Deploy Sentinel AI 2026-04-06_16:26:49

alphavantage_mcp.py

@@ -35,7 +35,12 @@ else:
 
 # --- FastAPI App & Alpha Vantage Client ---
 app = FastAPI(title="Aegis Alpha Vantage MCP Server")
-ts = TimeSeries(key=ALPHA_VANTAGE_API_KEY, output_format='json')
+ts = None
+if ALPHA_VANTAGE_API_KEY:
+    try:
+        ts = TimeSeries(key=ALPHA_VANTAGE_API_KEY, output_format='json')
+    except Exception as e:
+        logger.error(f"Failed to initialize Alpha Vantage TimeSeries: {e}")
 
 @app.post("/market_data")
 async def get_market_data(payload: dict):
@@ -58,6 +63,9 @@ async def get_market_data(payload: dict):
     logger.info(f"Received market data request for symbol: {symbol}, time_range: {time_range}")
 
     try:
+        if not ts:
+            raise ValueError("TimeSeries not initialized (missing API key).")
+            
         # Route to appropriate API based on time range
         if time_range == "INTRADAY":
             # Intraday data (last 4-6 hours, 5-min intervals)
@@ -220,6 +228,9 @@ async def get_company_overview(payload: dict):
     logger.info(f"Fetching company overview for {symbol}")
 
     try:
+        if not ALPHA_VANTAGE_API_KEY:
+            raise ValueError("Alpha Vantage API Key is missing.")
+
         url = "https://www.alphavantage.co/query"
         params = {
             "function": "OVERVIEW",
@@ -349,6 +360,9 @@ async def get_global_quote(payload: dict):
     logger.info(f"Fetching global quote for {symbol}")
 
     try:
+        if not ALPHA_VANTAGE_API_KEY:
+            raise ValueError("Alpha Vantage API Key is missing.")
+
         url = "https://www.alphavantage.co/query"
         params = {
             "function": "GLOBAL_QUOTE",

mcp_gateway.py

@@ -17,14 +17,26 @@ logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(name)s - %(level
 logger = logging.getLogger("MCP_Gateway")
 
 # --- Import Microservices for Consolidation ---
+tavily_app = None
 try:
     from tavily_mcp import app as tavily_app
+    logger.info("Successfully imported Tavily microservice.")
+except Exception as e:
+    logger.error(f"Failed to import Tavily microservice: {e}")
+
+alphavantage_app = None
+try:
     from alphavantage_mcp import app as alphavantage_app
+    logger.info("Successfully imported AlphaVantage microservice.")
+except Exception as e:
+    logger.error(f"Failed to import AlphaVantage microservice: {e}")
+
+private_app = None
+try:
     from private_mcp import app as private_app
-    logger.info("Successfully imported microservices for consolidation.")
-except ImportError as e:
-    logger.critical(f"Failed to import microservices: {e}")
-    raise
+    logger.info("Successfully imported Private microservice.")
+except Exception as e:
+    logger.error(f"Failed to import Private microservice: {e}")
 
 # --- Configuration (Updated for Monolithic Mode) ---
 # Default to internal mounted paths on the same port (8000)
@@ -45,9 +57,12 @@ app.add_middleware(
 )
 
 # --- Mount Microservices ---
-app.mount("/tavily", tavily_app)
-app.mount("/alphavantage", alphavantage_app)
-app.mount("/private", private_app)
+if tavily_app:
+    app.mount("/tavily", tavily_app)
+if alphavantage_app:
+    app.mount("/alphavantage", alphavantage_app)
+if private_app:
+    app.mount("/private", private_app)
 
 client = httpx.AsyncClient()
 

private_mcp.py

@@ -3,9 +3,11 @@ from fastapi import FastAPI, HTTPException
 import uvicorn
 import sqlite3
 import logging
-from langchain_core.prompts import ChatPromptTemplate
-from langchain_core.output_parsers import StrOutputParser
-from langchain_ollama import ChatOllama
+import os
+from dotenv import load_dotenv
+
+# --- Configuration ---
+load_dotenv()
 
 # --- Logging Setup ---
 logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(name)s - %(levelname)s - %(message)s')
@@ -14,26 +16,47 @@ logger = logging.getLogger("Private_MCP_Server")
 # --- Database Configuration ---
 DB_FILE = "portfolio.db"
 
