@dealcrawl/sdk 2.3.0 → 2.4.0

package/dist/index.d.mts

@@ -839,6 +839,19 @@ interface AgentStatusResponse extends JobStatusResponse {
     /** Final result when completed */
     result?: AgentResultResponse;
 }
+/** Schema generation response from /v1/agent/schema */
+interface SchemaGenerationResponse {
+    /** Generated JSON Schema for data extraction */
+    schema: Record<string, unknown>;
+    /** Refined prompt optimized for agent execution */
+    refinedPrompt: string;
+    /** Human-readable description of the schema */
+    schemaDescription: string;
+    /** Suggested follow-up questions if prompt was ambiguous */
+    suggestedQuestions?: string[];
+    /** Confidence score (0-1) in the generated schema */
+    confidence: number;
+}
 
 /**
  * Polling Utilities
@@ -1418,6 +1431,26 @@ interface AgentOptions {
         headers?: Record<string, string>;
     };
 }
+/** Context for schema generation from conversation */
+interface SchemaGenerationContext {
+    /** Specific domains/topics mentioned (e.g., ['marketing', 'web development']) */
+    domains?: string[];
+    /** Types of data to extract (e.g., ['free courses', 'discounts']) */
+    dataTypes?: string[];
+    /** Preferred output format */
+    format?: "json" | "csv" | "table";
+    /** Additional clarifications from user */
+    clarifications?: string[];
+}
+/** Options for generating a JSON Schema from natural language */
+interface SchemaGenerationOptions {
+    /** Natural language description of what data to extract (required, 5-2000 chars) */
+    prompt: string;
+    /** Optional context from conversation to refine the schema */
+    context?: SchemaGenerationContext;
+    /** LLM provider for generation (default: openai) */
+    model?: AgentModel;
+}
 
 /**
  * Account Resource
@@ -1713,6 +1746,53 @@ declare class AgentResource {
      * ```
      */
     withClaude(url: string, prompt: string, options?: Omit<AgentOptions, "url" | "prompt" | "model">): Promise<AgentJobResponse>;
+    /**
+     * Generate a JSON Schema from a natural language prompt
+     *
+     * This is useful for building extraction schemas without manual JSON writing.
+     * The generated schema can be used with the main agent.create() method.
+     *
+     * @param options - Schema generation options
+     * @returns Generated schema with refined prompt and confidence score
+     *
+     * @example Basic usage:
+     * ```ts
+     * const result = await client.agent.generateSchema({
+     *   prompt: "Find the best student deals on Coursera for marketing courses"
+     * });
+     *
+     * console.log(result.schema);
+     * // { type: "object", properties: { courses: { ... } } }
+     *
+     * console.log(result.refinedPrompt);
+     * // "Extract student offers for marketing courses..."
+     *
+     * // Use the generated schema with an agent
+     * const job = await client.agent.create({
+     *   url: "https://coursera.org",
+     *   prompt: result.refinedPrompt,
+     *   schema: result.schema
+     * });
+     * ```
+     *
+     * @example With context from conversation:
+     * ```ts
+     * const result = await client.agent.generateSchema({
+     *   prompt: "Find student deals on online courses",
+     *   context: {
+     *     domains: ["marketing", "web development"],
+     *     dataTypes: ["free courses", "discounts"],
+     *     format: "json"
+     *   }
+     * });
+     *
+     * if (result.confidence < 0.7 && result.suggestedQuestions) {
+     *   // Ask user for clarification
+     *   console.log("Please clarify:", result.suggestedQuestions);
+     * }
+     * ```
+     */
+    generateSchema(options: SchemaGenerationOptions): Promise<SchemaGenerationResponse>;
 }
 
 /**

package/dist/index.d.ts

@@ -839,6 +839,19 @@ interface AgentStatusResponse extends JobStatusResponse {
     /** Final result when completed */
     result?: AgentResultResponse;
 }
+/** Schema generation response from /v1/agent/schema */
+interface SchemaGenerationResponse {
+    /** Generated JSON Schema for data extraction */
+    schema: Record<string, unknown>;
+    /** Refined prompt optimized for agent execution */
+    refinedPrompt: string;
+    /** Human-readable description of the schema */
+    schemaDescription: string;
+    /** Suggested follow-up questions if prompt was ambiguous */
+    suggestedQuestions?: string[];
+    /** Confidence score (0-1) in the generated schema */
+    confidence: number;
+}
 
