?: number;
-    }): Promise<z.infer<T>>;
+    }): Promise<z.output<T>>;
     /**
      * Generate a structured object from a conversation-style message array.
      *
@@ -60,7 +61,7 @@ interface LLMClient {
         messages: Message[];
         schema: T;
         temperature?: number;
-    }): Promise<z.infer<T>>;
+    }): Promise<z.output<T>>;
 }
 
 export type { LLMClient, Message, MessageContentPart };

package/dist/shared/llm/types.d.ts

@@ -5,7 +5,8 @@ type MessageContentPart = {
     text: string;
 } | {
     type: "image";
-    image: string;
+    image: string | Uint8Array;
+    mediaType?: string;
 };
 type Message = {
     role: "user" | "assistant";
@@ -42,7 +43,7 @@ interface LLMClient {
         prompt: string;
         schema: T;
         temperature?: number;
-    }): Promise<z.infer<T>>;
+    }): Promise<z.output<T>>;
     /**
      * Generate a structured object from a conversation-style message array.
      *
@@ -60,7 +61,7 @@ interface LLMClient {
         messages: Message[];
         schema: T;
         temperature?: number;
-    }): Promise<z.infer<T>>;
+    }): Promise<z.output<T>>;
 }
 
 export type { LLMClient, Message, MessageContentPart };

package/dist/shared/state/session-state.cjs

@@ -21,6 +21,7 @@ __export(session_state_exports, {
   SESSION_STATE_VERSION: () => SESSION_STATE_VERSION,
   SessionStateFileSchema: () => SessionStateFileSchema,
   SessionStatusSchema: () => SessionStatusSchema,
+  SessionViewportSchema: () => SessionViewportSchema,
   parseSessionStateContent: () => parseSessionStateContent,
   parseSessionStateData: () => parseSessionStateData,
   serializeSessionState: () => serializeSessionState
@@ -35,13 +36,18 @@ const SessionStatusSchema = import_zod.z.enum([
   "failed",
   "exited"
 ]);
+const SessionViewportSchema = import_zod.z.object({
+  width: import_zod.z.number().int().min(1),
+  height: import_zod.z.number().int().min(1)
+});
 const SessionStateFileSchema = import_zod.z.object({
   version: import_zod.z.literal(SESSION_STATE_VERSION),
   port: import_zod.z.number().int().min(0).max(65535),
   pid: import_zod.z.number().int(),
   session: import_zod.z.string().min(1),
   startedAt: import_zod.z.string().datetime({ offset: true }),
-  status: SessionStatusSchema.optional()
+  status: SessionStatusSchema.optional(),
+  viewport: SessionViewportSchema.optional()
 });
 function formatIssues(error) {
   return error.issues.map((issue) => {
@@ -79,6 +85,7 @@ function serializeSessionState(state) {
   SESSION_STATE_VERSION,
   SessionStateFileSchema,
   SessionStatusSchema,
+  SessionViewportSchema,
   parseSessionStateContent,
   parseSessionStateData,
   serializeSessionState

package/dist/shared/state/session-state.d.cts

@@ -1,29 +1,35 @@
 import { z } from 'zod';
 
 declare const SESSION_STATE_VERSION = 1;
-declare const SessionStatusSchema: z.ZodEnum<["active", "paused", "completed", "failed", "exited"]>;
+declare const SessionStatusSchema: z.ZodEnum<{
+    active: "active";
+    paused: "paused";
+    completed: "completed";
+    failed: "failed";
+    exited: "exited";
+}>;
+declare const SessionViewportSchema: z.ZodObject<{
+    width: z.ZodNumber;
+    height: z.ZodNumber;
+}, z.core.$strip>;
 declare const SessionStateFileSchema: z.ZodObject<{
     version: z.ZodLiteral<1>;
     port: z.ZodNumber;
     pid: z.ZodNumber;
     session: z.ZodString;
     startedAt: z.ZodString;
-    status: z.ZodOptional<z.ZodEnum<["active", "paused", "completed", "failed", "exited"]>>;
-}, "strip", z.ZodTypeAny, {
-    version: 1;
-    port: number;
-    pid: number;
-    session: string;
-    startedAt: string;
-    status?: "active" | "paused" | "completed" | "failed" | "exited" | undefined;
-}, {
-    version: 1;
-    port: number;
-    pid: number;
-    session: string;
-    startedAt: string;
-    status?: "active" | "paused" | "completed" | "failed" | "exited" | undefined;
-}>;
+    status: z.ZodOptional<z.ZodEnum<{
+        active: "active";
+        paused: "paused";
+        completed: "completed";
+        failed: "failed";
+        exited: "exited";
+    }>>;
+    viewport: z.ZodOptional<z.ZodObject<{
+        width: z.ZodNumber;
+        height: z.ZodNumber;
+    }, z.core.$strip>>;
+}, z.core.$strip>;
 type SessionStatus = z.infer<typeof SessionStatusSchema>;
 type SessionStateFile = z.infer<typeof SessionStateFileSchema>;
 type SessionState = Omit<SessionStateFile, "version">;
@@ -31,4 +37,4 @@ declare function parseSessionStateData(rawState: unknown, source: string): Sessi
 declare function parseSessionStateContent(content: string, source: string): SessionState;
 declare function serializeSessionState(state: SessionState): SessionStateFile;
 
