@yourgpt/llm-sdk 1.5.1 → 1.5.3

package/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2024 YourGPT
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

package/dist/index.d.mts

@@ -413,6 +413,29 @@ interface HandleRequestOptions {
      */
     onFinish?: (result: HandleRequestResult) => Promise<void> | void;
 }
+/**
+ * Options for runtime.generate()
+ *
+ * @example
+ * ```typescript
+ * const result = await runtime.generate(body, {
+ *   onFinish: async ({ messages }) => {
+ *     await db.saveMessages(messages);
+ *   },
+ * });
+ * ```
+ */
+interface GenerateOptions {
+    /** AbortSignal for cancellation */
+    signal?: AbortSignal;
+    /** HTTP request for extracting headers (auth context) */
+    httpRequest?: Request;
+    /**
+     * Called after generation completes successfully.
+     * Use for server-side persistence.
+     */
+    onFinish?: (result: HandleRequestResult) => Promise<void> | void;
+}
 
 /**
  * StreamResult - Industry-standard streaming result object
@@ -447,6 +470,13 @@ interface HandleRequestOptions {
 interface StreamResultOptions {
     /** Additional headers to include in response */
     headers?: Record<string, string>;
+    /**
+     * Include token usage in response (default: false)
+     * Set to true for raw API access where you need usage data.
+     * When false, usage is stripped from client-facing responses
+     * but still available in onFinish callback for billing.
+     */
+    includeUsage?: boolean;
 }
 /**
  * Result passed to onFinish callback
@@ -598,11 +628,14 @@ declare class StreamResult {
      *
      * @example
      * ```typescript
+     * // Default: usage stripped for client-facing responses
      * const { text, messages, toolCalls } = await runtime.stream(body).collect();
-     * console.log('Response:', text);
+     *
+     * // Raw: include usage for server-side processing
+     * const { text, usage } = await runtime.stream(body).collect({ includeUsage: true });
      * ```
      */
-    collect(): Promise<CollectedResult>;
+    collect(options?: StreamResultOptions): Promise<CollectedResult>;
     /**
      * Get final text (convenience method)
      *
@@ -650,6 +683,11 @@ declare class StreamResult {
      * Call registered event handlers
      */
     private callEventHandlers;
+    /**
+     * Strip usage from result (usage is server-side only for billing)
+     * Client-facing APIs should not expose token usage
+     */
+    private stripUsageFromResult;
 }
 /**
  * Create a StreamResult from an async generator
@@ -662,6 +700,118 @@ declare class StreamResult {
  */
 declare function createStreamResult(generator: AsyncGenerator<StreamEvent>): StreamResult;
 
