File size: 11,830 Bytes
c817084 |
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 |
"""
AetherMap Client
Client para integração com AetherMap API - busca semântica, NER e análise de grafos.
"""
import httpx
import json
import io
from typing import List, Dict, Any, Optional
from dataclasses import dataclass, field
from datetime import datetime
import logging
from app.config import settings
logger = logging.getLogger(__name__)
# URL base do AetherMap (HuggingFace Space)
AETHERMAP_URL = getattr(settings, 'aethermap_url', 'https://madras1-aethermap.hf.space')
@dataclass
class ProcessResult:
"""Resultado do processamento de documentos"""
job_id: str
num_documents: int
num_clusters: int
num_noise: int
metrics: Dict[str, Any] = field(default_factory=dict)
cluster_analysis: Dict[str, Any] = field(default_factory=dict)
@dataclass
class SearchResult:
"""Resultado de busca semântica"""
summary: str # Resposta RAG gerada pelo LLM
results: List[Dict[str, Any]] = field(default_factory=list)
@dataclass
class EntityNode:
"""Nó de entidade no grafo"""
entity: str
entity_type: str
docs: int
degree: int = 0
centrality: float = 0.0
role: str = "peripheral" # hub, connector, peripheral
@dataclass
class EntityEdge:
"""Aresta do grafo de entidades"""
source_entity: str
target_entity: str
weight: int
reason: str
@dataclass
class EntityGraphResult:
"""Resultado da extração de entidades"""
nodes: List[EntityNode] = field(default_factory=list)
edges: List[EntityEdge] = field(default_factory=list)
hubs: List[Dict[str, Any]] = field(default_factory=list)
insights: Dict[str, Any] = field(default_factory=dict)
@dataclass
class GraphAnalysis:
"""Análise do grafo via LLM"""
analysis: str
key_entities: List[str] = field(default_factory=list)
relationships: List[str] = field(default_factory=list)
class AetherMapClient:
"""
Client para AetherMap API.
Funcionalidades:
- Processamento de documentos (embeddings + clusters)
- Busca semântica RAG (FAISS + BM25 + reranking + LLM)
- Extração de entidades NER
- Análise de grafo via LLM
"""
def __init__(self, base_url: str = None, timeout: float = 1800.0):
self.base_url = (base_url or AETHERMAP_URL).rstrip('/')
self.timeout = timeout
self._current_job_id: Optional[str] = None
@property
def current_job_id(self) -> Optional[str]:
"""Retorna o job_id atual"""
return self._current_job_id
async def process_documents(
self,
texts: List[str],
fast_mode: bool = True,
min_cluster_size: int = 0,
min_samples: int = 0
) -> ProcessResult:
"""
Processa uma lista de textos gerando embeddings e clusters.
Args:
texts: Lista de textos/documentos
fast_mode: Se True, usa PCA (rápido). Se False, usa UMAP (preciso)
min_cluster_size: Tamanho mínimo do cluster (0=auto)
min_samples: Mínimo de amostras (0=auto)
Returns:
ProcessResult com job_id e métricas
"""
# Criar arquivo TXT em memória
content = "\n".join(texts)
file_bytes = content.encode('utf-8')
try:
async with httpx.AsyncClient(timeout=self.timeout) as client:
files = {
'file': ('documents.txt', io.BytesIO(file_bytes), 'text/plain')
}
data = {
'n_samples': str(len(texts)),
'fast_mode': 'true' if fast_mode else 'false',
'min_cluster_size': str(min_cluster_size),
'min_samples': str(min_samples)
}
logger.info(f"AetherMap: Processando {len(texts)} documentos para {self.base_url}/process/")
response = await client.post(
f"{self.base_url}/process/",
files=files,
data=data
)
logger.info(f"AetherMap: Response status {response.status_code}")
if response.status_code != 200:
error_text = response.text[:500] if response.text else "No response body"
logger.error(f"AetherMap error: {response.status_code} - {error_text}")
raise Exception(f"AetherMap error: {response.status_code} - {error_text}")
result = response.json()
self._current_job_id = result.get('job_id')
metadata = result.get('metadata', {})
logger.info(f"AetherMap: Job criado {self._current_job_id}")
return ProcessResult(
job_id=self._current_job_id or "unknown",
num_documents=metadata.get('num_documents_processed', len(texts)),
num_clusters=metadata.get('num_clusters_found', 0),
num_noise=metadata.get('num_noise_points', 0),
metrics=result.get('metrics', {}),
cluster_analysis=result.get('cluster_analysis', {})
)
except httpx.TimeoutException:
logger.error(f"AetherMap: Timeout ao conectar com {self.base_url}")
raise Exception(f"Timeout: AetherMap Space pode estar dormindo. Tente novamente em alguns segundos.")
except httpx.ConnectError as e:
logger.error(f"AetherMap: Erro de conexão: {e}")
raise Exception(f"Erro de conexão com AetherMap: {e}")
except Exception as e:
logger.error(f"AetherMap: Erro inesperado: {e}")
raise
async def semantic_search(
self,
query: str,
job_id: str = None,
turbo_mode: bool = False
) -> SearchResult:
"""
Busca semântica RAG híbrida nos documentos processados.
