feat (backend): added conditional llm logic

Author: itsskofficial

src/server/.env.selfhost.template

@@ -25,11 +25,11 @@ CELERY_BROKER_URL=redis://:${REDIS_PASSWORD}@redis:6379/0
 CELERY_RESULT_BACKEND=redis://:${REDIS_PASSWORD}@redis:6379/0
 
 # --- LLM Provider Configuration ---
-# Set LLM_PROVIDER to "NOVITA" to use the Novita API, or "OLLAMA" to use a local Ollama instance.
-# If using OLLAMA, ensure it's running and accessible from Docker.
-LLM_PROVIDER=NOVITA
-NOVITA_API_KEY=<your-novita-api-key>
-NOVITA_MODEL_NAME=qwen/qwen3-4b-fp8 # The model to use from Novita
+# Standard OpenAI API compatible endpoint configuration.
+# For local Ollama, use the following. The Docker container will run Ollama internally.
+OPENAI_API_BASE_URL=http://localhost:11434
+OPENAI_MODEL_NAME=qwen2:1.5b
+OPENAI_API_KEY=ollama
 
 # --- 3rd Party API Keys (Optional but recommended for full functionality) ---
 POSTHOG_API_KEY=<your_posthog_project_api_key>

src/server/.env.template

@@ -31,7 +31,12 @@ CELERY_RESULT_BACKEND="redis://:<url_encoded_password>@<host>:<port>/0"
 # --- 3rd Party APIs ---
 POSTHOG_API_KEY=<your_posthog_project_api_key>
 POSTHOG_HOST=<your_posthog_host> 
-NOVITA_API_KEY=<your-novita-api-key>
+# --- OpenAI Compatible LLM Configuration ---
+# You can point this to any OpenAI API compatible service, including OpenAI,
+# Groq, Together, or a self-hosted vLLM or Ollama instance.
+OPENAI_API_BASE_URL=https://api.openai.com/v1
+OPENAI_API_KEY=<your_openai_api_key>
+OPENAI_MODEL_NAME=gpt-4o-mini
 NEWS_API_KEY=<your-news-api-key>
 GOOGLE_CLIENT_ID=
 GOOGLE_PROJECT_ID=

src/server/Dockerfile

@@ -1,6 +1,11 @@
 # Use a slim Python base image
 FROM python:3.11-slim
 
+# --- Build Arguments ---
+# These arguments are passed from docker-compose.yaml or the `docker build` command.
+ARG OPENAI_API_KEY
+ARG OPENAI_MODEL_NAME
+
 # Set environment variables
 ENV PYTHONDONTWRITEBYTECODE=1 \
     PYTHONUNBUFFERED=1 \
@@ -9,6 +14,8 @@ ENV PYTHONDONTWRITEBYTECODE=1 \
     CHROME_VERSION=138.0.7204.157 \
     CHROMEDRIVER_VERSION=138.0.7204.157 \
     CHROME_BIN=/opt/chrome/chrome
+# Set Ollama to listen on all interfaces within the container
+ENV OLLAMA_HOST=0.0.0.0
 
 # Install system dependencies
 RUN apt-get update && apt-get install -y --no-install-recommends \
@@ -45,6 +52,18 @@ RUN wget -q -O /tmp/chromedriver.zip "https://storage.googleapis.com/chrome-for-
     && chmod +x /usr/local/bin/chromedriver \
     && rm -f /tmp/chromedriver.zip
 
+# --- Conditional Ollama Installation ---
+# This checks if the API key is set to 'ollama'. If so, install Ollama and pull the specified model.
+# This is checked during the 'docker build' process.
+RUN if [ "$OPENAI_API_KEY" = "ollama" ]; then \
+    echo "Ollama API key detected. Installing Ollama..."; \
+    curl -fsSL https://ollama.com/install.sh | sh; \
+    echo "Pulling model: ${OPENAI_MODEL_NAME}"; \
+    /usr/local/bin/ollama pull ${OPENAI_MODEL_NAME}; \
+    else \
+    echo "Ollama configuration not detected. Skipping installation."; \
+    fi
+
 # Create required directories
 RUN mkdir -p /etc/supervisor/conf.d /var/log/supervisor
 

src/server/docker-compose.yaml

@@ -3,66 +3,68 @@
 # This file defines how to run your application using Docker Compose.
 # It now orchestrates the main application, Redis, and the WAHA service.
 
