Spaces:
Sleeping
Sleeping
File size: 11,154 Bytes
b41da6b 17dd328 8945179 b41da6b f30d224 b41da6b b1e0166 b41da6b f30d224 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 |
import os
import json
import random
import logging
from typing import List, Dict, Any
import streamlit as st
from dotenv import load_dotenv
import autogen
from autogen import AssistantAgent, UserProxyAgent
from langchain_community.embeddings import OpenAIEmbeddings
from langchain_community.vectorstores import Chroma
from langchain.docstore.document import Document
load_dotenv()
OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)
st.set_page_config(page_title="IT Support System (RAG)", layout="centered")
# Initialize session memory
if "chat_history" not in st.session_state:
st.session_state.chat_history = []
#Knowledge Base Setup
kb_path = os.path.join(os.path.dirname(__file__), 'kb.json')
with open(kb_path, encoding='utf-8') as f:
kb_entries = json.load(f)
docs: List[Document] = []
for entry in kb_entries:
docs.append(Document(
page_content=entry['answer'],
metadata={
'id': entry.get('id'),
'question': entry.get('question')
}
))
embeddings = OpenAIEmbeddings(openai_api_key=OPENAI_API_KEY)
vectordb = Chroma.from_documents(
documents=docs,
embedding=embeddings,
persist_directory='db/chroma'
)
retriever = vectordb.as_retriever(search_kwargs={"k": 3})
def retrieve_docs(query: str) -> str:
"""Retrieve relevant documentation from the knowledge base"""
hits = retriever.get_relevant_documents(query)
if not hits:
return "No relevant documentation found."
results = []
for doc in hits:
question = doc.metadata.get('question', 'FAQ')
results.append(f"**Q: {question}**\nA: {doc.page_content}")
return "\n\n".join(results)
def escalate_ticket(query: str, analysis: str = "") -> str:
"""Create a ticket for issues that need human intervention"""
ticket_id = f"TICKET-{random.randint(1000, 9999)}"
description = f"User Query: {query}\nAnalysis: {analysis}"
# In a real system, you would send this to a ticketing system like JIRA, ServiceNow, etc.
logger.info(f"Escalating issue with ticket {ticket_id}: {description}")
return f"Escalated issue. Created ticket {ticket_id}. A support technician will contact you shortly."
# LLM Configuration
llm_config = {
"config_list": [{
"model": "gpt-4",
"api_key": OPENAI_API_KEY,
}],
"seed": 42,
"temperature": 0.5,
}
# Agent Definitions
master_agent = AssistantAgent(
name="Master",
llm_config=llm_config,
system_message="""
You are the Master Agent that orchestrates the IT support workflow:
1. First determine if the query is IT-related. If not, provide a direct response explaining your limitations.
2. For IT-related queries, pass to the Planning Agent for execution plan development
3. Receive and review the complete workflow results from all agents
4. Provide a comprehensive yet concise final response to the user
Only handle one query at a time through the complete workflow.
"""
)
planning_agent = AssistantAgent(
name="Planning",
llm_config=llm_config,
system_message="""
You are the Planning Agent responsible for:
1. Validating if the user query is clear and complete
2. Refining the query if needed for better processing
3. Creating a structured execution plan with clear steps
4. Categorizing the IT issue type (hardware, software, network, access, etc.)
Provide your analysis as a structured output with sections for:
- Query Validation
- Issue Category
- Execution Plan
Always end your message with: "Forwarding to Analysis Agent"
"""
)
analysis_agent = AssistantAgent(
name="Analysis",
llm_config=llm_config,
system_message="""
You are the Analysis Agent responsible for:
1. Identifying key entities in the user query (devices, software, errors, etc.)
2. Determining severity level (Low, Medium, High, Critical)
3. Extracting technical details mentioned in the query
4. Structuring this information for the Resolution phase
Provide your analysis as structured output with sections for:
- Key Entities
- Technical Details
- Severity
- Analysis Summary
Always end your message with: "Forwarding to Resolution Agent"
"""
)
resolution_agent = AssistantAgent(
name="Resolution",
llm_config=llm_config,
system_message="""
You are the Resolution Agent responsible for:
1. Using available tools to retrieve relevant knowledge base articles or documentation
2. Applying the retrieved information to the specific user issue
3. Providing clear step-by-step instructions for resolution
4. Determining if the issue requires escalation to a human technician
If you can resolve the issue:
- Provide clear instructions
- Include any relevant documentation references
- End with "RESOLUTION COMPLETE"
If escalation is needed:
- Explain why the issue requires escalation
- Provide details that would help a technician understand the issue
- End with "ESCALATION NEEDED"
""",
function_map={"retrieve_docs": retrieve_docs}
)
escalation_agent = AssistantAgent(
name="Escalation",
llm_config=llm_config,
system_message="""
You are the Escalation Agent responsible for:
1. Creating support tickets for issues that cannot be resolved automatically
2. Providing the user with ticket tracking information
3. Setting expectations for next steps
4. Compiling all analysis from previous agents to assist human technicians
Format your response to be professional and reassuring to the user.