-export { SESSION_STATE_VERSION, type SessionState, type SessionStateFile, SessionStateFileSchema, type SessionStatus, SessionStatusSchema, parseSessionStateContent, parseSessionStateData, serializeSessionState };
+export { SESSION_STATE_VERSION, type SessionState, type SessionStateFile, SessionStateFileSchema, type SessionStatus, SessionStatusSchema, SessionViewportSchema, parseSessionStateContent, parseSessionStateData, serializeSessionState };

package/dist/shared/state/session-state.d.ts

@@ -1,29 +1,35 @@
 import { z } from 'zod';
 
 declare const SESSION_STATE_VERSION = 1;
-declare const SessionStatusSchema: z.ZodEnum<["active", "paused", "completed", "failed", "exited"]>;
+declare const SessionStatusSchema: z.ZodEnum<{
+    active: "active";
+    paused: "paused";
+    completed: "completed";
+    failed: "failed";
+    exited: "exited";
+}>;
+declare const SessionViewportSchema: z.ZodObject<{
+    width: z.ZodNumber;
+    height: z.ZodNumber;
+}, z.core.$strip>;
 declare const SessionStateFileSchema: z.ZodObject<{
     version: z.ZodLiteral<1>;
     port: z.ZodNumber;
     pid: z.ZodNumber;
     session: z.ZodString;
     startedAt: z.ZodString;
-    status: z.ZodOptional<z.ZodEnum<["active", "paused", "completed", "failed", "exited"]>>;
-}, "strip", z.ZodTypeAny, {
-    version: 1;
-    port: number;
-    pid: number;
-    session: string;
-    startedAt: string;
-    status?: "active" | "paused" | "completed" | "failed" | "exited" | undefined;
-}, {
-    version: 1;
-    port: number;
-    pid: number;
-    session: string;
-    startedAt: string;
-    status?: "active" | "paused" | "completed" | "failed" | "exited" | undefined;
-}>;
+    status: z.ZodOptional<z.ZodEnum<{
+        active: "active";
+        paused: "paused";
+        completed: "completed";
+        failed: "failed";
+        exited: "exited";
+    }>>;
+    viewport: z.ZodOptional<z.ZodObject<{
+        width: z.ZodNumber;
+        height: z.ZodNumber;
+    }, z.core.$strip>>;
+}, z.core.$strip>;
 type SessionStatus = z.infer<typeof SessionStatusSchema>;
 type SessionStateFile = z.infer<typeof SessionStateFileSchema>;
 type SessionState = Omit<SessionStateFile, "version">;
@@ -31,4 +37,4 @@ declare function parseSessionStateData(rawState: unknown, source: string): Sessi
 declare function parseSessionStateContent(content: string, source: string): SessionState;
 declare function serializeSessionState(state: SessionState): SessionStateFile;
 
-export { SESSION_STATE_VERSION, type SessionState, type SessionStateFile, SessionStateFileSchema, type SessionStatus, SessionStatusSchema, parseSessionStateContent, parseSessionStateData, serializeSessionState };
+export { SESSION_STATE_VERSION, type SessionState, type SessionStateFile, SessionStateFileSchema, type SessionStatus, SessionStatusSchema, SessionViewportSchema, parseSessionStateContent, parseSessionStateData, serializeSessionState };

package/dist/shared/state/session-state.js

@@ -7,13 +7,18 @@ const SessionStatusSchema = z.enum([
   "failed",
   "exited"
 ]);
+const SessionViewportSchema = z.object({
+  width: z.number().int().min(1),
+  height: z.number().int().min(1)
+});
 const SessionStateFileSchema = z.object({
   version: z.literal(SESSION_STATE_VERSION),
   port: z.number().int().min(0).max(65535),
   pid: z.number().int(),
   session: z.string().min(1),
   startedAt: z.string().datetime({ offset: true }),
-  status: SessionStatusSchema.optional()
+  status: SessionStatusSchema.optional(),
+  viewport: SessionViewportSchema.optional()
 });
 function formatIssues(error) {
   return error.issues.map((issue) => {
@@ -50,6 +55,7 @@ export {
   SESSION_STATE_VERSION,
   SessionStateFileSchema,
   SessionStatusSchema,
+  SessionViewportSchema,
   parseSessionStateContent,
   parseSessionStateData,
   serializeSessionState

