@jerome-benoit/sap-ai-provider 3.0.0 → 4.0.0-rc.1

package/README.md

@@ -3,6 +3,7 @@
 [![npm](https://img.shields.io/npm/v/@mymediset/sap-ai-provider/latest?label=npm&color=blue)](https://www.npmjs.com/package/@mymediset/sap-ai-provider)
 [![License: Apache-2.0](https://img.shields.io/badge/License-Apache%202.0-blue.svg)](https://opensource.org/licenses/Apache-2.0)
 [![Vercel AI SDK](https://img.shields.io/badge/Vercel%20AI%20SDK-6.0+-black.svg)](https://sdk.vercel.ai/docs)
+[![Language Model](https://img.shields.io/badge/Language%20Model-V3-green.svg)](https://sdk.vercel.ai/docs/ai-sdk-core/provider-interfaces)
 
 A community provider for SAP AI Core that integrates seamlessly with the Vercel AI SDK. Built on top of the official **@sap-ai-sdk/orchestration** package, this provider enables you to use SAP's enterprise-grade AI models through the familiar Vercel AI SDK interface.
 
@@ -12,15 +13,38 @@ A community provider for SAP AI Core that integrates seamlessly with the Vercel
 - [Quick Start](#quick-start)
 - [Quick Reference](#quick-reference)
 - [Installation](#installation)
+- [Provider Creation](#provider-creation)
+  - [Option 1: Factory Function (Recommended for Custom Configuration)](#option-1-factory-function-recommended-for-custom-configuration)
+  - [Option 2: Default Instance (Quick Start)](#option-2-default-instance-quick-start)
 - [Authentication](#authentication)
 - [Basic Usage](#basic-usage)
+  - [Text Generation](#text-generation)
+  - [Chat Conversations](#chat-conversations)
+  - [Streaming Responses](#streaming-responses)
+  - [Model Configuration](#model-configuration)
 - [Supported Models](#supported-models)
 - [Advanced Features](#advanced-features)
+  - [Tool Calling](#tool-calling)
+  - [Multi-modal Input (Images)](#multi-modal-input-images)
+  - [Data Masking (SAP DPI)](#data-masking-sap-dpi)
+  - [Content Filtering](#content-filtering)
 - [Configuration Options](#configuration-options)
 - [Error Handling](#error-handling)
+- [Troubleshooting](#troubleshooting)
+- [Performance](#performance)
+- [Security](#security)
+- [Debug Mode](#debug-mode)
 - [Examples](#examples)
 - [Migration Guides](#migration-guides)
+  - [Upgrading from v3.x to v4.x](#upgrading-from-v3x-to-v4x)
+  - [Upgrading from v2.x to v3.x](#upgrading-from-v2x-to-v3x)
+  - [Upgrading from v1.x to v2.x](#upgrading-from-v1x-to-v2x)
+- [Important Note](#important-note)
 - [Contributing](#contributing)
+- [Resources](#resources)
+  - [Documentation](#documentation)
+  - [Community](#community)
+  - [Related Projects](#related-projects)
 - [License](#license)
 
 ## Features
@@ -29,11 +53,12 @@ A community provider for SAP AI Core that integrates seamlessly with the Vercel
 - 🎯 **Tool Calling Support** - Full tool/function calling capabilities
 - 🧠 **Reasoning-Safe by Default** - Assistant reasoning parts are not forwarded unless enabled
 - 🖼️ **Multi-modal Input** - Support for text and image inputs
-- 📡 **Streaming Support** - Real-time text generation
+- 📡 **Streaming Support** - Real-time text generation with structured V3 blocks
 - 🔒 **Data Masking** - Built-in SAP DPI integration for privacy
 - 🛡️ **Content Filtering** - Azure Content Safety and Llama Guard support
 - 🔧 **TypeScript Support** - Full type safety and IntelliSense
 - 🎨 **Multiple Models** - Support for GPT-4, Claude, Gemini, Nova, and more
+- ⚡ **Language Model V3** - Latest Vercel AI SDK specification with enhanced streaming
 
 ## Quick Start
 
@@ -263,13 +288,13 @@ This provider supports all models available through SAP AI Core Orchestration se
 **Popular models:**
 
 - **OpenAI**: gpt-4o, gpt-4o-mini, gpt-4.1, o1, o3 (recommended for multi-tool apps)
-- **Anthropic Claude**: claude-3.5-sonnet, claude-4-opus
+- **Anthropic Claude**: anthropic--claude-3.5-sonnet, anthropic--claude-4-opus
 - **Google Gemini**: gemini-2.5-pro, gemini-2.0-flash
 
 ⚠️ **Important:** Google Gemini models have a 1 tool limit per request.
 