 /**
  * Polling Utilities
@@ -1418,6 +1431,26 @@ interface AgentOptions {
         headers?: Record<string, string>;
     };
 }
+/** Context for schema generation from conversation */
+interface SchemaGenerationContext {
+    /** Specific domains/topics mentioned (e.g., ['marketing', 'web development']) */
+    domains?: string[];
+    /** Types of data to extract (e.g., ['free courses', 'discounts']) */
+    dataTypes?: string[];
+    /** Preferred output format */
+    format?: "json" | "csv" | "table";
+    /** Additional clarifications from user */
+    clarifications?: string[];
+}
+/** Options for generating a JSON Schema from natural language */
+interface SchemaGenerationOptions {
+    /** Natural language description of what data to extract (required, 5-2000 chars) */
+    prompt: string;
+    /** Optional context from conversation to refine the schema */
+    context?: SchemaGenerationContext;
+    /** LLM provider for generation (default: openai) */
+    model?: AgentModel;
+}
 
 /**
  * Account Resource
@@ -1713,6 +1746,53 @@ declare class AgentResource {
      * ```
      */
     withClaude(url: string, prompt: string, options?: Omit<AgentOptions, "url" | "prompt" | "model">): Promise<AgentJobResponse>;
+    /**
+     * Generate a JSON Schema from a natural language prompt
+     *
+     * This is useful for building extraction schemas without manual JSON writing.
+     * The generated schema can be used with the main agent.create() method.
+     *
+     * @param options - Schema generation options
+     * @returns Generated schema with refined prompt and confidence score
+     *
+     * @example Basic usage:
+     * ```ts
+     * const result = await client.agent.generateSchema({
+     *   prompt: "Find the best student deals on Coursera for marketing courses"
+     * });
+     *
+     * console.log(result.schema);
+     * // { type: "object", properties: { courses: { ... } } }
+     *
+     * console.log(result.refinedPrompt);
+     * // "Extract student offers for marketing courses..."
+     *
+     * // Use the generated schema with an agent
+     * const job = await client.agent.create({
+     *   url: "https://coursera.org",
+     *   prompt: result.refinedPrompt,
+     *   schema: result.schema
+     * });
+     * ```
+     *
+     * @example With context from conversation:
+     * ```ts
+     * const result = await client.agent.generateSchema({
+     *   prompt: "Find student deals on online courses",
+     *   context: {
+     *     domains: ["marketing", "web development"],
+     *     dataTypes: ["free courses", "discounts"],
+     *     format: "json"
+     *   }
+     * });
+     *
+     * if (result.confidence < 0.7 && result.suggestedQuestions) {
+     *   // Ask user for clarification
+     *   console.log("Please clarify:", result.suggestedQuestions);
+     * }
+     * ```
+     */
+    generateSchema(options: SchemaGenerationOptions): Promise<SchemaGenerationResponse>;
 }
 
 /**

package/dist/index.js

@@ -784,6 +784,65 @@ var AgentResource = class {
       ...options
     });
   }
+  /**
+   * Generate a JSON Schema from a natural language prompt
+   *
+   * This is useful for building extraction schemas without manual JSON writing.
+   * The generated schema can be used with the main agent.create() method.
+   *
+   * @param options - Schema generation options
+   * @returns Generated schema with refined prompt and confidence score
+   *
+   * @example Basic usage:
+   * ```ts
+   * const result = await client.agent.generateSchema({
+   *   prompt: "Find the best student deals on Coursera for marketing courses"
+   * });
+   *
+   * console.log(result.schema);
+   * // { type: "object", properties: { courses: { ... } } }
+   *
+   * console.log(result.refinedPrompt);
+   * // "Extract student offers for marketing courses..."
+   *
+   * // Use the generated schema with an agent
+   * const job = await client.agent.create({
+   *   url: "https://coursera.org",
+   *   prompt: result.refinedPrompt,
+   *   schema: result.schema
+   * });
+   * ```
+   *
+   * @example With context from conversation:
+   * ```ts
+   * const result = await client.agent.generateSchema({
+   *   prompt: "Find student deals on online courses",
+   *   context: {
+   *     domains: ["marketing", "web development"],
+   *     dataTypes: ["free courses", "discounts"],
+   *     format: "json"
+   *   }
+   * });
+   *
+   * if (result.confidence < 0.7 && result.suggestedQuestions) {
+   *   // Ask user for clarification
+   *   console.log("Please clarify:", result.suggestedQuestions);
+   * }
+   * ```
+   */
+  async generateSchema(options) {
+    const body = {
+      prompt: options.prompt,
+      context: options.context,
+      model: options.model ?? "openai"
+    };
+    const result = await post(
+      this.ctx,
+      "/v1/agent/schema",
+      body
+    );
+    return result.data;
+  }
 };
 
 // src/resources/crawl.ts