working version

python/molina/content/vector.py

@@ -7,7 +7,8 @@
 from langchain_core.documents import Document
 # from langchain_community.vectorstores.chroma import Chroma 
 from langchain_community.embeddings import GPT4AllEmbeddings
-
+from content import documents
+from generators import gpt4all
 # last
 from langchain_chroma import Chroma
 
@@ -53,6 +54,9 @@ def create_vectordb(db_path: str, chunks: list[Document], embedding_model, verbo
 
     return db3 
 
+# -- ------------------------------------------------------------------------------- -- #
+# -- ------------------------------------------------------------------------------- -- #
+
 def search_vectordb(client, query):
 
     found_search = client.similarity_search(query)
@@ -62,3 +66,24 @@ def search_vectordb(client, query):
         found_l.append(found_search[i].model_dump()["page_content"])
 
     return found_l
+
+# -- ------------------------------------------------------------------------------- -- #
+# -- ------------------------------------------------------------------------------- -- #
+
+def setup_vectordb(db_path: str, docs_path: str):
+    """
+    Setup the vector database with documents.
+    """
+
+    k_documents = documents.load_documents(data_path=docs_path)
+    ks_documents = documents.split_text(k_documents, c_size=1000, c_overlap=200)
+
+    chromadb = create_vectordb(
+        db_path=db_path,
+        chunks=ks_documents,
+        embedding_model=gpt4all.embedding_model,
+        verbose=True
+    )
+
+    return chromadb
+

python/molina/generators/llama.py

@@ -11,6 +11,14 @@
 def embeddings_model():
     """
 
+    # to access to the model, for the case of meta-llama
+    # You need to agree to share your contact information
+
+    # For my user the following has already been granted
+    # https://huggingface.co/meta-llama/Llama-3.2-3B
+
+    # consequently, a login is needed 
+
     """
 
     checkpoint = "meta-llama/Llama-3.2-3B-Instruct"
@@ -43,7 +51,10 @@ def generation_model():
 # ---------------------------------------------------------------------------------- -- #
 # ---------------------------------------------------------------------------------- -- #
 
-def ask_query(retrieved_texts, generation_query):
+def ask_query(retrieved_texts, generation_query, generation_model):
+    """
+
+    """
 
     # Prepare the messages for the text generation pipeline
     messages = [
@@ -63,14 +74,9 @@ def ask_query(retrieved_texts, generation_query):
     ]
 
     # Generate a response using the text generation pipeline
-    generator = generation_model()
-    response = generator(messages, max_new_tokens=256) #[-1]["generated_text"][-1]["content"]
-    # response = response[-1]['generated_text'][-1]['content']
-
-    #print(f"Query: \n\t{query}")
-    #print(f"Context: \n\t{retrieved_texts}")
-    #print(f"Answer: \n\t{response}")
-
+    # generator = generation_model()
+    response = generation_model(messages, max_new_tokens=128)
     r_answer = {'query': generation_query, 'context': retrieved_texts, 'response': response}
 
     return r_answer
+

python/molina/main.py

@@ -1,8 +1,35 @@
 import warnings
 warnings.filterwarnings("ignore", category=UserWarning)
 
-from content import documents, vector
-from generators import llama, gpt4all
+from content import vector
+from generators import llama
+
+def chat_with_user(chromadb, model_gen):
+    """
+    Chat interface for user interaction.
+    """
+
+    print("Welcome to the Chatbot! Type 'exit' to quit.")
+
+    while True:
+        context_query = input("\nEnter context query: ")
+        if context_query.lower() == 'exit':
+            break
+
+        generation_query = input("Enter your question: ")
+        if generation_query.lower() == 'exit':
+            break
+
+        # Retrieve relevant texts based on the context query
+        retrieved_texts = vector.search_vectordb(chromadb, context_query)
+
+        # Generate an answer based on the retrieved texts
+        llama_answer = llama.ask_query(retrieved_texts, generation_query, model_gen)
+        a_specific = llama_answer['response'][-1]['generated_text'][-1]['content']
+
+        # Output the results
+        print(f'\nProvided Context:\n{retrieved_texts}\n')
+        print(f'Agent Answer:\n{a_specific}\n')
 
 def main():
 
@@ -15,40 +42,25 @@ def main():
     db_path = "knowledge/chroma"
     docs_path = "knowledge/molina_compatible"
 
-    wd_folder = "/Users/franciscome/git/iteralabs/molina"
-    in_knowledge = "/knowledge"
-    in_topic = "/molina_compatible"
-    in_folder = wd_folder + in_knowledge + in_topic
-
     # -- -------------------------------------------- Extract/Load/Transform Content -- #
     # -- --------------------------------------------------------------------------- -- #
 
-    k_documents = documents.load_documents(data_path=docs_path)
-    ks_documents = documents.split_text(k_documents, c_size=1000, c_overlap=100)
-
-    # -- ---------------------------------------------------------- Create Documents -- #
-    # -- ---------------------------------------------------------- ---------------- -- #
-
-    chromadb = vector.create_vectordb(db_path=db_path,
-                           chunks=ks_documents, 
-                           embedding_model=gpt4all.embedding_model,
-                           verbose=True)
+    chromadb = vector.setup_vectordb(db_path=db_path, docs_path=docs_path)
 
     # -- ------------------------------------------------------ Create Embeddings DB -- #
     # -- --------------------------------------------------------------------------- -- #
+
+    model_gen = llama.generation_model()
 
-    context_query = "large language models, llm"
-
-    generation_query = "what is the perplexity of a Retrieval-based language model?"
-
-    retrieved_texts = vector.search_vectordb(chromadb, context_query)   
-    llama_answer = llama.ask_query(retrieved_texts, generation_query)
-    a_specific = llama_answer['response'][-1]['generated_text'][-1]['content']
-
-    # -- --------------------------------------------------------- Output Formatting -- #
-    # -- --------------------------------------------------------------------------- -- #
-
-    print(f'\n agent answer: ', a_specific, '\n')
+    """
+
+    context: augmented retrieval generation RAG retrieval-based LLM Language Models  
+    question: what are the major drawbacks of Retrival augmented generation RAG models ?
+
+    """
+
+    # Start chat with user
+    chat_with_user(chromadb, model_gen)
 
 if __name__ == "__main__":
     main()