-- **Amazon Nova**: nova-pro, nova-lite
-- **Open Source**: mistralai-mistral-large, llama3.1-70b
+- **Amazon Nova**: amazon--nova-pro, amazon--nova-lite
+- **Open Source**: mistralai--mistral-large-instruct, meta--llama3.1-70b-instruct
 
 > **Note:** Model availability depends on your SAP AI Core tenant configuration, region, and subscription.
 
@@ -496,6 +521,23 @@ npx tsx examples/example-generate-text.ts
 
 ## Migration Guides
 
+### Upgrading from v3.x to v4.x
+
+Version 4.0 migrates from **LanguageModelV2** to **LanguageModelV3** specification (AI SDK 6.0+). **See the [Migration Guide](./MIGRATION_GUIDE.md#version-3x-to-4x-breaking-changes) for complete upgrade instructions.**
+
+**Key changes:**
+
+- **Finish Reason**: Changed from string to object (`result.finishReason.unified`)
+- **Usage Structure**: Nested format with detailed token breakdown (`result.usage.inputTokens.total`)
+- **Stream Events**: Structured blocks (`text-start`, `text-delta`, `text-end`) instead of simple deltas
+- **Warning Types**: Updated format with `feature` field for categorization
+
+**Impact by user type:**
+
+- High-level API users (`generateText`/`streamText`): ✅ Minimal impact (likely no changes)
+- Direct provider users: ⚠️ Update type imports (`LanguageModelV2` → `LanguageModelV3`)
+- Custom stream parsers: ⚠️ Update parsing logic for V3 structure
+
 ### Upgrading from v2.x to v3.x
 
 Version 3.0 standardizes error handling to use Vercel AI SDK native error types. **See the [Migration Guide](./MIGRATION_GUIDE.md#v2x--v30) for complete upgrade instructions.**
@@ -526,34 +568,28 @@ Version 2.0 uses the official SAP AI SDK. **See the [Migration Guide](./MIGRATIO
 
 We welcome contributions! Please see our [Contributing Guide](./CONTRIBUTING.md) for details.
 
-## License
-
-Apache License 2.0 - see [LICENSE](./LICENSE.md) for details.
-
-## Support
-
-- 📖 [Documentation](https://github.com/BITASIA/sap-ai-provider)
-- 🐛 [Issue Tracker](https://github.com/BITASIA/sap-ai-provider/issues)
+## Resources
 
-## Documentation
-
-### Guides
-
-- [Environment Setup](./ENVIRONMENT_SETUP.md) - Authentication and configuration setup
-- [Migration Guide](./MIGRATION_GUIDE.md) - Upgrading from v1.x to v2.x with step-by-step instructions
-- [curl API Testing](./CURL_API_TESTING_GUIDE.md) - Direct API testing for debugging
-
-### Reference
+### Documentation
 
+- [Migration Guide](./MIGRATION_GUIDE.md) - Version upgrade instructions (v1.x → v2.x → v3.x → v4.x)
 - [API Reference](./API_REFERENCE.md) - Complete API documentation with all types and functions
+- [Environment Setup](./ENVIRONMENT_SETUP.md) - Authentication and configuration setup
+- [Troubleshooting](./TROUBLESHOOTING.md) - Common issues and solutions
 - [Architecture](./ARCHITECTURE.md) - Internal architecture, design decisions, and request flows
+- [curl API Testing](./CURL_API_TESTING_GUIDE.md) - Direct API testing for debugging
 
-### Contributing
+### Community
 
-- [Contributing Guide](./CONTRIBUTING.md) - How to contribute to this project
+- 🐛 [Issue Tracker](https://github.com/BITASIA/sap-ai-provider/issues) - Report bugs and request features
+- 💬 [Discussions](https://github.com/BITASIA/sap-ai-provider/discussions) - Ask questions and share ideas
 
-## Related
+### Related Projects
 
 - [Vercel AI SDK](https://sdk.vercel.ai/) - The AI SDK this provider extends
 - [SAP AI SDK](https://sap.github.io/ai-sdk/) - Official SAP Cloud SDK for AI
 - [SAP AI Core Documentation](https://help.sap.com/docs/ai-core) - Official SAP AI Core docs
+
+## License
+
+Apache License 2.0 - see [LICENSE](./LICENSE.md) for details.

