checkExisting

import requests
import json
import os
import urllib3

# 1. Configuration
USER_KEY = os.getenv("USER_KEY")
MODEL_API = "https://granite-3-3-8b-instruct--apicast-production.apps.int.stc.ai.prod.us-east-1.aws.paas.redhat.com:443/v1/chat/completions"
MODEL_ID = "ibm-granite/granite-3.3-8b-instruct"

# This silences the 'InsecureRequestWarning' if we have to fallback to verify=False
urllib3.disable_warnings(urllib3.exceptions.InsecureRequestWarning)

def call_granite(messages, tools=None):
    headers = {
        "Authorization": f"Bearer {USER_KEY}",
        "Content-Type": "application/json"
    }
    
    payload = {
        "model": MODEL_ID,
        "messages": messages,
        "temperature": 0.2,
        "max_tokens": 500
    }
    
    if tools:
        payload["tools"] = tools

    # 1. Try to find the certificate in your current folder
    cert_file = "Current-IT-Root-CAs.pem"
    
    try:
        if os.path.exists(cert_file):
            response = requests.post(MODEL_API, headers=headers, json=payload, verify=cert_file)
        else:
            # 2. HACKATHON FALLBACK: If the file is missing, just skip verification to keep moving
            print(f"--- Warning: {cert_file} not found. Proceeding without verification ---")
            response = requests.post(MODEL_API, headers=headers, json=payload, verify=False)
        
        response.raise_for_status() # This will catch 401/500 errors
        return response.json()
    
    except requests.exceptions.RequestException as e:
        print(f"API Error: {e}")
        return None
# --- TOOL DEFINITION ---
tools = [
    {
        "type": "function",
        "function": {
            "name": "search_jira",
            "description": "Search issues.redhat.com for existing bug reports.",
            "parameters": {
                "type": "object",
                "properties": {
                    "jql": {
                        "type": "string",
                        "description": "The JQL query (e.g. 'text ~ \"lan tx queue\"')"
                    }
                },
                "required": ["jql"]
            }
        }
    }
]

# --- MOCK JIRA FUNCTION ---
# (Replace this later with a real requests call to Jira)
def search_jira(jql):
    print(f"\n[SYSTEM] Agent is searching Jira with: {jql}")
    # Mock data for your hackathon demo
    if "lan tx" in jql.lower():
        return [{"key": "RHEL-9921", "summary": "Failed to set LAN Tx queue on certain NICs", "status": "Closed"}]
    return []

# --- UPDATE YOUR call_granite CALL ---
# Change your loop to include 'tools=tools' in the call_granite function


# 2. Define the Agent's "Persona"
system_prompt = {
    "role": "system",
    "content": (
        "You are the 'CheckExisting' Agent. When a user provides a bug, extract the technical error code or unique symptoms. Use the search_jira tool. If the search returns results, compare the summaries. If a match is found, provide the link and say 'DUPLICATE DETECTED'. If no similar bugs exist, say 'UNIQUE ISSUE - PROCEEDING TO JIRA CREATION'"
    )
}

# 3. Main Loop
def run_agent():
    print("=== CheckExisting Agent Active ===")
    history = [system_prompt]

    while True:
        user_input = input("Enter Bug Description (or 'quit'): ")
        if user_input.lower() == 'quit': break

        history.append({"role": "user", "content": user_input})
    
        # 1. Call Granite with Tools
        result = call_granite(history, tools=tools)
        message = result['choices'][0]['message']

        # 2. Check if AI wants to use the tool
        if "tool_calls" in message:
            for tool_call in message["tool_calls"]:
            # Extract the JQL the AI wrote
                args = json.loads(tool_call["function"]["arguments"])
                jql = args.get("jql")
            
                # Execute Tool
                results = search_jira(jql)
            
                # Feed the Observation back to the AI
                history.append(message) # Add the assistant's request to history
                history.append({
                    "role": "tool",
                    "tool_call_id": tool_call["id"],
                    "name": "search_jira",
                    "content": json.dumps(results)
                })
            
            # 3. Final call to get the AI's summary of the data
            final_result = call_granite(history)
            print(f"\nAI Final Analysis: {final_result['choices'][0]['message']['content']}\n")
    else:
        print(f"\nAI Analysis: {message['content']}\n")

if __name__ == "__main__":
    run_agent()