use in orchestration

packages/gen-ai-hub/src/orchestration/orchestration-client.ts

@@ -1,8 +1,5 @@
 import { executeRequest, CustomRequestConfig } from '@sap-ai-sdk/core';
-import {
-  DeploymentResolver,
-  resolveDeployment
-} from '../utils/deployment-resolver.js';
+import { resolveDeployment } from '../utils/deployment-resolver.js';
 import {
   CompletionPostRequest,
   CompletionPostResponse
@@ -16,25 +13,31 @@ export class OrchestrationClient {
   /**
    * Creates a completion for the chat messages.
    * @param data - The input parameters for the chat completion.
-   * @param deploymentResolver - A deployment ID or a function to retrieve it.
+   * @param deploymentId - A deployment ID or undefined to retrieve it based on the given model.
    * @param requestConfig - Request configuration.
    * @returns The completion result.
    */
   async chatCompletion(
     data: OrchestrationCompletionParameters,
-    deploymentResolver: DeploymentResolver = () =>
-      resolveDeployment({ scenarioId: 'orchestration' }),
+    deploymentId?: string,
     requestConfig?: CustomRequestConfig
   ): Promise<CompletionPostResponse> {
     const body = constructCompletionPostRequest(data);
-    const deployment =
-      typeof deploymentResolver === 'function'
-        ? (await deploymentResolver()).id
-        : deploymentResolver;
+    deploymentId =
+      deploymentId ??
+      (
+        await resolveDeployment({
+          scenarioId: 'orchestration',
+          model: {
+            name: data.llmConfig.model_name,
+            version: data.llmConfig.model_version
+          }
+        })
+      ).id;
 
     const response = await executeRequest(
       {
-        url: `/inference/deployments/${deployment}/completion`
+        url: `/inference/deployments/${deploymentId}/completion`
       },
       body,
       requestConfig

sample-code/src/aiservice.ts

@@ -10,9 +10,12 @@ const openAiClient = new OpenAiClient();
  * @returns The answer from GPT.
  */
 export async function chatCompletion(): Promise<string> {
-  const response = await openAiClient.chatCompletion('gpt-35-turbo', {
-    messages: [{ role: 'user', content: 'What is the capital of France?' }]
-  });
+  const response = await openAiClient.chatCompletion(
+    {
+      messages: [{ role: 'user', content: 'What is the capital of France?' }]
+    },
+    'gpt-35-turbo'
+  );
   const assistantMessage = response.choices[0]
     .message as OpenAiChatAssistantMessage;
   return assistantMessage.content!;
@@ -23,8 +26,12 @@ export async function chatCompletion(): Promise<string> {
  * @returns An embedding vector.
  */
 export async function computeEmbedding(): Promise<number[]> {
-  const response = await openAiClient.embeddings('text-embedding-ada-002', {
-    input: 'Hello, world!'
-  });
+  const response = await openAiClient.embeddings(
+    {
+      input: 'Hello, world!'
+    },
+    'text-embedding-ada-002'
+  );
+
   return response.data[0].embedding;
 }