package/dist/index.cjs

@@ -24766,7 +24766,7 @@ var require_form_data = __commonJS({
     var parseUrl = require("url").parse;
     var fs = require("fs");
     var Stream = require("stream").Stream;
-    var crypto = require("crypto");
+    var crypto2 = require("crypto");
     var mime = require_mime_types();
     var asynckit = require_asynckit();
     var setToStringTag = require_es_set_tostringtag();
@@ -24972,7 +24972,7 @@ var require_form_data = __commonJS({
       return Buffer.concat([dataBuffer, Buffer.from(this._lastBoundary())]);
     };
     FormData2.prototype._generateBoundary = function() {
-      this._boundary = "--------------------------" + crypto.randomBytes(12).toString("hex");
+      this._boundary = "--------------------------" + crypto2.randomBytes(12).toString("hex");
     };
     FormData2.prototype.getLengthSync = function() {
       var knownLength = this._overheadLength + this._valueLength;
@@ -26202,7 +26202,7 @@ var require_axios = __commonJS({
   "node_modules/axios/dist/node/axios.cjs"(exports2, module2) {
     "use strict";
     var FormData$1 = require_form_data();
-    var crypto = require("crypto");
+    var crypto2 = require("crypto");
     var url = require("url");
     var proxyFromEnv = require_proxy_from_env();
     var http = require("http");
@@ -26217,7 +26217,7 @@ var require_axios = __commonJS({
       return e && typeof e === "object" && "default" in e ? e : { "default": e };
     }
     var FormData__default = /* @__PURE__ */ _interopDefaultLegacy(FormData$1);
-    var crypto__default = /* @__PURE__ */ _interopDefaultLegacy(crypto);
+    var crypto__default = /* @__PURE__ */ _interopDefaultLegacy(crypto2);
     var url__default = /* @__PURE__ */ _interopDefaultLegacy(url);
     var proxyFromEnv__default = /* @__PURE__ */ _interopDefaultLegacy(proxyFromEnv);
     var http__default = /* @__PURE__ */ _interopDefaultLegacy(http);
@@ -29897,12 +29897,14 @@ function convertToSAPMessages(prompt, options = {}) {
       }
       case "tool": {
         for (const part of message.content) {
-          const toolMessage = {
-            role: "tool",
-            tool_call_id: part.toolCallId,
-            content: JSON.stringify(part.output)
-          };
-          messages.push(toolMessage);
+          if (part.type === "tool-result") {
+            const toolMessage = {
+              role: "tool",
+              tool_call_id: part.toolCallId,
+              content: JSON.stringify(part.output)
+            };
+            messages.push(toolMessage);
+          }
         }
         break;
       }
@@ -30071,6 +30073,16 @@ See: https://help.sap.com/docs/sap-ai-core/sap-ai-core-service-guide/create-serv
 }
 