-# --- LLM Configuration (Local Llama 3) ---
-# This connects to the Ollama application running on your machine.
-# Make sure Ollama and the llama3 model are running.
-llm = ChatOllama(model="llama3", temperature=0)
-
-# --- Text-to-SQL Prompt Engineering ---
-# This prompt is carefully designed to make Llama 3 generate ONLY safe SQL queries.
-text_to_sql_prompt = ChatPromptTemplate.from_messages([
-    ("system", 
-     """You are a Text-to-SQL assistant. Convert the question to a read-only SQLite query for the 'holdings' table.
+# --- LLM Configuration (Google Gemini) ---
+# For cloud deployment on HF Spaces, we use Gemini instead of local Llama.
+sql_generation_chain = None
+try:
+    from langchain_google_genai import ChatGoogleGenerativeAI
+    
+    GOOGLE_API_KEY = os.getenv("GOOGLE_API_KEY")
+    if not GOOGLE_API_KEY:
+        try:
+            import toml
+            secrets_path = os.path.join(os.path.dirname(__file__), ".streamlit", "secrets.toml")
+            if os.path.exists(secrets_path):
+                secrets = toml.load(secrets_path)
+                GOOGLE_API_KEY = secrets.get("GOOGLE_API_KEY")
+        except Exception:
+            pass
+
+    if GOOGLE_API_KEY:
+        from langchain_core.prompts import ChatPromptTemplate
+        from langchain_core.output_parsers import StrOutputParser
+
+        llm = ChatGoogleGenerativeAI(model="gemini-2.5-flash", temperature=0, google_api_key=GOOGLE_API_KEY)
+        
+        text_to_sql_prompt = ChatPromptTemplate.from_messages([
+            ("system", 
+             """You are a Text-to-SQL assistant. Convert the question to a read-only SQLite query for the 'holdings' table.
 Schema: symbol (TEXT), shares (INTEGER), average_cost (REAL).
 RULES:
 1. SELECT only. No INSERT/UPDATE/DELETE.
 2. Output ONLY the SQL query. No markdown.
 """),
-    ("human", "Question: {question}")
-])
+            ("human", "Question: {question}")
+        ])
+        
+        sql_generation_chain = text_to_sql_prompt | llm | StrOutputParser()
+        logger.info("Successfully initialized Gemini LLM for Text-to-SQL.")
+    else:
+        logger.warning("GOOGLE_API_KEY not found. Private MCP will use regex fallback.")
 
-# Create the LangChain chain for Text-to-SQL
-sql_generation_chain = text_to_sql_prompt | llm | StrOutputParser()
+except Exception as e:
+    logger.error(f"Failed to initialize Gemini logic: {e}. Will use regex fallback.")
 
 # --- FastAPI App ---
 app = FastAPI(title="Aegis Private MCP Server")
@@ -127,8 +150,10 @@ async def get_portfolio_data(payload: dict):
     try:
         # Step 1: Generate the SQL query using the local LLM
         try:
+            if sql_generation_chain is None:
+                raise ValueError("sql_generation_chain not initialized (missing API key or imports).")
             generated_sql = await sql_generation_chain.ainvoke({"question": question})
-            logger.info(f"Llama 3 generated SQL: {generated_sql}")
+            logger.info(f"Gemini generated SQL: {generated_sql}")
         except Exception as llm_error:
             logger.warning(f"LLM generation failed (likely Ollama offline): {llm_error}. Using fallback logic.")
             # Fallback Logic: Dynamic symbol extraction

requirements.txt

@@ -10,8 +10,8 @@ httpx
 alpha_vantage
 fastapi
 uvicorn[standard]
-tavily
-langchain_ollama
+tavily-python
+langchain-ollama
 langchain-google-genai
 fpdf2
 pdfplumber
@@ -21,3 +21,4 @@ fredapi
 jinja2
 matplotlib
 requests
+toml

tavily_mcp.py

@@ -35,7 +35,12 @@ else:
 
 # --- FastAPI App & Tavily Client ---
 app = FastAPI(title="Aegis Tavily MCP Server")
-tavily = TavilyClient(api_key=TAVILY_API_KEY)
+tavily = None
+if TAVILY_API_KEY:
+    try:
+        tavily = TavilyClient(api_key=TAVILY_API_KEY)
+    except Exception as e:
+        logger.error(f"Failed to initialize TavilyClient: {e}")
 
 @app.post("/research")
 async def perform_research(payload: dict):
@@ -59,6 +64,9 @@ async def perform_research(payload: dict):
     # --- THIS IS THE CORRECTED LOGIC ---
     all_results = []
     try:
+        if not tavily:
+            raise ValueError("TavilyClient not initialized (missing API key).")
+            
         # Loop through each query and perform a search
         for query in queries:
             logger.info(f"Performing search for query: '{query}'")