Spaces:
Sleeping
Sleeping
File size: 20,334 Bytes
370480b ba2fc46 239dbce ba2fc46 239dbce ba2fc46 239dbce ba2fc46 239dbce ba2fc46 239dbce ba2fc46 239dbce ba2fc46 239dbce ba2fc46 239dbce ba2fc46 239dbce ba2fc46 239dbce ba2fc46 239dbce ba2fc46 239dbce ba2fc46 239dbce ba2fc46 239dbce 370480b ba2fc46 370480b 239dbce 370480b 239dbce 370480b 239dbce 370480b 239dbce 370480b 239dbce 370480b 239dbce 370480b 239dbce 370480b 239dbce 370480b ba2fc46 370480b 239dbce 370480b 239dbce 370480b 239dbce 370480b 239dbce 370480b 239dbce 370480b 239dbce | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 |
# # backend/src/services/chat_service.py
# import json
# from sqlalchemy.ext.asyncio import AsyncSession
# from sqlalchemy.future import select
# # --- Model Imports ---
# from backend.src.models.chat import ChatHistory
# from backend.src.models.integration import UserIntegration
# from backend.src.models.user import User
# # --- Dynamic Factory & Tool Imports ---
# from backend.src.services.llm.factory import get_llm_model
# from backend.src.services.vector_store.qdrant_adapter import get_vector_store
# from backend.src.services.security.pii_scrubber import PIIScrubber
# # --- Agents ---
# from backend.src.services.tools.secure_agent import get_secure_agent
# from backend.src.services.tools.nosql_agent import get_nosql_agent
# from backend.src.services.tools.cms_agent import get_cms_agent
# # --- Router ---
# from backend.src.services.routing.semantic_router import SemanticRouter
# # --- LangChain Core ---
# from langchain_core.messages import HumanMessage, AIMessage
# from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder
# # ==========================================
# # HELPER FUNCTIONS
# # ==========================================
# async def get_user_integrations(user_id: str, db: AsyncSession) -> dict:
# """Fetches active integrations and filters valid descriptions."""
# if not user_id: return {}
# query = select(UserIntegration).where(UserIntegration.user_id == user_id, UserIntegration.is_active == True)
# result = await db.execute(query)
# integrations = result.scalars().all()
# settings = {}
# for i in integrations:
# try:
# creds = json.loads(i.credentials)
# creds['provider'] = i.provider
# creds['schema_map'] = i.schema_map if i.schema_map else {}
# if i.profile_description:
# creds['description'] = i.profile_description
# settings[i.provider] = creds
# except (json.JSONDecodeError, TypeError):
# continue
# return settings
# async def save_chat_to_db(db: AsyncSession, session_id: str, human_msg: str, ai_msg: str, provider: str):
# """Saves chat history with PII redaction."""
# if not session_id: return
# safe_human = PIIScrubber.scrub(human_msg)
# safe_ai = PIIScrubber.scrub(ai_msg)
# new_chat = ChatHistory(
# session_id=session_id, human_message=safe_human, ai_message=safe_ai, provider=provider
# )
# db.add(new_chat)
# await db.commit()
# async def get_chat_history(session_id: str, db: AsyncSession):
# """Retrieves past conversation history."""
# if not session_id: return []
# query = select(ChatHistory).where(ChatHistory.session_id == session_id).order_by(ChatHistory.timestamp.asc())
# result = await db.execute(query)
# return result.scalars().all()
# async def get_bot_persona(user_id: str, db: AsyncSession):
# """Fetches custom Bot Name and Instructions from User table."""
# try:
# stmt = select(User).where(User.id == int(user_id))
# result = await db.execute(stmt)
# user = result.scalars().first()
# if user:
# return {
# "name": getattr(user, "bot_name", "OmniAgent"),
# "instruction": getattr(user, "bot_instruction", "You are a helpful AI assistant.")
# }
# except Exception as e:
# print(f"⚠️ Error fetching persona: {e}")
# pass
# return {"name": "OmniAgent", "instruction": "You are a helpful AI assistant."