 // src/sap-ai-chat-language-model.ts
+var StreamIdGenerator = class {
+  /**
+   * Generates a unique ID for a text block.
+   *
+   * @returns RFC 4122-compliant UUID string
+   */
+  generateTextBlockId() {
+    return crypto.randomUUID();
+  }
+};
 function validateModelParameters(params, warnings) {
   if (params.temperature !== void 0 && (params.temperature < 0 || params.temperature > 2)) {
     warnings.push({
@@ -30162,7 +30174,7 @@ function buildSAPToolParameters(schema) {
   };
 }
 var SAPAIChatLanguageModel = class {
-  specificationVersion = "v2";
+  specificationVersion = "v3";
   modelId;
   config;
   settings;
@@ -30185,10 +30197,14 @@ var SAPAIChatLanguageModel = class {
    * Checks if a URL is supported for file/image uploads.
    *
    * @param url - The URL to check
-   * @returns True if the URL protocol is HTTPS or data (content-type rules are enforced via supportedUrls)
+   * @returns True if the URL protocol is HTTPS or data with valid image format
    */
   supportsUrl(url) {
-    return url.protocol === "https:" || url.protocol === "data:";
+    if (url.protocol === "https:") return true;
+    if (url.protocol === "data:") {
+      return /^data:image\//i.test(url.href);
+    }
+    return false;
   }
   /**
    * Returns supported URL patterns for different content types.
@@ -30201,9 +30217,45 @@ var SAPAIChatLanguageModel = class {
     };
   }
   /**
-   * Gets the provider identifier.
+   * Generates text completion using SAP AI Core's Orchestration API.
+   *
+   * This method implements the `LanguageModelV3.doGenerate` interface,
+   * providing synchronous (non-streaming) text generation with support for:
+   * - Multi-turn conversations with system/user/assistant messages
+   * - Tool calling (function calling) with structured outputs
+   * - Multi-modal inputs (text + images)
+   * - Data masking via SAP DPI
+   * - Content filtering via Azure Content Safety or Llama Guard
+   *
+   * **Return Structure:**
+   * - Finish reason: `{ unified: string, raw?: string }`
+   * - Usage: Nested structure with token breakdown `{ inputTokens: { total, ... }, outputTokens: { total, ... } }`
+   * - Warnings: Array of warnings with `type` and optional `feature` field
+   *
+   * @param options - Generation options including prompt, tools, temperature, etc.
+   * @returns Promise resolving to generation result with content, usage, and metadata
+   *
+   * @throws {InvalidPromptError} If prompt format is invalid
+   * @throws {InvalidArgumentError} If arguments are malformed
+   * @throws {APICallError} If the SAP AI Core API call fails
+   *
+   * @example
+   * ```typescript
+   * const result = await model.doGenerate({
+   *   prompt: [
+   *     { role: 'user', content: [{ type: 'text', text: 'Hello!' }] }
+   *   ],
+   *   temperature: 0.7,
+   *   maxTokens: 100
+   * });
+   *
+   * console.log(result.content); // Array of V3 content parts
+   * console.log(result.finishReason.unified); // "stop", "length", "tool-calls", etc.
+   * console.log(result.usage.inputTokens.total); // Total input tokens
+   * ```
    *
-   * @returns The provider name ('sap-ai')
+   * @since 1.0.0
+   * @since 4.0.0 Updated to LanguageModelV3 interface
    */
   get provider() {
     return this.config.provider;
@@ -30270,11 +30322,11 @@ var SAPAIChatLanguageModel = class {
               const schemaRecord = jsonSchema;
               delete schemaRecord.$schema;
               parameters = buildSAPToolParameters(schemaRecord);
-            } catch {
+            } catch (error) {
               warnings.push({
-                type: "unsupported-tool",
-                tool,
-                details: "Failed to convert tool Zod schema to JSON Schema. Falling back to empty object schema."
+                type: "unsupported",
+                feature: `tool schema conversion for ${tool.name}`,
+                details: `Failed to convert tool Zod schema: ${error instanceof Error ? error.message : String(error)}. Falling back to empty object schema.`
               });
               parameters = buildSAPToolParameters({});
             }
@@ -30298,8 +30350,9 @@ var SAPAIChatLanguageModel = class {
           };
         } else {
           warnings.push({
-            type: "unsupported-tool",
-            tool
+            type: "unsupported",
+            feature: `tool type for ${tool.name}`,
+            details: "Only 'function' tool type is supported."
           });
           return null;
         }
@@ -30351,8 +30404,8 @@ var SAPAIChatLanguageModel = class {
     );
     if (options.toolChoice && options.toolChoice.type !== "auto") {
       warnings.push({
-        type: "unsupported-setting",
-        setting: "toolChoice",
+        type: "unsupported",
+        feature: "toolChoice",
         details: `SAP AI SDK does not support toolChoice '${options.toolChoice.type}'. Using default 'auto' behavior.`
       });
     }
@@ -30412,7 +30465,7 @@ var SAPAIChatLanguageModel = class {
   /**
    * Generates a single completion (non-streaming).
    *
-   * This method implements the `LanguageModelV2.doGenerate` interface,
+   * This method implements the `LanguageModelV3.doGenerate` interface,
    * sending a request to SAP AI Core and returning the complete response.
