fix raptor bugs (#928)

### What problem does this PR solve?

#922 
### Type of change

- [x] Bug Fix (non-breaking change which fixes an issue)

rag/llm/chat_model.py

@@ -359,7 +359,6 @@ class VolcEngineChat(Base):
         if system:
             history.insert(0, {"role": "system", "content": system})
         ans = ""
-        tk_count = 0
         try:
             req = {
                 "parameters": {
@@ -380,6 +379,7 @@ class VolcEngineChat(Base):
                 if resp.choices[0].finish_reason == "stop":
                     tk_count = resp.usage.total_tokens
                 yield ans
+
         except Exception as e:
             yield ans + "\n**ERROR**: " + str(e)
         yield tk_count

rag/raptor.py

@@ -95,6 +95,7 @@ class RecursiveAbstractiveProcessing4TreeOrganizedRetrieval:
                 gm.fit(reduced_embeddings)
                 probs = gm.predict_proba(reduced_embeddings)
                 lbls = [np.where(prob > self._threshold)[0] for prob in probs]
+                lbls = [lbl[0] if isinstance(lbl, np.ndarray) else lbl for lbl in lbls]
             lock = Lock()
             with ThreadPoolExecutor(max_workers=12) as executor:
                 threads = []