}
# # ==========================================
# # MAIN CHAT LOGIC (Ultra-Strict Isolated Mode)
# # ==========================================
# async def process_chat(message: str, session_id: str, user_id: str, db: AsyncSession):
# # 1. Fetch User Settings & Persona
# user_settings = await get_user_integrations(user_id, db)
# bot_persona = await get_bot_persona(user_id, db)
# # 2. LLM Check
# llm_creds = user_settings.get('groq') or user_settings.get('openai')
# if not llm_creds:
# return "Please configure your AI Model in Settings."
# # 3. Build Tool Map for Router
# tools_map = {}
# for provider, config in user_settings.items():
# if provider in ['sanity', 'sql', 'mongodb']:
# if config.get('description'):
# tools_map[provider] = config['description']
# # 4. SEMANTIC DECISION (Router)
# selected_provider = None
# if tools_map:
# router = SemanticRouter()
# selected_provider = router.route(message, tools_map)
# response_text = ""
# provider_name = "general_chat"
# # 5. Route to Winner (Agent Execution)
# if selected_provider:
# print(f"👉 [Router] Selected Tool: {selected_provider.upper()}")
# try:
# if selected_provider == 'sanity':
# schema = user_settings['sanity'].get('schema_map', {})
# agent = get_cms_agent(user_id=user_id, schema_map=schema, llm_credentials=llm_creds)
# res = await agent.ainvoke({"input": message})
# response_text = str(res.get('output', ''))
# provider_name = "cms_agent"
# elif selected_provider == 'sql':
# role = "admin" if user_id == '99' else "customer"
# agent = get_secure_agent(int(user_id), role, user_settings['sql'], llm_credentials=llm_creds)
# res = await agent.ainvoke({"input": message})
# response_text = str(res.get('output', ''))
# provider_name = "sql_agent"
# elif selected_provider == 'mongodb':
# agent = get_nosql_agent(user_id, user_settings['mongodb'], llm_credentials=llm_creds)
# res = await agent.ainvoke({"input": message})
# response_text = str(res.get('output', ''))
# provider_name = "nosql_agent"
# if not response_text or "error" in response_text.lower():
# response_text = ""
# except Exception as e:
# print(f"❌ Agent Execution Failed: {e}")
# response_text = ""
# # 6. Fallback / RAG (ULTRA-STRICT MODE 🛡️)
# if not response_text:
# print("👉 [Router] Executing Strict RAG Fallback...")
# try:
# llm = get_llm_model(credentials=llm_creds)
# # Context from Vector DB
# context = ""
# if 'qdrant' in user_settings:
# try:
# vector_store = get_vector_store(credentials=user_settings['qdrant'])
# docs = await vector_store.asimilarity_search(message, k=3)
# if docs:
# context = "\n\n".join([d.page_content for d in docs])
# except Exception as e:
# print(f"⚠️ RAG Warning: {e}")
# # --- 🔥 THE ULTRA-STRICT SYSTEM PROMPT ---
# system_instruction = f"""
# SYSTEM IDENTITY:
# You are the '{bot_persona['name']}'. You are a 'Knowledge-Isolated' AI Assistant for this specific platform.
# CORE MISSION:
# Your ONLY source of truth is the 'CONTEXT FROM KNOWLEDGE BASE' provided below.
# You must ignore ALL of your internal pre-trained general knowledge about the world, geography, famous people, or general facts.
# STRICT OPERATING RULES:
# 1. MANDATORY REFUSAL: If the user's question cannot be answered using ONLY the provided context, you MUST exactly say: "I apologize, but I am only authorized to provide information based on the provided database. This specific information is not currently available in my knowledge base."
# 2. NO HALLUCINATION: Never attempt to be helpful using outside information. If a fact (like 'Japan's location') is not in the context, you do NOT know it.
# 3. CONTEXT-ONLY: Your existence is bounded by the data below. If the data is empty, you cannot answer anything except greetings.
# 4. GREETINGS: You may respond to 'Hi' or 'Hello' by briefly identifying yourself as '{bot_persona['name']}' and asking what data the user is looking for.
