Updated

main.py

@@ -7,8 +7,8 @@ from spitch import Spitch
 from langchain.prompts import PromptTemplate
 from langchain_huggingface import HuggingFaceEndpoint
 from langdetect import detect, DetectorFactory
-from huggingface_hub.utils import HfHubHTTPError  # for quota error handling
-from smebuilder_vector import retriever  # <-- your retriever
+from huggingface_hub.utils import HfHubHTTPError
+from smebuilder_vector import retriever  # your retriever
 
 # ----------------- CONFIG -----------------
 DetectorFactory.seed = 0
@@ -25,14 +25,14 @@ if not SPITCH_API_KEY:
 os.environ["SPITCH_API_KEY"] = SPITCH_API_KEY
 spitch_client = Spitch()
 
-# HuggingFace LLM (better tuned for code generation)
+# HuggingFace LLM
 llm = HuggingFaceEndpoint(
     repo_id=HF_MODEL,
     temperature=0.7,
     top_p=0.9,
     do_sample=True,
     repetition_penalty=1.1,
-    max_new_tokens=2048  # bumped tokens
+    max_new_tokens=2048
 )
 
 # FastAPI app
@@ -143,8 +143,7 @@ async def stt_audio(file: UploadFile = File(...), lang_hint: str | None = None,
 
     suffix = os.path.splitext(file.filename)[1] or ".wav"
     with tempfile.NamedTemporaryFile(delete=False, suffix=suffix) as tf:
-        content = await file.read()
-        tf.write(content)
+        tf.write(await file.read())
         tmp_path = tf.name
 
     try:
@@ -156,7 +155,6 @@ async def stt_audio(file: UploadFile = File(...), lang_hint: str | None = None,
         resp = spitch_client.speech.transcribe(language="en", content=open(tmp_path, "rb").read())
 
     transcription = getattr(resp, "text", "") or (resp.get("text", "") if isinstance(resp, dict) else "")
-
     try:
         detected_lang = detect(transcription) if transcription.strip() else "en"
     except Exception:
@@ -171,7 +169,6 @@ async def stt_audio(file: UploadFile = File(...), lang_hint: str | None = None,
             translation = transcription
 
     reply = stt_chain.invoke({"speech": translation})
-
     return {
         "transcription": transcription,
         "detected_language": detected_lang,
@@ -189,7 +186,6 @@ def autodoc(req: AutoDocRequest, authorization: str | None = Header(None)):
 async def sme_generate(payload: dict = Body(...)):
     try:
         user_prompt = payload.get("user_prompt", "")
-        # retrieve context
         context_docs = retriever.get_relevant_documents(user_prompt)
         context = "\n".join([doc.page_content for doc in context_docs]) if context_docs else "No extra context"
         response = sme_chain.invoke({"user_prompt": user_prompt, "context": context})
@@ -205,8 +201,7 @@ async def sme_speech_generate(file: UploadFile = File(...), lang_hint: str | Non
 
     suffix = os.path.splitext(file.filename)[1] or ".wav"
     with tempfile.NamedTemporaryFile(delete=False, suffix=suffix) as tf:
-        content = await file.read()
-        tf.write(content)
+        tf.write(await file.read())
         tmp_path = tf.name
 
     try:
@@ -218,7 +213,6 @@ async def sme_speech_generate(file: UploadFile = File(...), lang_hint: str | Non
         resp = spitch_client.speech.transcribe(language="en", content=open(tmp_path, "rb").read())
 
     transcription = getattr(resp, "text", "") or (resp.get("text", "") if isinstance(resp, dict) else "")
-
     try:
         detected_lang = detect(transcription) if transcription.strip() else "en"
     except Exception:
@@ -233,10 +227,8 @@ async def sme_speech_generate(file: UploadFile = File(...), lang_hint: str | Non
             translation = transcription
 
     try:
-        # vector retrieval here too
         context_docs = retriever.get_relevant_documents(translation)
         context = "\n".join([doc.page_content for doc in context_docs]) if context_docs else "No extra context"
-
         sme_response = sme_chain.invoke({"user_prompt": translation, "context": context})
         return {
             "success": True,
@@ -250,7 +242,7 @@ async def sme_speech_generate(file: UploadFile = File(...), lang_hint: str | Non
             return {"success": False, "error": "⚠️ Token quota for today has been used. Please come back in 24 hours."}
         raise e
 
-# Hugging Face requires port 7860
+# ----------------- MAIN -----------------
 if __name__ == "__main__":
     import uvicorn
     uvicorn.run("main:app", host="0.0.0.0", port=7860, reload=False)