refactor(embeddings): centralize EMBEDDING_PROVIDER parsing

Author: PatrickSys

src/core/indexer.ts

@@ -21,7 +21,12 @@ import {
 } from '../types/index.js';
 import { analyzerRegistry } from './analyzer-registry.js';
 import { isCodeFile, isBinaryFile } from '../utils/language-detection.js';
-import { getEmbeddingProvider, getConfiguredDimensions, DEFAULT_MODEL } from '../embeddings/index.js';
+import {
+  getEmbeddingProvider,
+  getConfiguredDimensions,
+  DEFAULT_MODEL,
+  parseEmbeddingProviderName
+} from '../embeddings/index.js';
 import { getStorageProvider, CodeChunkWithEmbedding } from '../storage/index.js';
 import {
   LibraryUsageTracker,
@@ -240,14 +245,8 @@ export class CodebaseIndexer {
   }
 
   private mergeConfig(userConfig?: Partial<CodebaseConfig>): CodebaseConfig {
-    const envEmbeddingProvider = process.env.EMBEDDING_PROVIDER;
     const defaultEmbeddingProvider =
-      envEmbeddingProvider === 'openai' ||
-      envEmbeddingProvider === 'transformers' ||
-      envEmbeddingProvider === 'ollama' ||
-      envEmbeddingProvider === 'custom'
-        ? envEmbeddingProvider
-        : 'transformers';
+      parseEmbeddingProviderName(process.env.EMBEDDING_PROVIDER) ?? 'transformers';
 
     const defaultConfig: CodebaseConfig = {
       analyzers: {

src/embeddings/index.ts

@@ -1,7 +1,13 @@
 export * from './types.js';
 export * from './transformers.js';
 
-import { EmbeddingProvider, EmbeddingConfig, DEFAULT_EMBEDDING_CONFIG, DEFAULT_MODEL } from './types.js';
+import {
+  EmbeddingProvider,
+  EmbeddingConfig,
+  DEFAULT_EMBEDDING_CONFIG,
+  DEFAULT_MODEL,
+  parseEmbeddingProviderName
+} from './types.js';
 import { TransformersEmbeddingProvider, MODEL_CONFIGS } from './transformers.js';
 
 /**
@@ -12,7 +18,7 @@ import { TransformersEmbeddingProvider, MODEL_CONFIGS } from './transformers.js'
  * implementation) so new models are automatically handled without updating this function.
  */
 export function getConfiguredDimensions(config: Partial<EmbeddingConfig> = {}): number {
-  const provider = config.provider ?? (process.env.EMBEDDING_PROVIDER as string) ?? 'transformers';
+  const provider = config.provider ?? parseEmbeddingProviderName(process.env.EMBEDDING_PROVIDER) ?? 'transformers';
   const model = config.model ?? process.env.EMBEDDING_MODEL ?? DEFAULT_MODEL;
   if (provider === 'openai') return 1536; // text-embedding-3-small / ada-002
   // Look up from the same MODEL_CONFIGS the provider uses — avoids stale hardcoded guesses

src/embeddings/types.ts

@@ -18,13 +18,20 @@ export interface EmbeddingConfig {
   apiEndpoint?: string;
 }
 
+export function parseEmbeddingProviderName(value: unknown): EmbeddingConfig['provider'] | undefined {
+  if (value === 'transformers' || value === 'ollama' || value === 'openai' || value === 'custom') {
+    return value;
+  }
+  return undefined;
+}
+
 // Default: bge-small (fast, ~2min indexing, consumer-hardware safe)
 // Opt-in: set EMBEDDING_MODEL=onnx-community/granite-embedding-small-english-r2-ONNX for
 // better conceptual search at the cost of 5-10x slower indexing and higher RAM usage
 export const DEFAULT_MODEL = process.env.EMBEDDING_MODEL || 'Xenova/bge-small-en-v1.5';
 
 export const DEFAULT_EMBEDDING_CONFIG: EmbeddingConfig = {
-  provider: (process.env.EMBEDDING_PROVIDER as EmbeddingConfig['provider']) || 'transformers',
+  provider: parseEmbeddingProviderName(process.env.EMBEDDING_PROVIDER) ?? 'transformers',
   model: DEFAULT_MODEL,
   batchSize: 32,
   maxRetries: 3,