Update dependency @sap-ai-sdk/orchestration to v2 (#56)

* Update dependency @sap-ai-sdk/orchestration to v2

* Refactor orchestration client usage in chatCompletion functions

---------

Co-authored-by: renovate[bot] <29139614+renovate[bot]@users.noreply.github.com>
Co-authored-by: Daniel Richter <[email protected]>

Author: renovate[bot]

pr-review/package-lock.json

@@ -21,7 +21,7 @@
     "@octokit/plugin-throttling": "8.2.0",
     "@octokit/webhooks": "13.9.1",
     "@sap-ai-sdk/ai-api": "2.1.0",
-    "@sap-ai-sdk/orchestration": "1.18.0",
+    "@sap-ai-sdk/orchestration": "2.1.0",
     "axios": "1.13.1",
     "minimatch": "10.1.1",
     "mollitia": "0.2.0",

pr-review/package.json

@@ -1,5 +1,5 @@
 import * as core from "@actions/core"
-import { ChatMessage, OrchestrationClient, TokenUsage } from "@sap-ai-sdk/orchestration"
+import { ChatMessage, OrchestrationClient } from "@sap-ai-sdk/orchestration"
 import { isAxiosError } from "axios"
 import { inspect } from "node:util"
 import { z } from "zod"
@@ -30,22 +30,24 @@ export async function chatCompletion(messages: ChatMessage[]): Promise<string> {
     core.info("Use the OrchestrationClient to call the model")
     const orchestrationClient = new OrchestrationClient(
       {
-        llm: {
-          model_name: config.model,
-          model_params: config.modelParameters,
-          model_version: config.modelVersion,
-        },
-        templating: {
-          template: messages,
+        promptTemplating: {
+          model: {
+            name: config.model,
+            params: config.modelParameters,
+            version: config.modelVersion,
+          },
+          prompt: {
+            template: messages,
+          },
         },
       },
       config.deploymentConfig,
     )
     const completion = await orchestrationClient.chatCompletion()
-    core.info(inspect(completion.data, { depth: undefined, colors: true }))
+    core.info(inspect(completion._data, { depth: undefined, colors: true }))
 
-    modelName = (completion.data?.module_results?.llm?.model_name as string) ?? modelName
-    const tokenUsage: TokenUsage = completion.getTokenUsage()
+    modelName = completion._data.final_result.model ?? modelName
+    const tokenUsage = completion.getTokenUsage()
     promptTokens += tokenUsage.prompt_tokens
     completionTokens += tokenUsage.completion_tokens
 
@@ -68,27 +70,29 @@ export async function chatCompletionWithJsonSchema<T extends z.ZodTypeAny>(zodSc
   try {
     core.info("Use the OrchestrationClient to call the model")
     const orchestrationClient = new OrchestrationClient({
-      llm: {
-        model_name: config.model,
-        model_params: config.modelParameters,
-        model_version: config.modelVersion,
-      },
-      templating: {
-        template: [
-          ...messages,
-          {
-            role: "user",
-            content: `Always return a plain JSON document with the following schema:\n\n ${JSON.stringify(jsonSchema, undefined, 2)} `,
-          },
-        ],
+      promptTemplating: {
+        model: {
+          name: config.model,
+          params: config.modelParameters,
+          version: config.modelVersion,
+        },
+        prompt: {
+          template: [
+            ...messages,
+            {
+              role: "user",
+              content: `Always return a plain JSON document with the following schema:\n\n ${JSON.stringify(jsonSchema, undefined, 2)} `,
+            },
+          ],
+        },
       },
     })
     const completion = await orchestrationClient.chatCompletion()
     core.info(inspect(completion.rawResponse.data, { depth: undefined, colors: true }))
     responseJson = `${completion.getContent()}`
 
-    modelName = (completion.data?.module_results?.llm?.model_name as string) ?? modelName
-    const tokenUsage: TokenUsage = completion.getTokenUsage()
+    modelName = completion._data.final_result.model ?? modelName
+    const tokenUsage = completion.getTokenUsage()
     promptTokens += tokenUsage.prompt_tokens
     completionTokens += tokenUsage.completion_tokens
   } catch (error) {

pr-review/src/ai-core-client.ts

@@ -20,7 +20,7 @@
     "@octokit/plugin-throttling": "8.2.0",
     "@octokit/webhooks": "13.9.1",
     "@sap-ai-sdk/ai-api": "2.1.0",
-    "@sap-ai-sdk/orchestration": "1.18.0",
+    "@sap-ai-sdk/orchestration": "2.1.0",
     "axios": "1.13.1",
     "minimatch": "10.1.1",
     "mollitia": "0.2.0",

pr-summary/package-lock.json

@@ -1,5 +1,5 @@
 import * as core from "@actions/core"
-import { ChatMessage, OrchestrationClient, TokenUsage } from "@sap-ai-sdk/orchestration"
+import { ChatMessage, OrchestrationClient } from "@sap-ai-sdk/orchestration"
 import { isAxiosError } from "axios"
 import { inspect } from "node:util"
 import { config } from "./config.js"
@@ -26,22 +26,24 @@ export async function chatCompletion(messages: ChatMessage[]): Promise<string> {
     core.info("Call OrchestrationClient")
     const orchestrationClient = new OrchestrationClient(
       {
-        llm: {
-          model_name: config.model,
-          model_params: config.modelParameters,
-          model_version: config.modelVersion,
-        },
-        templating: {
-          template: messages,
+        promptTemplating: {
+          model: {
+            name: config.model,
+            params: config.modelParameters,
+            version: config.modelVersion,
+          },
+          prompt: {
+            template: messages,
+          },
         },
       },
       config.deploymentConfig,
     )
     const completion = await orchestrationClient.chatCompletion()
-    core.info(inspect(completion.data, { depth: undefined, colors: true }))
+    core.info(inspect(completion._data, { depth: undefined, colors: true }))
 
-    modelName = (completion.data?.module_results?.llm?.model_name as string) ?? modelName
-    const tokenUsage: TokenUsage = completion.getTokenUsage()
+    modelName = completion._data.final_result.model ?? modelName
+    const tokenUsage = completion.getTokenUsage()
     promptTokens += tokenUsage.prompt_tokens
     completionTokens += tokenUsage.completion_tokens