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "libretto",
-  "version": "0.3.1",
+  "version": "0.4.0",
   "description": "AI-powered browser automation library and CLI built on Playwright",
   "license": "MIT",
   "repository": {
@@ -8,6 +8,7 @@
     "url": "https://github.com/saffron-health/libretto"
   },
   "type": "module",
+  "packageManager": "pnpm@9.15.4",
   "publishConfig": {
     "access": "public"
   },
@@ -26,16 +27,32 @@
       "require": "./dist/index.cjs"
     }
   },
+  "scripts": {
+    "postinstall": "playwright install chromium",
+    "build": "pnpm run build:runtime && pnpm run build:cli",
+    "build:runtime": "tsup --config tsup.config.ts",
+    "build:cli": "tsup --config tsup.cli.config.ts",
+    "type-check": "tsc --noEmit",
+    "test": "pnpm run build && vitest run",
+    "eval": "pnpm run build && vitest run --config vitest.evals.config.ts",
+    "benchmark": "pnpm run build && tsx benchmarks/run.ts",
+    "test:watch": "vitest",
+    "cli": "node dist/index.js",
+    "prepack": "pnpm run build"
+  },
   "peerDependencies": {
-    "@ai-sdk/anthropic": "^3.0.0",
-    "@ai-sdk/google-vertex": "^4.0.0",
-    "@ai-sdk/openai": "^3.0.0",
-    "zod": ">=3.0.0"
+    "@ai-sdk/anthropic": "^3.0.58",
+    "@ai-sdk/google": "^3.0.51",
+    "@ai-sdk/google-vertex": "^4.0.80",
+    "@ai-sdk/openai": "^3.0.41"
   },
   "peerDependenciesMeta": {
     "@ai-sdk/anthropic": {
       "optional": true
     },
+    "@ai-sdk/google": {
+      "optional": true
+    },
     "@ai-sdk/google-vertex": {
       "optional": true
     },
@@ -44,34 +61,23 @@
     }
   },
   "devDependencies": {
-    "@anthropic-ai/claude-agent-sdk": "^0.2.73",
-    "@ai-sdk/anthropic": "^3.0.53",
-    "@ai-sdk/google-vertex": "^4.0.72",
-    "@ai-sdk/openai": "^3.0.39",
-    "@types/node": "^25.3.3",
-    "@types/yargs": "^17.0.33",
-    "openai": "^6.27.0",
-    "tsup": "^8.0.0",
-    "typescript": "^5.7.0",
-    "vitest": "^4.0.18",
-    "zod": "^3.25.0"
+    "@anthropic-ai/claude-agent-sdk": "^0.2.75",
+    "@ai-sdk/anthropic": "^3.0.58",
+    "@ai-sdk/google": "^3.0.51",
+    "@ai-sdk/google-vertex": "^4.0.80",
+    "@ai-sdk/openai": "^3.0.41",
+    "@types/node": "^25.5.0",
+    "@types/yargs": "^17.0.35",
+    "openai": "^6.29.0",
+    "tsup": "^8.5.1",
+    "typescript": "^5.9.3",
+    "vitest": "^4.1.0"
   },
   "dependencies": {
-    "ai": "^6.0.110",
-    "playwright": "^1.52.0",
-    "tsx": "^4.19.2",
-    "yargs": "^17.7.2"
-  },
-  "scripts": {
-    "postinstall": "npx playwright install chromium",
-    "build": "pnpm run build:runtime && pnpm run build:cli",
-    "build:runtime": "tsup --config tsup.config.ts",
-    "build:cli": "tsup --config tsup.cli.config.ts",
-    "type-check": "tsc --noEmit",
-    "test": "pnpm run build && vitest run",
-    "eval": "pnpm run build && vitest run --config vitest.evals.config.ts",
-    "benchmark": "pnpm run build && tsx benchmarks/run.ts",
-    "test:watch": "vitest",
-    "cli": "node dist/index.js"
+    "ai": "^6.0.116",
+    "playwright": "^1.58.2",
+    "tsx": "^4.21.0",
+    "yargs": "^18.0.0",
+    "zod": "^4.3.6"
   }
-}
+}