Create app.py

app.py

@@ -0,0 +1,110 @@
+# app.py
+import os
+import json
+import requests
+from flask import Flask, request, jsonify
+
+app = Flask(__name__)
+
+# ---- API KEYS ----
+SKYVERN_API_KEY = "eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJleHAiOjQ5MDMwNTkxMjcsInN1YiI6Im9fNDM5ODczMDQ4OTgwNTgyMjQwIn0.IlKt9HFbDg1dqbv2qHxH5aRU4dwn4MFzxHXUNhzL5BA"
+GEMINI_API_KEY = "AIzaSyB5DoxsabUkiUxfEVTcoMuzH-1hfJ0ZLH0"  # replace with your actual Gemini key
+
+# Skyvern endpoint
+SKYVERN_BASE_URL = "https://api.skyvern.com/v1"  # replace with correct Skyvern API base
+# Gemini endpoint
+GEMINI_BASE_URL = "https://api.openai.com/v1beta"  # Gemini API base
+
+def run_skyvern_task(task_instruction):
+    """
+    Sends a task to Skyvern API for execution.
+    Returns a dict with task status, result, and screenshot URL (if any).
+    """
+    headers = {
+        "Authorization": f"Bearer {SKYVERN_API_KEY}",
+        "Content-Type": "application/json"
+    }
+    payload = {
+        "instructions": task_instruction,
+        "returnArtifacts": True
+    }
+    try:
+        response = requests.post(f"{SKYVERN_BASE_URL}/tasks", headers=headers, json=payload)
+        data = response.json()
+        if response.status_code != 200:
+            return {"error": data, "success": False}
+        # Extract screenshot if available
+        screenshot_url = None
+        if "artifacts" in data and isinstance(data["artifacts"], list):
+            for artifact in data["artifacts"]:
+                if artifact.get("type") in ["screenshot", "image"]:
+                    screenshot_url = artifact.get("url")
+                    break
+        return {
+            "success": True,
+            "result": data.get("result") or "Task completed",
+            "screenshot": screenshot_url,
+            "raw": data
+        }
+    except Exception as e:
+        return {"error": str(e), "success": False}
+
+def get_gemini_response(prompt_text):
+    """
+    Sends the user's natural language command to Gemini API to interpret instructions.
+    Returns the processed text for Skyvern task.
+    """
+    headers = {
+        "Authorization": f"Bearer {GEMINI_API_KEY}",
+        "Content-Type": "application/json"
+    }
+    payload = {
+        "model": "gemini-pro",
+        "input": prompt_text
+    }
+    try:
+        response = requests.post(f"{GEMINI_BASE_URL}/generateContent", headers=headers, json=payload)
+        data = response.json()
+        if response.status_code != 200:
+            return {"error": data, "success": False}
+        return {"success": True, "text": data.get("outputText") or prompt_text}
+    except Exception as e:
+        return {"error": str(e), "success": False}
+
+@app.route("/execute", methods=["POST"])
+def execute_command():
+    """
+    Accepts JSON payload: {"command": "user command in natural language"}
+    Processes it via Gemini for natural language understanding
+    Then sends task to Skyvern
+    Returns JSON with result and screenshot URL
+    """
+    data = request.get_json()
+    if not data or "command" not in data:
+        return jsonify({"error": "Command is required", "success": False}), 400
+
+    user_command = data["command"]
+
+    # 1. Process natural language via Gemini
+    gemini_response = get_gemini_response(user_command)
+    if not gemini_response.get("success"):
+        return jsonify({"error": gemini_response.get("error"), "success": False}), 500
+
+    processed_command = gemini_response.get("text")
+
+    # 2. Send processed command to Skyvern
+    skyvern_response = run_skyvern_task(processed_command)
+    if not skyvern_response.get("success"):
+        return jsonify({"error": skyvern_response.get("error"), "success": False}), 500
+
+    return jsonify({
+        "success": True,
+        "user_command": user_command,
+        "processed_command": processed_command,
+        "result": skyvern_response.get("result"),
+        "screenshot": skyvern_response.get("screenshot"),
+        "raw": skyvern_response.get("raw")
+    })
+
+if __name__ == "__main__":
+    app.run(host="0.0.0.0", port=8000, debug=True)