-version: "3.8"
-
 services:
-  # 1. The Main Sentient Application (FastAPI, Nginx, Celery, MCPs)
-  sentient-app:
-    build:
-      context: .
-      dockerfile: Dockerfile
-    container_name: sentient-app
-    restart: unless-stopped
-    env_file:
-      - ./.env
-    ports:
-      # Exposes Nginx (running on port 80 in the container) to port 5000 on the host.
-      # This is the primary public-facing port for your entire application.
-      - "5000:80"
-    depends_on:
-      - redis
-      - waha
-    networks:
-      - sentient_network
+    # 1. The Main Sentient Application (FastAPI, Nginx, Celery, MCPs)
+    sentient-app:
+        build:
+            context: .
+            dockerfile: Dockerfile
+            # Pass variables from the .env file as build arguments for conditional builds
+            args:
+                - OPENAI_API_KEY=${OPENAI_API_KEY}
+                - OPENAI_MODEL_NAME=${OPENAI_MODEL_NAME}
+        container_name: sentient-app
+        restart: unless-stopped
+        env_file:
+            - ./.env
+        ports:
+            # Exposes Nginx (running on port 80 in the container) to port 5000 on the host.
+            # This is the primary public-facing port for your entire application.
+            - "5000:80"
+        depends_on:
+            - redis
+            - waha
+        networks:
+            - sentient_network
 
-  # 2. Redis for Celery Broker & Backend
-  redis:
-    image: redis:7-alpine
-    container_name: redis-db
-    restart: unless-stopped
-    # Use the password from the .env file, ensuring it's quoted
-    command: redis-server --requirepass "${REDIS_PASSWORD}"
-    volumes:
-      # Use a named volume for persistent Redis data
-      - redis-data:/data
-    networks:
-      - sentient_network
-    env_file:
-      - ./.env
-    # No ports are exposed to the host. Communication is internal and secure.
+    # 2. Redis for Celery Broker & Backend
+    redis:
+        image: redis:7-alpine
+        container_name: redis-db
+        restart: unless-stopped
+        # Use the password from the .env file, ensuring it's quoted
+        command: redis-server --requirepass "${REDIS_PASSWORD}"
+        volumes:
+            # Use a named volume for persistent Redis data
+            - redis-data:/data
+        networks:
+            - sentient_network
+        env_file:
+            - ./.env
+        # No ports are exposed to the host. Communication is internal and secure.
 
-  # 3. WAHA (WhatsApp HTTP API) Service
-  waha:
-    image: devlikeapro/waha:noweb
-    container_name: waha-notifier
-    restart: unless-stopped
-    volumes:
-      # Use a named volume for persistent WhatsApp session data
-      - waha-sessions:/app/.sessions
-    networks:
-      - sentient_network
-    env_file:
-      - ./.env
-    ports:
-    - "8000:3000"
-    # No ports are exposed to the host. The sentient-app communicates with it internally.
+    # 3. WAHA (WhatsApp HTTP API) Service
+    waha:
+        image: devlikeapro/waha:noweb
+        container_name: waha-notifier
+        restart: unless-stopped
+        volumes:
+            # Use a named volume for persistent WhatsApp session data
+            - waha-sessions:/app/.sessions
+        networks:
+            - sentient_network
+        env_file:
+            - ./.env
+        ports:
+            - "8000:3000"
+        # No ports are exposed to the host. The sentient-app communicates with it internally.
 
 # Defines the shared network for all services
 networks:
-  sentient_network:
-    driver: bridge
+    sentient_network:
+        driver: bridge
 
 # Defines the named volumes for data persistence
 volumes:
-  redis-data:
-  waha-sessions:
+    redis-data:
+    waha-sessions:

src/server/main/config.py

@@ -315,31 +315,19 @@
     }
 }
 
-# --- Service Provider Configuration ---
-# These variables allow for flexible switching between different service providers.
-LLM_PROVIDER = os.getenv("LLM_PROVIDER", "OLLAMA") # Options: "OLLAMA", "NOVITA"
-# STT_PROVIDER = os.getenv("STT_PROVIDER", "FASTER_WHISPER") # Voice removed
-# TTS_PROVIDER = os.getenv("TTS_PROVIDER", "ORPHEUS") # Voice removed
-
-# --- Service-Specific API Keys and Paths ---
-# ELEVENLABS_API_KEY = os.getenv("ELEVENLABS_API_KEY") # Voice removed
-# ELEVENLABS_VOICE_ID = os.getenv("ELEVENLABS_VOICE_ID", "JBFqnCBsd6RMkjVDRZzb") # Voice removed
-
-# Voice related model paths and settings removed
-
-
 # LLM Endpoint Configuration
-OLLAMA_BASE_URL = os.getenv("OLLAMA_BASE_URL", "http://localhost:11434") 
-OLLAMA_MODEL_NAME = os.getenv("OLLAMA_MODEL_NAME", "qwen3:4b") 
-NOVITA_API_KEY = os.getenv("NOVITA_API_KEY")
-NOVITA_MODEL_NAME = os.getenv("NOVITA_MODEL_NAME", "qwen/qwen3-4b-fp8")
+# --- OpenAI API Standard Configuration ---
+# This can point to OpenAI, Ollama, Groq, or any other compatible service.
+OPENAI_API_BASE_URL = os.getenv("OPENAI_API_BASE_URL", "http://localhost:11434")
+OPENAI_MODEL_NAME = os.getenv("OPENAI_MODEL_NAME", "qwen2:1.5b")
+OPENAI_API_KEY = os.getenv("OPENAI_API_KEY", "ollama") # Default key for Ollama
+
 # MCP Server URLs
 PROGRESS_UPDATER_MCP_SERVER_URL=os.getenv("PROGRESS_UPDATER_MCP_SERVER_URL", "http://localhost:9011/sse")
 CHAT_TOOLS_MCP_SERVER_URL=os.getenv("CHAT_TOOLS_MCP_SERVER_URL", "http://localhost:9013/sse") # For agent action handoff
 SUPERMEMORY_MCP_BASE_URL = os.getenv("SUPERMEMORY_MCP_BASE_URL", "https://mcp.supermemory.ai/")
 SUPERMEMORY_MCP_ENDPOINT_SUFFIX = os.getenv("SUPERMEMORY_MCP_ENDPOINT_SUFFIX", "/sse")
 