# 5. PROHIBITED TOPICS: Do not discuss any topic that is not present in the provided context.
# CONTEXT FROM KNOWLEDGE BASE:
# ---------------------------
# {context if context else "THE DATABASE IS CURRENTLY EMPTY. DO NOT PROVIDE ANY INFORMATION."}
# ---------------------------
# """
# # History Load
# history = await get_chat_history(session_id, db)
# formatted_history = []
# for chat in history:
# formatted_history.append(HumanMessage(content=chat.human_message))
# if chat.ai_message: formatted_history.append(AIMessage(content=chat.ai_message))
# # LLM Chain Setup
# prompt = ChatPromptTemplate.from_messages([
# ("system", system_instruction),
# MessagesPlaceholder(variable_name="chat_history"),
# ("human", "{question}")
# ])
# chain = prompt | llm
# ai_response = await chain.ainvoke({"chat_history": formatted_history, "question": message})
# response_text = ai_response.content
# provider_name = "rag_fallback"
# except Exception as e:
# print(f"❌ Fallback Error: {e}")
# response_text = "I apologize, but I am currently unable to process your request due to a system error."
# # 7. Save to DB
# await save_chat_to_db(db, session_id, message, response_text, provider_name)
# return response_text
import json
from sqlalchemy.ext.asyncio import AsyncSession
from sqlalchemy.future import select
from qdrant_client.http import models # <--- NEW IMPORT (Filter ke liye)
# --- Model Imports ---
from backend.src.models.chat import ChatHistory
from backend.src.models.integration import UserIntegration
from backend.src.models.user import User
# --- Dynamic Factory & Tool Imports ---
from backend.src.services.llm.factory import get_llm_model
from backend.src.services.vector_store.qdrant_adapter import get_vector_store
from backend.src.services.security.pii_scrubber import PIIScrubber
# --- Agents ---
from backend.src.services.tools.secure_agent import get_secure_agent
from backend.src.services.tools.nosql_agent import get_nosql_agent
from backend.src.services.tools.cms_agent import get_cms_agent
# --- Router ---
from backend.src.services.routing.semantic_router import SemanticRouter
# --- LangChain Core ---
from langchain_core.messages import HumanMessage, AIMessage
from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder
# ==========================================
# HELPER FUNCTIONS
# ==========================================
async def get_user_integrations(user_id: str, db: AsyncSession) -> dict:
"""Fetches active integrations and filters valid descriptions."""
if not user_id: return {}
query = select(UserIntegration).where(UserIntegration.user_id == user_id, UserIntegration.is_active == True)
result = await db.execute(query)
integrations = result.scalars().all()
settings = {}
for i in integrations:
try:
creds = json.loads(i.credentials)
creds['provider'] = i.provider
creds['schema_map'] = i.schema_map if i.schema_map else {}
if i.profile_description:
creds['description'] = i.profile_description
settings[i.provider] = creds
except (json.JSONDecodeError, TypeError):
continue
return settings
async def save_chat_to_db(db: AsyncSession, session_id: str, human_msg: str, ai_msg: str, provider: str):
"""Saves chat history with PII redaction."""
if not session_id: return
safe_human = PIIScrubber.scrub(human_msg)
safe_ai = PIIScrubber.scrub(ai_msg)
new_chat = ChatHistory(
session_id=session_id, human_message=safe_human, ai_message=safe_ai, provider=provider
)
db.add(new_chat)
await db.commit()
async def get_chat_history(session_id: str, db: AsyncSession):
"""Retrieves past conversation history."""
if not session_id: return []
query = select(ChatHistory).where(ChatHistory.session_id == session_id).order_by(ChatHistory.timestamp.asc())
result = await db.execute(query)
return result.scalars().all()
async def get_bot_persona(user_id: str, db: AsyncSession):
"""Fetches custom Bot Name and Instructions from User table."""
try:
stmt = select(User).where(User.id == int(user_id))
result = await db.execute(stmt)
user = result.scalars().first()
if user:
return {
"name": getattr(user, "bot_name", "OmniAgent"),
"instruction": getattr(user, "bot_instruction", "You are a helpful AI assistant.")
}
except Exception as e:
print(f"⚠️ Error fetching persona: {e}")
pass
return {"name": "OmniAgent", "instruction": "You are a helpful AI assistant."}
# ==========================================
# MAIN CHAT LOGIC (Ultra-Strict Isolated Mode)
# ==========================================
async def process_chat(message: str, session_id: str, user_id: str, db: AsyncSession):
# 1. Fetch User Settings & Persona
user_settings = await get_user_integrations(user_id, db)
bot_persona = await get_bot_persona(user_id, db)
# 2. LLM Check
llm_creds = user_settings.get('groq') or user_settings.get('openai')
if not llm_creds:
return "Please configure your AI Model in Settings."
