"""CodeGPT Backend — Flask API server. Supports Groq (cloud) and Ollama (local). Deploy to Render/Railway/PythonAnywhere for 24/7 access without your laptop. Usage: python server.py # Starts on port 5050 python server.py --port 8080 # Custom port GROQ_API_KEY=gsk_... python server.py # Use Groq cloud backend Endpoints: POST /chat — Send messages, get AI response (streaming) POST /chat/quick — Non-streaming single response GET /models — List available models GET /health — Health check GET /config — Get server config (model, provider) """ import json import os import sys import time from datetime import datetime from flask import Flask, request, jsonify, Response, stream_with_context # Optional: Groq for cloud inference try: from groq import Groq HAS_GROQ = True except ImportError: HAS_GROQ = False # Optional: requests for Ollama import requests as http_requests app = Flask(__name__) # --- Config --- GROQ_API_KEY = os.environ.get("GROQ_API_KEY", "") OLLAMA_URL = os.environ.get("OLLAMA_URL", "http://localhost:11434") DEFAULT_MODEL = os.environ.get("CODEGPT_MODEL", "llama-3.2-3b-preview") # Groq model OLLAMA_MODEL = os.environ.get("OLLAMA_MODEL", "llama3.2") PORT = int(os.environ.get("PORT", 5050)) # Auto-detect provider if GROQ_API_KEY and HAS_GROQ: PROVIDER = "groq" else: PROVIDER = "ollama" SYSTEM_PROMPT = """You are an AI modeled after a highly technical, system-focused developer mindset. Be direct, concise, and dense with information. No fluff, no filler, no emojis. Give conclusions first, then minimal necessary explanation. Focus on: AI, coding, automation, cybersecurity, system design. Blunt but intelligent. Slightly dark tone is acceptable. Keep responses concise for mobile reading.""" PERSONAS = { "Default": SYSTEM_PROMPT, "Hacker": "You are a cybersecurity expert. Technical jargon, CVEs, attack vectors. Defensive security only. Concise.", "Teacher": "You are a patient programming teacher. Step by step, analogies, examples. Concise.", "Roast": "You are a brutally sarcastic code reviewer. Roast then fix. Dark humor. Concise.", "Minimal": "Shortest possible answer. One line if possible. Code only.", } # Stats server_stats = {"requests": 0, "start": time.time()} # --- Groq Backend --- def query_groq(messages, model, system, stream=False): """Query Groq cloud API.""" client = Groq(api_key=GROQ_API_KEY) full_messages = [{"role": "system", "content": system}] full_messages.extend(messages) if stream: completion = client.chat.completions.create( model=model or DEFAULT_MODEL, messages=full_messages, stream=True, max_tokens=4096, ) def generate(): full = [] for chunk in completion: delta = chunk.choices[0].delta if delta and delta.content: full.append(delta.content) yield json.dumps({ "content": delta.content, "done": False, }) + "\n" yield json.dumps({ "content": "", "done": True, "full_response": "".join(full), "model": model or DEFAULT_MODEL, "provider": "groq", }) + "\n" return generate() else: completion = client.chat.completions.create( model=model or DEFAULT_MODEL, messages=full_messages, max_tokens=4096, ) content = completion.choices[0].message.content usage = completion.usage return { "content": content, "model": model or DEFAULT_MODEL, "provider": "groq", "tokens": { "input": usage.prompt_tokens if usage else 0, "output": usage.completion_tokens if usage else 0, }, } # --- Ollama Backend --- def query_ollama(messages, model, system, stream=False): """Query local Ollama.""" ollama_messages = [{"role": "system", "content": system}] ollama_messages.extend(messages) if stream: try: response = http_requests.post( f"{OLLAMA_URL}/api/chat", json={"model": model or OLLAMA_MODEL, "messages": ollama_messages, "stream": True}, stream=True, timeout=120, ) response.raise_for_status() except Exception as e: def error_gen(): yield json.dumps({"content": f"Error: {e}", "done": True}) + "\n" return error_gen() def generate(): for line in response.iter_lines(): if not line: continue try: chunk = json.loads(line) except json.JSONDecodeError: continue content = chunk.get("message", {}).get("content", "") done = chunk.get("done", False) out = {"content": content, "done": done} if done: ec = chunk.get("eval_count", 0) td = chunk.get("total_duration", 0) ds = td / 1e9 if td else 0 tps = ec / ds if ds > 0 else 0 out["model"] = model or OLLAMA_MODEL out["provider"] = "ollama" out["tokens"] = {"output": ec} out["stats"] = f"{ec} tok | {ds:.1f}s | {tps:.0f} tok/s" yield json.dumps(out) + "\n" return generate() else: response = http_requests.post( f"{OLLAMA_URL}/api/chat", json={"model": model or OLLAMA_MODEL, "messages": ollama_messages, "stream": False}, timeout=120, ) response.raise_for_status() data = response.json() content = data.get("message", {}).get("content", "") ec = data.get("eval_count", 0) td = data.get("total_duration", 0) ds = td / 1e9 if td else 0 return { "content": content, "model": model or OLLAMA_MODEL, "provider": "ollama", "tokens": {"output": ec}, "stats": f"{ec} tok | {ds:.1f}s", } # --- Routes --- @app.route("/health", methods=["GET"]) def health(): uptime = int(time.time() - server_stats["start"]) return jsonify({ "status": "ok", "provider": PROVIDER, "model": DEFAULT_MODEL if PROVIDER == "groq" else OLLAMA_MODEL, "uptime": uptime, "requests": server_stats["requests"], }) @app.route("/config", methods=["GET"]) def get_config(): return jsonify({ "provider": PROVIDER, "model": DEFAULT_MODEL if PROVIDER == "groq" else OLLAMA_MODEL, "personas": list(PERSONAS.keys()), }) @app.route("/models", methods=["GET"]) def list_models(): if PROVIDER == "groq": # Groq available models models = [ "llama-3.2-3b-preview", "llama-3.2-1b-preview", "llama-3.3-70b-versatile", "llama-3.1-8b-instant", "gemma2-9b-it", "mixtral-8x7b-32768", ] else: try: resp = http_requests.get(f"{OLLAMA_URL}/api/tags", timeout=5) models = [m["name"] for m in resp.json().get("models", [])] except Exception: models = [] return jsonify({"models": models, "provider": PROVIDER}) @app.route("/chat", methods=["POST"]) def chat_stream(): """Streaming chat endpoint.""" server_stats["requests"] += 1 data = request.get_json() if not data: return jsonify({"error": "Invalid JSON"}), 400 messages = data.get("messages", []) model = data.get("model", "") persona = data.get("persona", "Default") system = PERSONAS.get(persona, SYSTEM_PROMPT) if not messages: return jsonify({"error": "No messages"}), 400 # Input validation: cap message count and content length if len(messages) > 100: messages = messages[-100:] for msg in messages: if not isinstance(msg, dict) or "role" not in msg or "content" not in msg: return jsonify({"error": "Invalid message format"}), 400 if len(msg["content"]) > 50000: msg["content"] = msg["content"][:50000] try: if PROVIDER == "groq": gen = query_groq(messages, model, system, stream=True) else: gen = query_ollama(messages, model, system, stream=True) return Response( stream_with_context(gen), content_type="application/x-ndjson", ) except Exception as e: return jsonify({"error": str(e)}), 500 @app.route("/chat/quick", methods=["POST"]) def chat_quick(): """Non-streaming chat endpoint.""" server_stats["requests"] += 1 data = request.get_json() if not data: return jsonify({"error": "Invalid JSON"}), 400 messages = data.get("messages", []) model = data.get("model", "") persona = data.get("persona", "Default") system = PERSONAS.get(persona, SYSTEM_PROMPT) if not messages: return jsonify({"error": "No messages"}), 400 # Input validation if len(messages) > 100: messages = messages[-100:] for msg in messages: if not isinstance(msg, dict) or "role" not in msg or "content" not in msg: return jsonify({"error": "Invalid message format"}), 400 if len(msg["content"]) > 50000: msg["content"] = msg["content"][:50000] try: if PROVIDER == "groq": result = query_groq(messages, model, system, stream=False) else: result = query_ollama(messages, model, system, stream=False) return jsonify(result) except Exception as e: return jsonify({"error": str(e)}), 500 # --- Main --- if __name__ == "__main__": print("=" * 50) print(" CodeGPT Backend Server") print("=" * 50) print(f" Provider: {PROVIDER}") if PROVIDER == "groq": print(f" Model: {DEFAULT_MODEL}") print(f" API Key: ****{GROQ_API_KEY[-4:]}") else: print(f" Model: {OLLAMA_MODEL}") print(f" Ollama: {OLLAMA_URL}") print(f" Port: {PORT}") print(f" Personas: {', '.join(PERSONAS.keys())}") print("=" * 50) print(f" http://localhost:{PORT}") print(f" http://0.0.0.0:{PORT}") print(" Ctrl+C to stop.\n") app.run(host="0.0.0.0", port=PORT, debug=False)