nodebb-plugin-search-agent 0.0.9 → 0.0.91

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "nodebb-plugin-search-agent",
-  "version": "0.0.9",
+  "version": "0.0.91",
   "description": "NodeBB plugin that adds a floating chat assistant to help users find relevant forum topics using TF-IDF text similarity",
   "main": "library.js",
   "author": "Racheli Bayfus",

package/services/embeddingService.js

@@ -11,6 +11,12 @@ const OPENAI_EMBEDDINGS_PATH = '/v1/embeddings';
 const EMBEDDING_MODEL = 'text-embedding-3-small';
 const MAX_RETRIES = 3;
 const RETRY_DELAY_MS = 500;
+// text-embedding-3-small supports 8 192 tokens; ~4 chars/token → cap at 30 000 chars (~7 500 tokens)
+const MAX_CHARS = 30000;
+
+function truncate(text) {
+	return text.length > MAX_CHARS ? text.slice(0, MAX_CHARS) : text;
+}
 
 /**
  * Performs an HTTPS POST request to the OpenAI embeddings endpoint.
@@ -95,8 +101,9 @@ async function embed(text) {
 		throw new Error('OPENAI_API_KEY environment variable is not set');
 	}
 
-	winston().verbose(`[search-agent] embeddingService: generating embedding for text (${text.length} chars)`);
-	const response = await withRetry(() => requestEmbeddings(apiKey, text));
+	const safe = truncate(text);
+	winston().verbose(`[search-agent] embeddingService: generating embedding for text (${safe.length} chars)`);
+	const response = await withRetry(() => requestEmbeddings(apiKey, safe));
 	winston().verbose('[search-agent] embeddingService: embedding generated successfully');
 	return response.data[0].embedding;
 }
@@ -122,8 +129,9 @@ async function embedBatch(texts) {
 		throw new Error('OPENAI_API_KEY environment variable is not set');
 	}
 
-	winston().verbose(`[search-agent] embeddingService: generating batch embeddings for ${texts.length} text(s)`);
-	const response = await withRetry(() => requestEmbeddings(apiKey, texts));
+	const safeTexts = texts.map(truncate);
+	winston().verbose(`[search-agent] embeddingService: generating batch embeddings for ${safeTexts.length} text(s)`);
+	const response = await withRetry(() => requestEmbeddings(apiKey, safeTexts));
 	winston().verbose(`[search-agent] embeddingService: batch embeddings generated successfully (${texts.length} vector(s))`);
 
 	// OpenAI returns items sorted by index field, but sort explicitly to be safe