Always include the ticket ID and expected follow-up timeframe.
""",
function_map={"escalate_ticket": escalate_ticket}
)
def handle_it_query(query: str) -> str:
"""Process IT queries through the multi-agent workflow"""
query = query.strip()
if not query:
return "Please enter an IT question or issue."
workflow_logs = {"query": query}
try:
# First determine if it's an IT issue through the Master Agent
master_proxy = UserProxyAgent(
name="MasterProxy", human_input_mode="NEVER", code_execution_config=False
)
master_prompt = f"User query: '{query}'. First, determine if this is an IT-related issue."
master_proxy.initiate_chat(master_agent, message=master_prompt, max_turns=1)
initial_assessment = master_proxy.chat_messages[master_agent][-1]["content"]
workflow_logs["initial_assessment"] = initial_assessment
# If not IT-related, return the response directly
if "NOT IT-RELATED" in initial_assessment.upper():
return initial_assessment
# Planning
plan_proxy = UserProxyAgent(
name="PlanningProxy", human_input_mode="NEVER", code_execution_config=False
)
plan_proxy.initiate_chat(planning_agent, message=query, max_turns=1)
planning_output = plan_proxy.chat_messages[planning_agent][-1]["content"]
workflow_logs["planning"] = planning_output
logger.info(f"Planning completed: {len(planning_output)} chars")
# 2: Analysis
analysis_proxy = UserProxyAgent(
name="AnalysisProxy", human_input_mode="NEVER", code_execution_config=False
)
analysis_proxy.initiate_chat(analysis_agent, message=planning_output, max_turns=1)
analysis_output = analysis_proxy.chat_messages[analysis_agent][-1]["content"]
workflow_logs["analysis"] = analysis_output
logger.info(f"Analysis completed: {len(analysis_output)} chars")
# 3: Resolution
res_proxy = UserProxyAgent(
name="ResolutionProxy", human_input_mode="NEVER", code_execution_config=False
)
resolution_input = f"User Query: {query}\n\nPlanning: {planning_output}\n\nAnalysis: {analysis_output}"
res_proxy.initiate_chat(resolution_agent, message=resolution_input, max_turns=1)
resolution_output = res_proxy.chat_messages[resolution_agent][-1]["content"]
workflow_logs["resolution"] = resolution_output
logger.info(f"Resolution completed: {len(resolution_output)} chars")
# Escalation if needed
escalation_output = None
if "ESCALATION NEEDED" in resolution_output.upper():
esc_proxy = UserProxyAgent(
name="EscalationProxy", human_input_mode="NEVER", code_execution_config=False
)
escalation_input = f"Original Query: {query}\n\nAnalysis: {analysis_output}\n\nResolution Attempt: {resolution_output}"
esc_proxy.initiate_chat(escalation_agent, message=escalation_input, max_turns=1)
escalation_output = esc_proxy.chat_messages[escalation_agent][-1]["content"]
workflow_logs["escalation"] = escalation_output
logger.info(f"Escalation completed: {len(escalation_output)} chars")
# Master summarizes
final_master_proxy = UserProxyAgent(
name="FinalMasterProxy", human_input_mode="NEVER", code_execution_config=False
)
if escalation_output:
final_prompt = (
f"Complete workflow results for query: '{query}':\n\n"
f"Planning: {planning_output}\n\n"
f"Analysis: {analysis_output}\n\n"
f"Resolution: {resolution_output}\n\n"
f"Escalation: {escalation_output}\n\n"
f"Synthesize these results into a clear, helpful response for the user."
)
else:
final_prompt = (
f"Complete workflow results for query: '{query}':\n\n"
f"Planning: {planning_output}\n\n"
f"Analysis: {analysis_output}\n\n"
f"Resolution: {resolution_output}\n\n"
f"Synthesize these results into a clear, helpful response for the user."
)
final_master_proxy.initiate_chat(master_agent, message=final_prompt, max_turns=1)
final_response = final_master_proxy.chat_messages[master_agent][-1]["content"]
workflow_logs["final_response"] = final_response
# Save to memory
st.session_state.chat_history.append({
"user": query,
"assistant": final_response,
"workflow_logs": workflow_logs
})
return final_response
except Exception as e:
logger.error(f"Error in workflow: {e}", exc_info=True)
return f"An error occurred during processing: {str(e)}\n\nPlease try rephrasing your question."
st.title("AI Help Desk")
st.write("Ask any IT support question and our multi-agent system will assist you.")
with st.form(key="query_form", clear_on_submit=True):
user_input = st.text_area("Describe your IT issue:", height=100)
show_logs = st.checkbox("Show workflow details", value=False)
submitted = st.form_submit_button("Submit")
if submitted:
if not user_input:
st.error("Please type a message before submitting.")
else:
with st.spinner("Processing your request through our agent workflow..."):
response = handle_it_query(user_input)
st.markdown("### Response")
st.write(response)
|