 print(f"[{datetime.datetime.now()}] [MainServer_Config] Configuration loaded. AUTH0_DOMAIN: {'SET' if AUTH0_DOMAIN else 'NOT SET'}")
-print(f"[{datetime.datetime.now()}] [MainServer_Config] LLM Provider: {LLM_PROVIDER}")
-# print(f"[{datetime.datetime.now()}] [MainServer_Config] STT Provider: {STT_PROVIDER}") # Voice removed
-# print(f"[{datetime.datetime.now()}] [MainServer_Config] TTS Provider: {TTS_PROVIDER}") # Voice removed
+print(f"[{datetime.datetime.now()}] [MainServer_Config] LLM Endpoint: {OPENAI_API_BASE_URL}")
+print(f"[{datetime.datetime.now()}] [MainServer_Config] LLM Model: {OPENAI_MODEL_NAME}")

src/server/main/llm.py

@@ -4,11 +4,8 @@
 from qwen_agent.agents import Assistant
 from qwen_agent.llm import get_chat_model
 
-from main.config import (
-    LLM_PROVIDER,
-    OLLAMA_MODEL_NAME, OLLAMA_BASE_URL,
-    NOVITA_MODEL_NAME, NOVITA_API_KEY
-)
+from main.config import (OPENAI_API_BASE_URL, OPENAI_API_KEY,
+                         OPENAI_MODEL_NAME)
 
 logger = logging.getLogger(__name__)
 
@@ -18,42 +15,20 @@ def get_qwen_assistant(system_message: str = DEFAULT_SYSTEM_PROMPT, function_lis
     """
     Initializes and returns a Qwen Assistant agent configured for the current environment.
     """
-    llm_cfg = {}
-    if LLM_PROVIDER == "OLLAMA":
-        # Ollama configuration for Qwen Agent (expects OpenAI-compatible v1 endpoint)
-        # Ensure OLLAMA_BASE_URL is like "http://localhost:11434"
-        ollama_v1_url = f"{OLLAMA_BASE_URL.rstrip('/')}/v1"
-        llm_cfg = {
-            'model': OLLAMA_MODEL_NAME,  # e.g., "llama3.2:3b"
-            'model_server': ollama_v1_url,
-            'api_key': 'ollama',  # Placeholder, Ollama doesn't typically require a key via API
-        }
-        logger.info(f"Qwen Agent configured for LLM_PROVIDER='OLLAMA': model={OLLAMA_MODEL_NAME}, server={ollama_v1_url}")
-    elif LLM_PROVIDER == "NOVITA":
-        # Novita configuration for Qwen Agent
-        novita_v1_url = "https://api.novita.ai/v3/openai"
-        llm_cfg = {
-            'model': NOVITA_MODEL_NAME,
-            'model_server': novita_v1_url,
-            'api_key': NOVITA_API_KEY,
-        }
-        # For Novita, some models might need specific routing prefixes if not handled by model_server directly
-        # e.g. 'openrouter/meta-llama/llama-3.1-8b-instruct'
-        # However, Qwen's OpenAI client usually takes the model name as passed.
-        # The 'model_server' determines the endpoint.
-        logger.info(f"Qwen Agent configured for LLM_PROVIDER='NOVITA': model={NOVITA_MODEL_NAME}, server={novita_v1_url}")
-    else:
-        logger.error(f"Invalid LLM_PROVIDER: '{LLM_PROVIDER}'. Must be 'OLLAMA' or 'NOVITA'.")
-        raise ValueError(f"Invalid LLM_PROVIDER configured: {LLM_PROVIDER}")
-
-    if not llm_cfg.get('model'):
+    # Qwen-agent's `Assistant` uses an OpenAI-compatible interface.
+    # We map our standard environment variables to its expected config format.
+    # Note: `model_server` for qwen-agent is the `base_url`.
+    llm_cfg = {
+        'model': OPENAI_MODEL_NAME,
+        'model_server': f"{OPENAI_API_BASE_URL.rstrip('/')}/v1",
+        'api_key': OPENAI_API_KEY,
+    }
+    logger.info(f"Qwen Agent configured with model='{OPENAI_MODEL_NAME}' and server='{llm_cfg['model_server']}'")
+    if not OPENAI_MODEL_NAME:
         logger.error("LLM model name is not configured. Qwen Agent cannot be initialized.")
         raise ValueError("LLM model configuration error.")
 