+/**
+ * GenerateResult - Result from non-streaming generation
+ *
+ * Similar to Vercel AI SDK's generateText() result.
+ * Provides both raw access and formatted response methods.
+ *
+ * @example
+ * ```typescript
+ * const result = await runtime.generate(body);
+ *
+ * // Raw access
+ * console.log(result.text);
+ * console.log(result.toolCalls);
+ *
+ * // CopilotChat format
+ * res.json(result.toResponse());
+ * ```
+ */
+
+interface GenerateResultData {
+    text: string;
+    messages: DoneEventMessage[];
+    toolCalls: Array<{
+        id: string;
+        name: string;
+        args: Record<string, unknown>;
+    }>;
+    toolResults: Array<{
+        id: string;
+        result: unknown;
+    }>;
+    requiresAction: boolean;
+    error?: {
+        message: string;
+        code?: string;
+    };
+}
+/**
+ * Response format compatible with CopilotChat
+ */
+interface CopilotChatResponse {
+    success: boolean;
+    content: string;
+    messages?: DoneEventMessage[];
+    toolCalls?: Array<{
+        id: string;
+        name: string;
+        args: Record<string, unknown>;
+    }>;
+    toolResults?: Array<{
+        id: string;
+        result: unknown;
+    }>;
+    requiresAction?: boolean;
+    error?: {
+        message: string;
+        code?: string;
+    };
+}
+/**
+ * GenerateResult - Non-streaming generation result
+ */
+declare class GenerateResult {
+    private data;
+    constructor(data: GenerateResultData);
+    /**
+     * Generated text content
+     */
+    get text(): string;
+    /**
+     * All messages from the conversation
+     */
+    get messages(): DoneEventMessage[];
+    /**
+     * Tool calls made during generation
+     */
+    get toolCalls(): GenerateResultData["toolCalls"];
+    /**
+     * Results from tool executions
+     */
+    get toolResults(): GenerateResultData["toolResults"];
+    /**
+     * Whether client action is required (e.g., tool approval)
+     */
+    get requiresAction(): boolean;
+    /**
+     * Error if generation failed
+     */
+    get error(): GenerateResultData["error"];
+    /**
+     * Whether generation was successful
+     */
+    get success(): boolean;
+    /**
+     * Convert to CopilotChat-compatible JSON response
+     *
+     * @example
+     * ```typescript
+     * // Express
+     * res.json(result.toResponse());
+     *
+     * // Next.js
+     * return Response.json(result.toResponse());
+     * ```
+     */
+    toResponse(): CopilotChatResponse;
+    /**
+     * Convert to raw object (without methods)
+     */
+    toJSON(): GenerateResultData;
+}
+
 /**
  * Copilot SDK Runtime
  *
@@ -845,23 +995,42 @@ declare class Runtime {
      * console.log('Response:', text);
      * res.json({ response: text });
      *
-     * // With onFinish for usage tracking
-     * const result = await runtime.chat(body, {
-     *   onFinish: ({ usage }) => console.log('Tokens:', usage?.totalTokens)
-     * });
+     * // Usage is included in result - strip before sending to client
+     * const { usage, ...clientResult } = await runtime.chat(body);
+     * await billing.record(usage);
+     * res.json(clientResult);
      * ```
      */
     chat(request: ChatRequest, options?: {
         signal?: AbortSignal;
-        onFinish?: (result: {
-            messages: DoneEventMessage[];
-            usage?: {
-                promptTokens: number;
-                completionTokens: number;
-                totalTokens: number;
-            };
-        }) => Promise<void> | void;
     }): Promise<CollectedResult>;
+    /**
+     * Generate a complete response (non-streaming)
+     *
+     * Like Vercel AI SDK's generateText() - clean, non-streaming API.
+     * Returns GenerateResult with .toResponse() for CopilotChat format.
+     *
+     * @example
+     * ```typescript
+     * // Simple usage
+     * const result = await runtime.generate(body);
+     * console.log(result.text);
+     *
+     * // CopilotChat format response (Express)
+     * res.json(result.toResponse());
+     *
+     * // CopilotChat format response (Next.js)
+     * return Response.json(result.toResponse());
+     *
+     * // With persistence callback
+     * const result = await runtime.generate(body, {
+     *   onFinish: async ({ messages }) => {
+     *     await db.saveMessages(messages);
+     *   },
+     * });
+     * ```
+     */
+    generate(request: ChatRequest, options?: GenerateOptions): Promise<GenerateResult>;
     /**
      * Create Express-compatible handler middleware
      *
@@ -1164,4 +1333,4 @@ declare function generateThreadId(): string;
  */
 declare function generateToolCallId(): string;
 
-export { AIProvider, ActionDefinition, type ActionRequest, AgentLoopConfig, type AgentLoopOptions, type ChatRequest, type CollectedResult, DEFAULT_MAX_ITERATIONS, DoneEventMessage, GenerateTextParams, GenerateTextResult, LLMAdapter, Message, type RequestContext, Runtime, type RuntimeConfig, StreamEvent, StreamResult, type StreamResultOptions, StreamTextParams, StreamTextResult, TokenUsageRaw, Tool, ToolCallInfo, ToolContext, ToolDefinition, ToolResponse, createEventStream, createExpressHandler, createExpressMiddleware, createHonoApp, createNextHandler, createNodeHandler, createRuntime, createSSEHeaders, createSSEResponse, createStreamResult, createTextStreamHeaders, createTextStreamResponse, formatSSEData, formatToolsForAnthropic, formatToolsForGoogle, formatToolsForOpenAI, generateMessageId, generateText, generateThreadId, generateToolCallId, pipeSSEToResponse, pipeTextToResponse, runAgentLoop, streamText, tool };
+export { AIProvider, ActionDefinition, type ActionRequest, AgentLoopConfig, type AgentLoopOptions, type ChatRequest, type CollectedResult, type CopilotChatResponse, DEFAULT_MAX_ITERATIONS, DoneEventMessage, GenerateResult, type GenerateResultData, GenerateTextParams, GenerateTextResult, LLMAdapter, Message, type RequestContext, Runtime, type RuntimeConfig, StreamEvent, StreamResult, type StreamResultOptions, StreamTextParams, StreamTextResult, TokenUsageRaw, Tool, ToolCallInfo, ToolContext, ToolDefinition, ToolResponse, createEventStream, createExpressHandler, createExpressMiddleware, createHonoApp, createNextHandler, createNodeHandler, createRuntime, createSSEHeaders, createSSEResponse, createStreamResult, createTextStreamHeaders, createTextStreamResponse, formatSSEData, formatToolsForAnthropic, formatToolsForGoogle, formatToolsForOpenAI, generateMessageId, generateText, generateThreadId, generateToolCallId, pipeSSEToResponse, pipeTextToResponse, runAgentLoop, streamText, tool };