# 3. Build Tool Map for Router
tools_map = {}
for provider, config in user_settings.items():
if provider in ['sanity', 'sql', 'mongodb']:
if config.get('description'):
tools_map[provider] = config['description']
# 4. SEMANTIC DECISION (Router)
selected_provider = None
if tools_map:
router = SemanticRouter()
selected_provider = router.route(message, tools_map)
response_text = ""
provider_name = "general_chat"
# 5. Route to Winner (Agent Execution)
if selected_provider:
print(f"👉 [Router] Selected Tool: {selected_provider.upper()}")
try:
if selected_provider == 'sanity':
schema = user_settings['sanity'].get('schema_map', {})
agent = get_cms_agent(user_id=user_id, schema_map=schema, llm_credentials=llm_creds)
res = await agent.ainvoke({"input": message})
response_text = str(res.get('output', ''))
provider_name = "cms_agent"
elif selected_provider == 'sql':
role = "admin" if user_id == '99' else "customer"
agent = get_secure_agent(int(user_id), role, user_settings['sql'], llm_credentials=llm_creds)
res = await agent.ainvoke({"input": message})
response_text = str(res.get('output', ''))
provider_name = "sql_agent"
elif selected_provider == 'mongodb':
agent = get_nosql_agent(user_id, user_settings['mongodb'], llm_credentials=llm_creds)
res = await agent.ainvoke({"input": message})
response_text = str(res.get('output', ''))
provider_name = "nosql_agent"
if not response_text or "error" in response_text.lower():
response_text = ""
except Exception as e:
print(f"❌ Agent Execution Failed: {e}")
response_text = ""
# 6. Fallback / RAG (ULTRA-STRICT MODE 🛡️)
if not response_text:
print("👉 [Router] Executing Strict RAG Fallback...")
try:
llm = get_llm_model(credentials=llm_creds)
# Context from Vector DB
context = ""
if 'qdrant' in user_settings:
try:
vector_store = get_vector_store(credentials=user_settings['qdrant'])
# 🔥 SECURITY FIX: FILTER BY USER_ID 🔥
# Hum ensure kar rahe hain ke LangChain sirf ISI USER ka data uthaye.
# QdrantAdapter mein humne metadata_payload_key="metadata" set kiya tha.
# Isliye key "metadata.user_id" hogi.
user_filter = models.Filter(
must=[
models.FieldCondition(
key="metadata.user_id",
match=models.MatchValue(value=str(user_id))
)
]
)
# Ab search mein filter pass karein
docs = await vector_store.asimilarity_search(
message,
k=3,
filter=user_filter
)
if docs:
context = "\n\n".join([d.page_content for d in docs])
except Exception as e:
print(f"⚠️ RAG Warning: {e}")
# --- 🔥 THE ULTRA-STRICT SYSTEM PROMPT ---
system_instruction = f"""
SYSTEM IDENTITY:
You are the '{bot_persona['name']}'. You are a 'Knowledge-Isolated' AI Assistant for this specific platform.
CORE MISSION:
Your ONLY source of truth is the 'CONTEXT FROM KNOWLEDGE BASE' provided below.
You must ignore ALL of your internal pre-trained general knowledge about the world, geography, famous people, or general facts.
STRICT OPERATING RULES:
1. MANDATORY REFUSAL: If the user's question cannot be answered using ONLY the provided context, you MUST exactly say: "I apologize, but I am only authorized to provide information based on the provided database. This specific information is not currently available in my knowledge base."
2. NO HALLUCINATION: Never attempt to be helpful using outside information. If a fact (like 'Japan's location') is not in the context, you do NOT know it.
3. CONTEXT-ONLY: Your existence is bounded by the data below. If the data is empty, you cannot answer anything except greetings.
4. GREETINGS: You may respond to 'Hi' or 'Hello' by briefly identifying yourself as '{bot_persona['name']}' and asking what data the user is looking for.
5. PROHIBITED TOPICS: Do not discuss any topic that is not present in the provided context.
CONTEXT FROM KNOWLEDGE BASE:
---------------------------
{context if context else "THE DATABASE IS CURRENTLY EMPTY. DO NOT PROVIDE ANY INFORMATION."}
---------------------------
"""
# History Load
history = await get_chat_history(session_id, db)
formatted_history = []
for chat in history:
formatted_history.append(HumanMessage(content=chat.human_message))
if chat.ai_message: formatted_history.append(AIMessage(content=chat.ai_message))
# LLM Chain Setup
prompt = ChatPromptTemplate.from_messages([
("system", system_instruction),
MessagesPlaceholder(variable_name="chat_history"),
("human", "{question}")
])
chain = prompt | llm
ai_response = await chain.ainvoke({"chat_history": formatted_history, "question": message})
response_text = ai_response.content
provider_name = "rag_fallback"
except Exception as e:
print(f"❌ Fallback Error: {e}")
response_text = "I apologize, but I am currently unable to process your request due to a system error."
# 7. Save to DB
await save_chat_to_db(db, session_id, message, response_text, provider_name)
return response_text |