     try:
-        # Initialize the LLM part of the agent
-        # llm_instance = get_chat_model(llm_cfg) # Can also pass llm_cfg directly to Assistant
-
         # Initialize the Assistant agent
         bot = Assistant(
             llm=llm_cfg,

src/server/supervisord.conf

@@ -3,6 +3,17 @@
 [supervisord]
 nodaemon=true
 user=root
+environment=DISPLAY=":99"
+
+[program:xvfb]
+command=/usr/bin/Xvfb :99 -screen 0 1024x768x24
+priority=1
+autostart=true
+autorestart=true
+stdout_logfile=/dev/stdout
+stdout_logfile_maxbytes=0
+stderr_logfile=/dev/stderr
+stderr_logfile_maxbytes=0
 
 [program:nginx]
 command=/usr/sbin/nginx -g "daemon off;"
@@ -14,6 +25,17 @@ stdout_logfile_maxbytes=0
 stderr_logfile=/dev/stderr
 stderr_logfile_maxbytes=0
 
+[program:ollama]
+command=/bin/sh -c "if [ -f /usr/local/bin/ollama ]; then /usr/local/bin/ollama serve; else echo Ollama not installed, skipping... && tail -f /dev/null; fi"
+directory=/app
+autostart=true
+autorestart=true
+priority=5
+stdout_logfile=/dev/stdout
+stdout_logfile_maxbytes=0
+stderr_logfile=/dev/stderr
+stderr_logfile_maxbytes=0
+
 [program:main-server]
 command=uvicorn main.app:app --host 0.0.0.0 --port 5000 --lifespan on
 directory=/app
@@ -235,18 +257,3 @@ stdout_logfile=/dev/stdout
 stdout_logfile_maxbytes=0
 stderr_logfile=/dev/stderr
 stderr_logfile_maxbytes=0
-
-[supervisord]
-nodaemon=true
-user=root
-environment=DISPLAY=":99"
-
-[program:xvfb]
-command=/usr/bin/Xvfb :99 -screen 0 1024x768x24
-priority=1
-autostart=true
-autorestart=true
-stdout_logfile=/dev/stdout
-stdout_logfile_maxbytes=0
-stderr_logfile=/dev/stderr
-stderr_logfile_maxbytes=0

src/server/workers/executor/config.py

@@ -13,12 +13,10 @@
 MONGO_URI = os.getenv("MONGO_URI", "mongodb://localhost:27017/")
 MONGO_DB_NAME = os.getenv("MONGO_DB_NAME", "sentient_dev_db")
 
-LLM_PROVIDER = os.getenv("LLM_PROVIDER", "OLLAMA")
-OLLAMA_BASE_URL = os.getenv("OLLAMA_BASE_URL", "http://localhost:11434")
-OLLAMA_MODEL_NAME = os.getenv("OLLAMA_MODEL_NAME", "qwen3:4b")
-NOVITA_API_KEY = os.getenv("NOVITA_API_KEY")
-NOVITA_MODEL_NAME = os.getenv("NOVITA_MODEL_NAME", "qwen/qwen3-4b-fp8")
-
+# OpenAI API Standard Configuration
+OPENAI_API_BASE_URL = os.getenv("OPENAI_API_BASE_URL", "http://localhost:11434")
+OPENAI_MODEL_NAME = os.getenv("OPENAI_MODEL_NAME", "qwen2:1.5b")
+OPENAI_API_KEY = os.getenv("OPENAI_API_KEY", "ollama")
 SUPERMEMORY_MCP_BASE_URL = os.getenv("SUPERMEMORY_MCP_BASE_URL", "https://mcp.supermemory.ai/")
 SUPERMEMORY_MCP_ENDPOINT_SUFFIX = os.getenv("SUPERMEMORY_MCP_ENDPOINT_SUFFIX", "/sse")