package/dist/index.d.ts

@@ -413,6 +413,29 @@ interface HandleRequestOptions {
      */
     onFinish?: (result: HandleRequestResult) => Promise<void> | void;
 }
+/**
+ * Options for runtime.generate()
+ *
+ * @example
+ * ```typescript
+ * const result = await runtime.generate(body, {
+ *   onFinish: async ({ messages }) => {
+ *     await db.saveMessages(messages);
+ *   },
+ * });
+ * ```
+ */
+interface GenerateOptions {
+    /** AbortSignal for cancellation */
+    signal?: AbortSignal;
+    /** HTTP request for extracting headers (auth context) */
+    httpRequest?: Request;
+    /**
+     * Called after generation completes successfully.
+     * Use for server-side persistence.
+     */
+    onFinish?: (result: HandleRequestResult) => Promise<void> | void;
+}
 
 /**
  * StreamResult - Industry-standard streaming result object
@@ -447,6 +470,13 @@ interface HandleRequestOptions {
 interface StreamResultOptions {
     /** Additional headers to include in response */
     headers?: Record<string, string>;
+    /**
+     * Include token usage in response (default: false)
+     * Set to true for raw API access where you need usage data.
+     * When false, usage is stripped from client-facing responses
+     * but still available in onFinish callback for billing.
+     */
+    includeUsage?: boolean;
 }
 /**
  * Result passed to onFinish callback
@@ -598,11 +628,14 @@ declare class StreamResult {
      *
      * @example
      * ```typescript
+     * // Default: usage stripped for client-facing responses
      * const { text, messages, toolCalls } = await runtime.stream(body).collect();
-     * console.log('Response:', text);
+     *
+     * // Raw: include usage for server-side processing
+     * const { text, usage } = await runtime.stream(body).collect({ includeUsage: true });
      * ```
      */
-    collect(): Promise<CollectedResult>;
+    collect(options?: StreamResultOptions): Promise<CollectedResult>;
     /**
      * Get final text (convenience method)
      *
@@ -650,6 +683,11 @@ declare class StreamResult {
      * Call registered event handlers
      */
     private callEventHandlers;
+    /**
+     * Strip usage from result (usage is server-side only for billing)
+     * Client-facing APIs should not expose token usage
+     */
+    private stripUsageFromResult;
 }
 /**
  * Create a StreamResult from an async generator
@@ -662,6 +700,118 @@ declare class StreamResult {
  */
 declare function createStreamResult(generator: AsyncGenerator<StreamEvent>): StreamResult;
 
