integrate feedback, add raw message conversation method

Author: edkaya

genai/config.py

@@ -10,12 +10,14 @@
     [
         "api_key_openai",
         "waitress",
-        "api_openwebui"
+        "api_openwebui",
+        "base_url"
     ],
 )
 
 Config = ConfigT(
     api_key_openai=environ.get("API_SECRET_OPENAI_MINE"),
     waitress=environ.get("USE_WAITRESS", "false").lower() == "true",
-    api_openwebui=environ.get("API_OPENWEBUI")
+    api_openwebui=environ.get("API_OPENWEBUI"),
+    base_url=environ.get("BASE_URL")
 )

genai/controller/generate_controller.py

@@ -6,7 +6,7 @@
 from genai.rag.ingestion_pipeline import IngestionPipeline
 from genai.vector_database.qdrant_vdb import QdrantVDB
 from genai.rag.llm.chat_model import ChatModel
-from genai.service.rag_service import retrieve_similar_docs, prepare_prompt
+from genai.service.rag_service import retrieve_similar_docs, prepare_prompt, process_raw_messages
 
 
 # Set Logging
@@ -78,25 +78,33 @@ def upload_file():
 
 @generate_bp.route('/genai/generate', methods=['POST'])
 def generate():
-    """API Endpoint for generating recipe response based on document retrieval
+    """
+    API Endpoint for generating recipe responses using retrieved context.
 
-    This endpoint processes user queries against a vector database of recipes
-    and returns AI-generated responses using retrieved context.
+    This endpoint processes a user query against a vector database of recipes
+    and returns an AI-generated response using both retrieved context and
+    the full conversation history provided in the request.
 
     Request Body:
         query (str): The user's recipe-related query
-        conversation_id (str): Unique identifier for the conversation thread
+        messages (List[Dict]): Full conversation history, each with 'role' and 'content'
+            Example:
+            [
+                {"role": "user", "content": "I have eggs and tomatoes."},
+                {"role": "assistant", "content": "You could make shakshuka."}
+            ]
 
     Returns:
-        JSON response containing the generated recipe response or error message
+        JSON response containing:
+            - 'response': The generated assistant reply
     """
     data = request.get_json()
 
-    if not data or "query" not in data or "conversation_id" not in data:
-        return jsonify({"error": "Missing 'query' or 'conversation_id'"}), 400
+    if not data or "query" not in data or "messages" not in data:
+        return jsonify({"error": "Missing 'query' or 'messages'"}), 400
 
     query = data["query"]
-    # conversation_id = data["conversation_id"] # will be used in the future
+    messages_raw = data["messages"]
 
     try:
         collection_name = "recipes"
@@ -106,8 +114,8 @@ def generate():
             vector_store = qdrant.create_and_get_vector_storage(
                 collection_name
             )
-            # todo: retrieve messages from chat history as BaseMessage
-            messages = []
+            # turn raw message into BaseMessage type
+            messages = process_raw_messages(messages_raw)
             retrieved_docs = retrieve_similar_docs(vector_store, query)
             prompt = prepare_prompt(
                 llm.get_system_prompt(),

genai/rag/llm/chat_model.py

@@ -14,7 +14,10 @@ def _generate(self, messages: List[BaseMessage],
                   stop=None,
                   **kwargs) -> ChatResult:
         prompt = "\n".join([
-            msg.content for msg in messages if isinstance(msg, HumanMessage)
+            f"User: {m.content}" if isinstance(m, HumanMessage)
+            else f"Assistant: {m.content}" if isinstance(m, AIMessage)
+            else ""
+            for m in messages
             ])
         response_text = generate_response(self.model_name, prompt)
 

genai/service/openwebui_service.py

@@ -2,7 +2,7 @@
 
 from genai.config import Config
 
-BASE_URL = "https://gpu.aet.cit.tum.de/"
+BASE_URL = Config.base_url
 
 
 def generate_response(model_name: str, prompt: str):

genai/service/rag_service.py

@@ -1,7 +1,7 @@
-from typing import List
+from typing import List, Dict
 
 from langchain_qdrant import QdrantVectorStore
-from langchain_core.messages import BaseMessage, HumanMessage
+from langchain_core.messages import BaseMessage, HumanMessage, AIMessage
 from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder
 # from genai.rag.llm.chat_model import ChatModel
 
@@ -33,6 +33,21 @@ def prepare_prompt(system_prompt: str,
 
     return prompt
 
+def process_raw_messages(raw_messages: List[Dict]) -> List[BaseMessage]:
+    """Turns raw messages into BaseMessages, so they can be passed into LLM"""
+    processed_messages = []
+    for msg in raw_messages:
+        role = msg.get("role")
+        content = msg.get("content")
+
+        if role == "user":
+            processed_messages.append(HumanMessage(content=content))
+
+        elif role == "assistant":
+            processed_messages.append(AIMessage(content=content))
+
+    return processed_messages
+
 # For testing purposes
 # if __name__ == "__main__":
 #     msg = HumanMessage(content="My name is John Doe.")