Args:
query: Termo de busca
job_id: ID do job (se não fornecido, usa o último)
turbo_mode: Se True, busca mais rápida (menos precisa)
Returns:
SearchResult com resumo e resultados
"""
job_id = job_id or self._current_job_id
if not job_id:
raise ValueError("Nenhum job_id disponível. Processe documentos primeiro.")
async with httpx.AsyncClient(timeout=self.timeout) as client:
data = {
'query': query,
'job_id': job_id,
'turbo_mode': 'true' if turbo_mode else 'false'
}
logger.info(f"AetherMap: Buscando '{query}'...")
response = await client.post(
f"{self.base_url}/search/",
data=data
)
if response.status_code != 200:
raise Exception(f"AetherMap search error: {response.status_code} - {response.text}")
result = response.json()
return SearchResult(
summary=result.get('summary', ''),
results=result.get('results', [])
)
async def extract_entities(self, job_id: str = None) -> EntityGraphResult:
"""
Extrai entidades nomeadas (NER) e cria grafo de conexões.
Args:
job_id: ID do job (se não fornecido, usa o último)
Returns:
EntityGraphResult com nós, arestas e insights
"""
job_id = job_id or self._current_job_id
if not job_id:
raise ValueError("Nenhum job_id disponível. Processe documentos primeiro.")
async with httpx.AsyncClient(timeout=self.timeout) as client:
data = {'job_id': job_id}
logger.info(f"AetherMap: Extraindo entidades...")
response = await client.post(
f"{self.base_url}/entity_graph/",
data=data
)
if response.status_code != 200:
raise Exception(f"AetherMap entity_graph error: {response.status_code} - {response.text}")
result = response.json()
# Converter para dataclasses
nodes = [
EntityNode(
entity=n.get('entity', ''),
entity_type=n.get('type', ''),
docs=n.get('docs', 0),
degree=n.get('degree', 0),
centrality=n.get('centrality', 0.0),
role=n.get('role', 'peripheral')
)
for n in result.get('nodes', [])
]
edges = [
EntityEdge(
source_entity=e.get('source_entity', ''),
target_entity=e.get('target_entity', ''),
weight=e.get('weight', 0),
reason=e.get('reason', '')
)
for e in result.get('edges', [])
]
return EntityGraphResult(
nodes=nodes,
edges=edges,
hubs=result.get('hubs', []),
insights=result.get('insights', {})
)
async def analyze_graph(self, job_id: str = None) -> GraphAnalysis:
"""
Usa LLM para analisar o Knowledge Graph e extrair insights.
Args:
job_id: ID do job (se não fornecido, usa o último)
Returns:
GraphAnalysis com análise textual
"""
job_id = job_id or self._current_job_id
if not job_id:
raise ValueError("Nenhum job_id disponível. Processe documentos primeiro.")
async with httpx.AsyncClient(timeout=self.timeout) as client:
data = {'job_id': job_id}
logger.info(f"AetherMap: Analisando grafo com LLM...")
response = await client.post(
f"{self.base_url}/analyze_graph/",
data=data
)
if response.status_code != 200:
raise Exception(f"AetherMap analyze_graph error: {response.status_code} - {response.text}")
result = response.json()
return GraphAnalysis(
analysis=result.get('analysis', ''),
key_entities=result.get('key_entities', []),
relationships=result.get('relationships', [])
)
async def describe_clusters(self, job_id: str = None) -> Dict[str, Any]:
"""
Usa LLM para descrever cada cluster encontrado.
Args:
job_id: ID do job (se não fornecido, usa o último)
Returns:
Dict com insights por cluster
"""
job_id = job_id or self._current_job_id
if not job_id:
raise ValueError("Nenhum job_id disponível. Processe documentos primeiro.")
async with httpx.AsyncClient(timeout=self.timeout) as client:
data = {'job_id': job_id}
logger.info(f"AetherMap: Descrevendo clusters...")
response = await client.post(
f"{self.base_url}/describe_clusters/",
data=data
)
if response.status_code != 200:
raise Exception(f"AetherMap describe_clusters error: {response.status_code} - {response.text}")
return response.json()
# Instância global do client
aethermap = AetherMapClient()
|