Merge pull request #39 from carlosmada22/15-web-app

Added web interface

Author: carlosmada22

desi_vectordb/chroma.sqlite3

@@ -154,9 +154,41 @@ def run_query_interface(config: DesiConfig) -> None:
 
 
 def run_web_interface(config: DesiConfig) -> None:
-    """Run the web interface (placeholder - not implemented yet)."""
-    logger.info("🌐 Web interface not implemented yet, starting CLI interface...")
-    run_query_interface(config)
+    """Run the web interface using FastAPI and Gradio."""
+    logger.info("🌐 Starting web interface...")
+    try:
+        import uvicorn
+
+        # Update the global API base URL in the app module
+        import desi.web.app as app_module
+
+        app_module.api_base_url = f"http://{config.web_host}:{config.web_port}"
+
+        logger.info(
+            f"Web interface will be available at: http://{config.web_host}:{config.web_port}"
+        )
+        logger.info("Press Ctrl+C to stop the server")
+
+        # Run the server
+        uvicorn.run(
+            "desi.web.app:app",
+            host=config.web_host,
+            port=config.web_port,
+            reload=config.web_debug,
+            log_level="info",
+        )
+
+    except ImportError as e:
+        logger.error(f"Failed to import required packages: {e}")
+        logger.error(
+            "Please install required packages: pip install fastapi uvicorn gradio"
+        )
+        logger.info("Falling back to CLI interface...")
+        run_query_interface(config)
+    except Exception as e:
+        logger.error(f"Failed to start web interface: {e}")
+        logger.info("Falling back to CLI interface...")
+        run_query_interface(config)
 
 
 def main():
@@ -242,8 +274,7 @@ def main():
 
     # Step 4: Start the interface
     if args.web:
-        logger.info("🌐 Web interface not implemented yet, starting CLI interface...")
-        run_query_interface(config)
+        run_web_interface(config)
     else:
         run_query_interface(config)
 

main.py

@@ -38,8 +38,9 @@ dependencies = [
     "PyYAML",
     "requests",
     "scikit-learn",
-    "flask",
-    "flask-cors",
+    "fastapi",
+    "uvicorn[standard]",
+    "gradio",
     "chromadb",
 ]
 

pyproject.toml

@@ -8,6 +8,7 @@
 """
 
 import logging
+import re
 import sqlite3
 import uuid
 from typing import Dict, List, Tuple, TypedDict
@@ -153,27 +154,28 @@ def rewrite_query_node(self, state: GraphState) -> Dict:
 
         history_str = "\n".join([f"{msg['role']}: {msg['content']}" for msg in history])
 
-        rewrite_prompt = f"""
-Given the following conversation history and a user's follow-up question, rewrite the user's question to be a standalone question that can be understood without the context of the chat history.
-
-This standalone question will be used to search a knowledge base.
-
-If the user's question is already a good standalone question, simply return it as is.
+        rewrite_prompt = f"""Based on the chat history below, rewrite the user's follow-up question into a single, self-contained question ideal for a vector search. DO NOT add any commentary or explanation.
 
 <Conversation History>
 {history_str}
 </Conversation History>
 
 User's Follow-up Question: "{user_query}"
 
-Standalone Question:
-"""
+Standalone Question:"""
+
+        # Get the raw output from the LLM
+        raw_output = self.rewrite_llm.invoke(rewrite_prompt)
+
+        # Clean the output to remove any "thought" blocks
+        cleaned_query = re.sub(
+            r"<think>.*?</think>", "", str(raw_output.content), flags=re.DOTALL
+        ).strip()
 
-        rewritten_query = self.rewrite_llm.invoke(rewrite_prompt).content
         logger.info(
-            f"Original query: '{user_query}' -> Rewritten query: '{rewritten_query}'"
+            f"Original query: '{user_query}' -> Rewritten query: '{cleaned_query}'"
         )
-        return {"rewritten_query": rewritten_query}
+        return {"rewritten_query": cleaned_query}
 
     # --- Node Definitions for the Graph ---
 
@@ -325,9 +327,9 @@ def start_chat_session(self):
     SQLITE_DB_PATH = "./data/conversation_memory.db"
     CONVERSATION_HISTORY_LIMIT = 20
     # Value for boosting dswiki chunks
-    DSWIKI_BOOST_VALUE = 0.2
+    DSWIKI_BOOST_VALUE = 0.1
     # A score of 0.7 means we discard any chunk with less than 70% similarity.
-    RELEVANCE_THRESHOLD = 0.7
+    RELEVANCE_THRESHOLD = 0.4
     # The model used by your RAG engine
     LLM_MODEL = "qwen3"
 

src/desi/query/conversation_engine.py

@@ -287,9 +287,9 @@ def query(
     # --- Standalone Execution Example ---
     CHROMA_PERSIST_DIRECTORY = "./desi_vectordb"
     # Value for boosting dswiki chunks
-    DSWIKI_BOOST_VALUE = 0.15
+    DSWIKI_BOOST_VALUE = 0.1
     # A score of 0.3 means we discard any chunk with less than 30% similarity.
-    RELEVANCE_THRESHOLD = 0.7
+    RELEVANCE_THRESHOLD = 0.4
     # Path to the prompt template
     PROMPT_TEMPLATE_PATH = "./prompts/desi_query_prompt.md"