    *
    * **Features:**
@@ -30420,6 +30473,13 @@ var SAPAIChatLanguageModel = class {
    * - Multi-modal input (text + images)
    * - Data masking (if configured)
    * - Content filtering (if configured)
+   * - Abort signal support (via Promise.race)
+   *
+   * **Note on Abort Signal:**
+   * The abort signal implementation uses Promise.race to reject the promise when
+   * the signal is aborted. However, this does not cancel the underlying HTTP request
+   * to SAP AI Core - the request continues executing on the server. This is a
+   * limitation of the SAP AI SDK's chatCompletion API.
    *
    * @param options - Generation options including prompt, tools, and settings
    * @returns Promise resolving to the generation result with content, usage, and metadata
@@ -30492,7 +30552,7 @@ var SAPAIChatLanguageModel = class {
         Object.entries(responseHeadersRaw).flatMap(([key, value]) => {
           if (typeof value === "string") return [[key, value]];
           if (Array.isArray(value)) {
-            const strings = value.filter((item) => typeof item === "string").join(",");
+            const strings = value.filter((item) => typeof item === "string").join("; ");
             return strings.length > 0 ? [[key, strings]] : [];
           }
           if (typeof value === "number" || typeof value === "boolean") {
@@ -30533,9 +30593,17 @@ var SAPAIChatLanguageModel = class {
         content,
         finishReason,
         usage: {
-          inputTokens: tokenUsage.prompt_tokens,
-          outputTokens: tokenUsage.completion_tokens,
-          totalTokens: tokenUsage.total_tokens
+          inputTokens: {
+            total: tokenUsage.prompt_tokens,
+            noCache: tokenUsage.prompt_tokens,
+            cacheRead: void 0,
+            cacheWrite: void 0
+          },
+          outputTokens: {
+            total: tokenUsage.completion_tokens,
+            text: tokenUsage.completion_tokens,
+            reasoning: void 0
+          }
         },
         providerMetadata: {
           "sap-ai": {
@@ -30566,19 +30634,28 @@ var SAPAIChatLanguageModel = class {
   /**
    * Generates a streaming completion.
    *
-   * This method implements the `LanguageModelV2.doStream` interface,
+   * This method implements the `LanguageModelV3.doStream` interface,
    * sending a streaming request to SAP AI Core and returning a stream of response parts.
    *
    * **Stream Events:**
-   * - `stream-start` - Stream initialization
+   * - `stream-start` - Stream initialization with warnings
    * - `response-metadata` - Response metadata (model, timestamp)
-   * - `text-start` - Text generation starts
-   * - `text-delta` - Incremental text chunks
-   * - `text-end` - Text generation completes
-   * - `tool-call` - Tool call detected
+   * - `text-start` - Text block begins (with unique ID)
+   * - `text-delta` - Incremental text chunks (with block ID)
+   * - `text-end` - Text block completes (with accumulated text)
+   * - `tool-input-start` - Tool input begins
+   * - `tool-input-delta` - Tool input chunk
+   * - `tool-input-end` - Tool input completes
+   * - `tool-call` - Complete tool call
    * - `finish` - Stream completes with usage and finish reason
    * - `error` - Error occurred
    *
+   * **Stream Structure:**
+   * - Text blocks have explicit lifecycle with unique IDs
+   * - Finish reason format: `{ unified: string, raw?: string }`
+   * - Usage format: `{ inputTokens: { total, ... }, outputTokens: { total, ... } }`
+   * - Warnings only in `stream-start` event
+   *
    * @param options - Streaming options including prompt, tools, and settings
    * @returns Promise resolving to stream and request metadata
    *
@@ -30594,8 +30671,13 @@ var SAPAIChatLanguageModel = class {
    *   if (part.type === 'text-delta') {
    *     process.stdout.write(part.delta);
    *   }
+   *   if (part.type === 'text-end') {
+   *     console.log('Block complete:', part.id, part.text);
+   *   }
    * }
    * ```
+   *
+   * @since 4.0.0
    */
   async doStream(options) {
     try {
@@ -30623,12 +30705,25 @@ var SAPAIChatLanguageModel = class {
         options.abortSignal,
         { promptTemplating: { include_usage: true } }
       );
+      const idGenerator = new StreamIdGenerator();
+      let textBlockId = null;
       const streamState = {
-        finishReason: "unknown",
+        finishReason: {
+          unified: "other",
+          raw: void 0
+        },
         usage: {
-          inputTokens: void 0,
-          outputTokens: void 0,
-          totalTokens: void 0
+          inputTokens: {
+            total: void 0,
+            noCache: void 0,
+            cacheRead: void 0,
+            cacheWrite: void 0
+          },
+          outputTokens: {
+            total: void 0,
+            text: void 0,
+            reasoning: void 0
+          }
         },
         isFirstChunk: true,
         activeText: false
@@ -30656,19 +30751,25 @@ var SAPAIChatLanguageModel = class {
               }
               const deltaToolCalls = chunk.getDeltaToolCalls();
               if (Array.isArray(deltaToolCalls) && deltaToolCalls.length > 0) {
-                streamState.finishReason = "tool-calls";
