feat(watsonX): add region environment variable

Signed-off-by: Lukáš Janeček <[email protected]>
i-am-bee · Oct 18, 2024 · e0c1654 · e0c1654
1 parent fcd47f1
commit e0c1654
Show file tree

Hide file tree

Showing 4 changed files with 10 additions and 3 deletions.
diff --git a/.env.example b/.env.example
@@ -58,6 +58,7 @@ OPENAI_API_KEY=
 # https://www.ibm.com/products/watsonx-ai
 WATSONX_API_KEY=
 WATSONX_PROJECT_ID=
+WATSONX_REGION= # optional
 
 BAM_API_KEY=
 

diff --git a/src/config.ts b/src/config.ts
@@ -96,6 +96,7 @@ export const BAM_API_KEY = getEnv('BAM_API_KEY', null);
 
 export const WATSONX_API_KEY = getEnv('WATSONX_API_KEY', null);
 export const WATSONX_PROJECT_ID = getEnv('WATSONX_PROJECT_ID', null);
+export const WATSONX_REGION = getEnv('WATSONX_REGION', null);
 
 export const CAIKIT_URL = getEnv('CAIKIT_URL', null);
 export const CAIKIT_CA_CERT = getEnv('CAIKIT_CA_CERT', null);

diff --git a/src/embedding/factory.ts b/src/embedding/factory.ts
@@ -34,7 +34,8 @@ import {
   OLLAMA_URL,
   OPENAI_API_KEY,
   WATSONX_API_KEY,
-  WATSONX_PROJECT_ID
+  WATSONX_PROJECT_ID,
+  WATSONX_REGION
 } from '@/config';
 
 export function getDefaultEmbeddingModel(backend: EmbeddingBackend = EMBEDDING_BACKEND) {
@@ -81,7 +82,8 @@ export async function createEmbeddingAdapter(
       const llm = new WatsonXLLM({
         modelId: 'foobar',
         apiKey: WATSONX_API_KEY,
-        projectId: WATSONX_PROJECT_ID
+        projectId: WATSONX_PROJECT_ID,
+        region: WATSONX_REGION ?? undefined
       });
       // @ts-expect-error use protected property
       const client = llm.client;

diff --git a/src/runs/execution/factory.ts b/src/runs/execution/factory.ts
@@ -48,7 +48,8 @@ import {
   OLLAMA_URL,
   OPENAI_API_KEY,
   WATSONX_API_KEY,
-  WATSONX_PROJECT_ID
+  WATSONX_PROJECT_ID,
+  WATSONX_REGION
 } from '@/config';
 
 export function getDefaultModel(backend: LLMBackend = LLM_BACKEND) {
@@ -137,6 +138,7 @@ export function createChatLLM(run: Loaded<Run>, backend: LLMBackend = LLM_BACKEN
       return WatsonXChatLLM.fromPreset(run.model as WatsonXChatLLMPresetModel, {
         apiKey: WATSONX_API_KEY,
         projectId: WATSONX_PROJECT_ID,
+        region: WATSONX_REGION ?? undefined,
         parameters: (parameters) => ({
           ...parameters,
           top_p: run.topP ?? parameters.top_p,
@@ -187,6 +189,7 @@ export function createCodeLLM(backend: LLMBackend = LLM_BACKEND) {
         modelId: 'meta-llama/llama-3-1-70b-instruct',
         apiKey: WATSONX_API_KEY,
         projectId: WATSONX_PROJECT_ID,
+        region: WATSONX_REGION ?? undefined,
         parameters: {
           decoding_method: 'greedy',
           include_stop_sequence: false,