+/**
+ * GenerateResult - Result from non-streaming generation
+ *
+ * Similar to Vercel AI SDK's generateText() result.
+ * Provides both raw access and formatted response methods.
+ *
+ * @example
+ * ```typescript
+ * const result = await runtime.generate(body);
+ *
+ * // Raw access
+ * console.log(result.text);
+ * console.log(result.toolCalls);
+ *
+ * // CopilotChat format
+ * res.json(result.toResponse());
+ * ```
+ */
+
+interface GenerateResultData {
+    text: string;
+    messages: DoneEventMessage[];
+    toolCalls: Array<{
+        id: string;
+        name: string;
+        args: Record<string, unknown>;
+    }>;
+    toolResults: Array<{
+        id: string;
+        result: unknown;
+    }>;
+    requiresAction: boolean;
+    error?: {
+        message: string;
+        code?: string;
+    };
+}
+/**
+ * Response format compatible with CopilotChat
+ */
+interface CopilotChatResponse {
+    success: boolean;
+    content: string;
+    messages?: DoneEventMessage[];
+    toolCalls?: Array<{
+        id: string;
+        name: string;
+        args: Record<string, unknown>;
+    }>;
+    toolResults?: Array<{
+        id: string;
+        result: unknown;
+    }>;
+    requiresAction?: boolean;
+    error?: {
+        message: string;
+        code?: string;
+    };
+}
+/**
+ * GenerateResult - Non-streaming generation result
+ */
+declare class GenerateResult {
+    private data;
+    constructor(data: GenerateResultData);
+    /**
+     * Generated text content
+     */
+    get text(): string;
+    /**
+     * All messages from the conversation
+     */
+    get messages(): DoneEventMessage[];
+    /**
+     * Tool calls made during generation
+     */
+    get toolCalls(): GenerateResultData["toolCalls"];
+    /**
+     * Results from tool executions
+     */
+    get toolResults(): GenerateResultData["toolResults"];
+    /**
+     * Whether client action is required (e.g., tool approval)
+     */
+    get requiresAction(): boolean;
+    /**
+     * Error if generation failed
+     */
+    get error(): GenerateResultData["error"];
+    /**
+     * Whether generation was successful
+     */
+    get success(): boolean;
+    /**
+     * Convert to CopilotChat-compatible JSON response
+     *
+     * @example
+     * ```typescript
+     * // Express
+     * res.json(result.toResponse());
+     *
+     * // Next.js
+     * return Response.json(result.toResponse());
+     * ```
+     */
+    toResponse(): CopilotChatResponse;
+    /**
+     * Convert to raw object (without methods)
+     */
+    toJSON(): GenerateResultData;
+}
+
 /**
  * Copilot SDK Runtime
  *
@@ -845,23 +995,42 @@ declare class Runtime {
      * console.log('Response:', text);
      * res.json({ response: text });
      *
-     * // With onFinish for usage tracking
-     * const result = await runtime.chat(body, {
-     *   onFinish: ({ usage }) => console.log('Tokens:', usage?.totalTokens)
-     * });
+     * // Usage is included in result - strip before sending to client
+     * const { usage, ...clientResult } = await runtime.chat(body);
+     * await billing.record(usage);
+     * res.json(clientResult);
      * ```
      */
     chat(request: ChatRequest, options?: {
         signal?: AbortSignal;
-        onFinish?: (result: {
-            messages: DoneEventMessage[];
-            usage?: {
-                promptTokens: number;
-                completionTokens: number;
-                totalTokens: number;
-            };
-        }) => Promise<void> | void;
     }): Promise<CollectedResult>;
+    /**
+     * Generate a complete response (non-streaming)
+     *
+     * Like Vercel AI SDK's generateText() - clean, non-streaming API.
+     * Returns GenerateResult with .toResponse() for CopilotChat format.
+     *
+     * @example
+     * ```typescript
+     * // Simple usage
+     * const result = await runtime.generate(body);
+     * console.log(result.text);
+     *
+     * // CopilotChat format response (Express)
+     * res.json(result.toResponse());
+     *
+     * // CopilotChat format response (Next.js)
+     * return Response.json(result.toResponse());
+     *
+     * // With persistence callback
+     * const result = await runtime.generate(body, {
+     *   onFinish: async ({ messages }) => {
+     *     await db.saveMessages(messages);
+     *   },
+     * });
+     * ```
+     */
+    generate(request: ChatRequest, options?: GenerateOptions): Promise<GenerateResult>;
     /**
      * Create Express-compatible handler middleware
      *
@@ -1164,4 +1333,4 @@ declare function generateThreadId(): string;
  */
 declare function generateToolCallId(): string;
 
-export { AIProvider, ActionDefinition, type ActionRequest, AgentLoopConfig, type AgentLoopOptions, type ChatRequest, type CollectedResult, DEFAULT_MAX_ITERATIONS, DoneEventMessage, GenerateTextParams, GenerateTextResult, LLMAdapter, Message, type RequestContext, Runtime, type RuntimeConfig, StreamEvent, StreamResult, type StreamResultOptions, StreamTextParams, StreamTextResult, TokenUsageRaw, Tool, ToolCallInfo, ToolContext, ToolDefinition, ToolResponse, createEventStream, createExpressHandler, createExpressMiddleware, createHonoApp, createNextHandler, createNodeHandler, createRuntime, createSSEHeaders, createSSEResponse, createStreamResult, createTextStreamHeaders, createTextStreamResponse, formatSSEData, formatToolsForAnthropic, formatToolsForGoogle, formatToolsForOpenAI, generateMessageId, generateText, generateThreadId, generateToolCallId, pipeSSEToResponse, pipeTextToResponse, runAgentLoop, streamText, tool };
+export { AIProvider, ActionDefinition, type ActionRequest, AgentLoopConfig, type AgentLoopOptions, type ChatRequest, type CollectedResult, type CopilotChatResponse, DEFAULT_MAX_ITERATIONS, DoneEventMessage, GenerateResult, type GenerateResultData, GenerateTextParams, GenerateTextResult, LLMAdapter, Message, type RequestContext, Runtime, type RuntimeConfig, StreamEvent, StreamResult, type StreamResultOptions, StreamTextParams, StreamTextResult, TokenUsageRaw, Tool, ToolCallInfo, ToolContext, ToolDefinition, ToolResponse, createEventStream, createExpressHandler, createExpressMiddleware, createHonoApp, createNextHandler, createNodeHandler, createRuntime, createSSEHeaders, createSSEResponse, createStreamResult, createTextStreamHeaders, createTextStreamResponse, formatSSEData, formatToolsForAnthropic, formatToolsForGoogle, formatToolsForOpenAI, generateMessageId, generateText, generateThreadId, generateToolCallId, pipeSSEToResponse, pipeTextToResponse, runAgentLoop, streamText, tool };