Feat: Desacoplamento

This commit is contained in:
2026-03-12 15:02:26 -03:00
parent 08be8e314d
commit 8c3c23c65f
11 changed files with 231 additions and 163 deletions

View File

View File

@@ -0,0 +1,46 @@
import json
import time
from utils.langgraph_agent import RULES, run_agent
async def evaluate_servico(servico: dict, guia: dict, file_content: str) -> dict:
codigo_servico_raw = str(servico.get("codigoServico", ""))
code = "".join(c for c in codigo_servico_raw if c.isdigit())
if code not in RULES:
return {
"codigoServico": codigo_servico_raw,
"resultado": "SKIPPED",
"motivo": f"Codigo '{code}' nao encontrado nas regras",
"agentOutput": "",
"tempoAgentSegundos": 0,
}
query_data = {
"atendimento": guia.get("atendimento", {}),
"guia": guia.get("guia", {}),
"servico": servico,
"historico": guia.get("historico", {}),
}
query = json.dumps(query_data, indent=2, ensure_ascii=False)
t0 = time.time()
try:
result = await run_agent(query, code, file_content)
agent_output = result["response"]
input_tokens = result["input_tokens"]
output_tokens = result["output_tokens"]
except Exception as e:
agent_output = f"ERROR: {str(e)}"
input_tokens = 0
output_tokens = 0
return {
"codigoServico": codigo_servico_raw,
"resultado": "Aprovado" if "".join(c for c in agent_output.lower() if c.isalpha()).startswith("aprov") else "Reprovado",
"agentOutput": agent_output,
"input_tokens": input_tokens,
"output_tokens": output_tokens,
"tempoAgentSegundos": round(time.time() - t0, 2),
}

View File

@@ -0,0 +1,75 @@
import asyncio
import io
from pathlib import Path
from urllib.parse import urlparse
import boto3
from PyPDF2 import PdfReader, PdfWriter
from utils.config import AWS_REGION
from utils.secrets_manager import SECRETS
_s3_input = boto3.client(
"s3",
aws_access_key_id=SECRETS["AWS_ACCESS_KEY"],
aws_secret_access_key=SECRETS["AWS_SECRET_KEY"],
region_name=AWS_REGION,
)
_textract = boto3.client("textract", region_name=AWS_REGION)
def parse_s3_uri(s3_uri: str) -> tuple[str, str]:
parsed = urlparse(s3_uri)
if parsed.scheme != "s3":
raise ValueError(f"Not an S3 URI: {s3_uri}")
bucket = parsed.netloc
key = parsed.path.lstrip("/")
if not bucket or not key:
raise ValueError(f"Invalid S3 URI: {s3_uri}")
return bucket, key
def _extract_text_from_textract_response(response: dict) -> str:
if not response:
return ""
return "\n".join(
block["Text"] for block in response.get("Blocks", [])
if block["BlockType"] == "LINE"
)
def _split_pdf_pages(pdf_bytes: bytes) -> list[bytes]:
reader = PdfReader(io.BytesIO(pdf_bytes))
pages = []
for page in reader.pages:
writer = PdfWriter()
writer.add_page(page)
buf = io.BytesIO()
writer.write(buf)
pages.append(buf.getvalue())
return pages
def _textract_detect_bytes(file_bytes: bytes) -> str:
response = _textract.detect_document_text(Document={"Bytes": file_bytes})
return _extract_text_from_textract_response(response)
async def extract_text_from_s3_document(bucket: str, key: str) -> tuple[str, int]:
file_bytes = await asyncio.to_thread(
lambda: _s3_input.get_object(Bucket=bucket, Key=key)["Body"].read()
)
file_ext = Path(key).suffix.lower()
if file_ext in [".png", ".jpg", ".jpeg"]:
text = await asyncio.to_thread(_textract_detect_bytes, file_bytes)
return text, 1
if file_ext == ".pdf":
page_bytes_list = await asyncio.to_thread(_split_pdf_pages, file_bytes)
texts = await asyncio.gather(*[
asyncio.to_thread(_textract_detect_bytes, p) for p in page_bytes_list
])
return "\n".join(texts), len(page_bytes_list)
return "", 0

View File

@@ -0,0 +1,29 @@
import asyncio
import json
from datetime import datetime
import boto3
from utils.config import AWS_REGION, OUTPUT_BUCKET, API_VERSION
_s3_output = boto3.client("s3", region_name=AWS_REGION)
async def save_results(results: list[dict]) -> None:
timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
async def _save_one(guia_result: dict):
numero_guia = guia_result.get("guia", {}).get("codigoGuiaLocal", "unknown")
key = f"{API_VERSION}/{numero_guia}_{timestamp}.json"
await asyncio.to_thread(
_s3_output.put_object,
Bucket=OUTPUT_BUCKET,
Key=key,
Body=json.dumps(guia_result, ensure_ascii=False),
ContentType="application/json",
)
try:
await asyncio.gather(*[_save_one(g) for g in results])
except Exception:
pass