Update app.py

app.py

@@ -15,7 +15,7 @@ HF_MODEL_NAME = "all-MiniLM-L6-v2"
 LLM_REPO = "HuggingFaceH4/zephyr-7b-beta"
 HUGGINGFACE_API_TOKEN = os.getenv("HUGGINGFACE_API_KEY")
 
-# Set Hugging Face API token for litellm
+# Set Hugging Face API token
 os.environ["HUGGINGFACEHUB_API_TOKEN"] = HUGGINGFACE_API_TOKEN 
 
 # Language Configuration
@@ -31,14 +31,11 @@ LANGUAGE_CODES = {
 def speech_to_text(audio_file, language_code):
     """Convert uploaded audio file to text"""
     recognizer = sr.Recognizer()
-
     try:
-        # Save uploaded file to temporary location
         with tempfile.NamedTemporaryFile(delete=False) as tmp_file:
             tmp_file.write(audio_file.getvalue())
             audio_path = tmp_file.name
 
-        # Convert to WAV if necessary
         if not audio_path.endswith('.wav'):
             audio = AudioSegment.from_file(audio_path)
             wav_path = audio_path + ".wav"
@@ -72,21 +69,15 @@ def load_knowledge_base():
         return None
 
 def setup_agents(language='en'):
-    prompts = {
-        'en': "Explain space concepts clearly in English",
-        'es': "Explica conceptos espaciales en español",
-        'fr': "Expliquez les concepts spatiaux en français",
-        'de': "Erklären Sie Raumfahrtkonzepte auf Deutsch",
-        'zh': "用中文清楚解释空间概念",
-        'ar': "اشرح مفاهيم الفضاء باللغة العربية"
-    }
-
     researcher = Agent(
         role="Multilingual Space Analyst",
         goal="Analyze and validate space information",
         backstory="Expert in multilingual space data analysis with NASA mission experience.",
         verbose=True,
-        llm=LLM_REPO,
+        llm={
+            "model": "HuggingFaceH4/zephyr-7b-beta",
+            "provider": "huggingface"
+        },
         llm_kwargs={
             "temperature": 0.4,
             "max_length": 512
@@ -99,7 +90,10 @@ def setup_agents(language='en'):
         goal=f"Explain complex concepts in {language} using simple terms",
         backstory=f"Multilingual science communicator specializing in {language} explanations.",
         verbose=True,
-        llm=LLM_REPO,
+        llm={
+            "model": "HuggingFaceH4/zephyr-7b-beta",
+            "provider": "huggingface"
+        },
         llm_kwargs={
             "temperature": 0.5,
             "max_length": 612
@@ -120,8 +114,7 @@ def process_question(question, target_lang='en'):
             NASA Context: {nasa_data.get('explanation', '')}
             Language: {target_lang}""",
             agent=researcher,
-            expected_output="3 verified technical points",
-            output_file="research.md"
+            expected_output="3 verified technical points"
         )
 
         explain_task = Task(
@@ -142,14 +135,12 @@ def process_question(question, target_lang='en'):
         return f"Error: {str(e)}"
 
 # Streamlit Interface
-st.title("🚀COSMOLAB (Multilingual Space Agent)")
+st.title("🚀 COSMOLAB (Multilingual Space Agent)")
 st.markdown("### Ask space questions in any language!")
 
-# Single language selection for both input and output
 selected_lang = st.selectbox("Select Language", list(LANGUAGE_CODES.keys()))
-lang_code = LANGUAGE_CODES[selected_lang].split('-')[0]  # Extract base language code
+lang_code = LANGUAGE_CODES[selected_lang].split('-')[0]
 
-# Input Method
 input_method = st.radio("Input Method", ["Text", "Audio File"])
 
 question = ""
@@ -157,7 +148,6 @@ if input_method == "Text":
     question = st.text_input(f"Your space question in {selected_lang}:", "")
 else:
     audio_file = st.file_uploader("Upload audio file", type=["wav", "mp3", "ogg"])
-
     if audio_file is not None:
         with st.spinner("Processing audio..."):
             question = speech_to_text(audio_file, LANGUAGE_CODES[selected_lang])