+                streamState.finishReason = {
+                  unified: "tool-calls",
+                  raw: void 0
+                };
               }
               const deltaContent = chunk.getDeltaContent();
-              if (typeof deltaContent === "string" && deltaContent.length > 0 && streamState.finishReason !== "tool-calls") {
+              if (typeof deltaContent === "string" && deltaContent.length > 0 && streamState.finishReason.unified !== "tool-calls") {
                 if (!streamState.activeText) {
-                  controller.enqueue({ type: "text-start", id: "0" });
+                  textBlockId = idGenerator.generateTextBlockId();
+                  controller.enqueue({ type: "text-start", id: textBlockId });
                   streamState.activeText = true;
                 }
-                controller.enqueue({
-                  type: "text-delta",
-                  id: "0",
-                  delta: deltaContent
-                });
+                if (textBlockId) {
+                  controller.enqueue({
+                    type: "text-delta",
+                    id: textBlockId,
+                    delta: deltaContent
+                  });
+                }
               }
               if (Array.isArray(deltaToolCalls) && deltaToolCalls.length > 0) {
                 for (const toolCallChunk of deltaToolCalls) {
@@ -30718,7 +30819,7 @@ var SAPAIChatLanguageModel = class {
               const chunkFinishReason = chunk.getFinishReason();
               if (chunkFinishReason) {
                 streamState.finishReason = mapFinishReason(chunkFinishReason);
-                if (streamState.finishReason === "tool-calls") {
+                if (streamState.finishReason.unified === "tool-calls") {
                   const toolCalls2 = Array.from(toolCallsInProgress.values());
                   for (const tc of toolCalls2) {
                     if (tc.didEmitCall) {
@@ -30747,8 +30848,8 @@ var SAPAIChatLanguageModel = class {
                       input: tc.arguments
                     });
                   }
-                  if (streamState.activeText) {
-                    controller.enqueue({ type: "text-end", id: "0" });
+                  if (streamState.activeText && textBlockId) {
+                    controller.enqueue({ type: "text-end", id: textBlockId });
                     streamState.activeText = false;
                   }
                 }
@@ -30784,20 +30885,24 @@ var SAPAIChatLanguageModel = class {
                 input: tc.arguments
               });
             }
-            if (streamState.activeText) {
-              controller.enqueue({ type: "text-end", id: "0" });
+            if (streamState.activeText && textBlockId) {
+              controller.enqueue({ type: "text-end", id: textBlockId });
             }
             const finalFinishReason = streamResponse.getFinishReason();
             if (finalFinishReason) {
               streamState.finishReason = mapFinishReason(finalFinishReason);
             } else if (didEmitAnyToolCalls) {
-              streamState.finishReason = "tool-calls";
+              streamState.finishReason = {
+                unified: "tool-calls",
+                raw: void 0
+              };
             }
             const finalUsage = streamResponse.getTokenUsage();
             if (finalUsage) {
-              streamState.usage.inputTokens = finalUsage.prompt_tokens;
-              streamState.usage.outputTokens = finalUsage.completion_tokens;
-              streamState.usage.totalTokens = finalUsage.total_tokens;
+              streamState.usage.inputTokens.total = finalUsage.prompt_tokens;
+              streamState.usage.inputTokens.noCache = finalUsage.prompt_tokens;
+              streamState.usage.outputTokens.total = finalUsage.completion_tokens;
+              streamState.usage.outputTokens.text = finalUsage.completion_tokens;
             }
             controller.enqueue({
               type: "finish",
@@ -30828,8 +30933,7 @@ var SAPAIChatLanguageModel = class {
         stream: transformedStream,
         request: {
           body: requestBody
-        },
-        warnings: warningsOut
+        }
       };
     } catch (error) {
       throw convertToAISDKError(error, {
@@ -30841,27 +30945,28 @@ var SAPAIChatLanguageModel = class {
   }
 };
 function mapFinishReason(reason) {
-  if (!reason) return "unknown";
+  const raw = reason;
+  if (!reason) return { unified: "other", raw };
   switch (reason.toLowerCase()) {
     case "stop":
     case "end_turn":
     case "stop_sequence":
     case "eos":
-      return "stop";
+      return { unified: "stop", raw };
     case "length":
     case "max_tokens":
     case "max_tokens_reached":
-      return "length";
+      return { unified: "length", raw };
     case "tool_calls":
     case "tool_call":
     case "function_call":
-      return "tool-calls";
+      return { unified: "tool-calls", raw };
     case "content_filter":
-      return "content-filter";
+      return { unified: "content-filter", raw };
     case "error":
-      return "error";
+      return { unified: "error", raw };
     default:
-      return "other";
+      return { unified: "other", raw };
   }
 }