Merge branch 'main' of https://huggingface.co/spaces/kk53/rag_lithuania

app.py

@@ -44,6 +44,26 @@ def load_model():
     n_threads = 2# Verbose is required to pass to the callback manager
 )
     st.success("loaded the second NLP model from Hugging Face!") 
+
+    model_2_name = "TheBloke/zephyr-7B-beta-GGUF"
+    model_2base_name = "zephyr-7b-beta.Q4_K_M.gguf"
+    model_path_model = hf_hub_download(
+    repo_id=model_2_name,
+    filename=model_2base_name,
+    cache_dir= '/content/models' # Directory for the model
+)   
+    callback_manager = CallbackManager([StreamingStdOutCallbackHandler()])
+    llm = LlamaCpp(
+    model_path=model_path_model,
+    temperature=0.75,
+    max_tokens=2500,
+    top_p=1,
+    callback_manager=callback_manager,
+    verbose=True, 
+    n_ctx=2048,
+    n_threads = 2# Verbose is required to pass to the callback manager
+)
+    st.success("loaded the second NLP model from Hugging Face!") 
 #     prompt_template = "<|system|>\
 # </s>\
 # <|user|>\
@@ -51,6 +71,8 @@ def load_model():
 # <|assistant|>"
 #     template = prompt_template
 #     prompt = PromptTemplate.from_template(template)
+
+    return model, llm
 #     callback_manager = CallbackManager([StreamingStdOutCallbackHandler()])
 #     llm = LlamaCpp(
 #     model_path=model_path_model,
@@ -66,6 +88,9 @@ def load_model():
     
 st.title("Please ask your question on Lithuanian rules for foreigners.")
 model,llm  = load_model()
+pc = Pinecone(api_key=apikeys)
+index = pc.Index("law")
+model,llm  = load_model()
 pc = Pinecone(api_key="003117b0-6caf-4de4-adf9-cc49da6587e6")
 index = pc.Index("law")
 question = st.text_input("Enter your question:")
@@ -81,10 +106,14 @@ response_t = response['matches'][0]['metadata']['text']
 st.header("Answer:")
 st.write(response_t)
 
-# if question:
-#     # Perform Question Answering
-#     answer = qa_chain(context=context, question=question)
-    
-#     # Display the answer
-#     st.header("Answer:")
-#     st.write(answer)
+query = model.create_embedding(question)
+q = query['data'][0]['embedding']
+response = index.query(
+  vector=q,
+  top_k=1,
+  include_metadata = True,
+  namespace = "ns1"
+)
+response_t = response['matches'][0]['metadata']['text']
+st.header("Answer:")
+st.write(response_t)