File size: 7,721 Bytes
9a10687 270c1c7 9a10687 270c1c7 9a10687 270c1c7 9a10687 270c1c7 9a10687 270c1c7 9a10687 |
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 |
"""
Chat Service - Intelligent chat with RAG capabilities
Uses local database + Lancer for comprehensive responses
"""
import httpx
from typing import Optional, List, Dict, Any
from sqlalchemy.orm import Session
from app.config import settings
from app.models.entity import Entity, Relationship
LANCER_URL = "https://madras1-lancer.hf.space/api/v1"
SYSTEM_PROMPT = """Você é um assistente de inteligência do NUMIDIUM.
Você tem acesso a um grafo de conhecimento com entidades e relacionamentos,
e pode pesquisar na web para informações atualizadas.
Responda em português brasileiro de forma clara e direta.
Se não tiver certeza, diga que não sabe em vez de inventar."""
class ChatService:
"""Chat service with RAG using local database and Lancer"""
def __init__(self):
self.api_url = "https://api.cerebras.ai/v1/chat/completions"
self.conversation_history: Dict[str, List[Dict[str, str]]] = {}
def _get_history(self, session_id: Optional[str]) -> List[Dict[str, str]]:
key = session_id or "default"
if key not in self.conversation_history:
self.conversation_history[key] = []
return self.conversation_history[key]
def clear_history(self, session_id: Optional[str] = None):
"""Clear conversation history"""
key = session_id or "default"
self.conversation_history.pop(key, None)
def _get_local_context(self, query: str, db: Session, limit: int = 5) -> str:
"""Get relevant entities from local database"""
# Search entities by name
entities = db.query(Entity).filter(
Entity.name.ilike(f"%{query}%")
).limit(limit).all()
# Also search by description
if len(entities) < limit:
desc_entities = db.query(Entity).filter(
Entity.description.ilike(f"%{query}%")
).limit(limit - len(entities)).all()
entities.extend(desc_entities)
if not entities:
# Try splitting query into words
words = query.split()
for word in words:
if len(word) > 3:
word_entities = db.query(Entity).filter(
Entity.name.ilike(f"%{word}%")
).limit(2).all()
entities.extend(word_entities)
if not entities:
return ""
context_parts = []
seen_ids = set()
for entity in entities:
if entity.id in seen_ids:
continue
seen_ids.add(entity.id)
ctx = f"• {entity.name} ({entity.type})"
if entity.description:
ctx += f": {entity.description[:200]}"
# Get relationships
relationships = db.query(Relationship).filter(
(Relationship.source_id == entity.id) |
(Relationship.target_id == entity.id)
).limit(5).all()
if relationships:
related = []
for rel in relationships:
if rel.source_id == entity.id:
target = db.query(Entity).filter(Entity.id == rel.target_id).first()
if target:
related.append(f"{rel.type} → {target.name}")
else:
source = db.query(Entity).filter(Entity.id == rel.source_id).first()
if source:
related.append(f"{source.name} → {rel.type}")
if related:
ctx += f" | Relações: {', '.join(related[:3])}"
context_parts.append(ctx)
return "\n".join(context_parts)
async def _get_web_context(self, query: str) -> str:
"""Get context from Lancer web search"""
try:
async with httpx.AsyncClient(timeout=30.0) as client:
response = await client.post(
f"{LANCER_URL}/search",
json={
"query": query,
"max_results": 5,
"include_answer": True
}
)
if response.status_code == 200:
data = response.json()
if data.get("answer"):
return f"Informações da web:\n{data['answer'][:1000]}"
return ""
except Exception as e:
print(f"Lancer error: {e}")
return ""
async def _call_llm(self, messages: List[Dict[str, str]]) -> str:
"""Call Cerebras LLM"""
try:
async with httpx.AsyncClient(timeout=60.0) as client:
response = await client.post(
self.api_url,
headers={
"Authorization": f"Bearer {settings.cerebras_api_key}",
"Content-Type": "application/json"
},
json={
"model": "qwen-3-32b",
"messages": messages,
"temperature": 0.7,
"max_tokens": 2048
}
)
if response.status_code == 200:
data = response.json()
return data["choices"][0]["message"]["content"]
else:
return f"Erro na API: {response.status_code}"
except Exception as e:
return f"Erro: {str(e)}"
async def chat(
self,
message: str,
db: Session,
use_web: bool = True,
use_history: bool = True,
session_id: Optional[str] = None
) -> Dict[str, Any]:
"""Process chat message with RAG"""
history = self._get_history(session_id)
# Get local context
local_context = self._get_local_context(message, db)
# Get web context if enabled
web_context = ""
if use_web:
web_context = await self._get_web_context(message)
# Build context
context_parts = []
if local_context:
context_parts.append(f"📊 Conhecimento local:\n{local_context}")
if web_context:
context_parts.append(f"🌐 {web_context}")
context = "\n\n".join(context_parts) if context_parts else "Nenhum contexto disponível."
# Build messages
messages = [{"role": "system", "content": SYSTEM_PROMPT}]
if use_history and history:
messages.extend(history[-6:])
user_message = f"""Contexto:
{context}
Pergunta: {message}"""
messages.append({"role": "user", "content": user_message})
# Call LLM
response = await self._call_llm(messages)
# Store history
if use_history:
history.append({"role": "user", "content": message})
history.append({"role": "assistant", "content": response})
return {
"answer": response,
"local_context_used": bool(local_context),
"web_context_used": bool(web_context),
"entities_found": local_context.count("•") if local_context else 0
}
# Singleton
chat_service = ChatService()
|