Spaces:
Sleeping
Sleeping
File size: 14,502 Bytes
fd842a4 5234643 fd842a4 5234643 fd842a4 5234643 fd842a4 5234643 fd842a4 5234643 fd842a4 5234643 fd842a4 5234643 fd842a4 5234643 fd842a4 5234643 fd842a4 5234643 fd842a4 5234643 | 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 | from __future__ import annotations
import base64
import json
import logging
from pathlib import Path
from typing import Any
logger = logging.getLogger(__name__)
EXTRACTION_INSTRUCTIONS = (
"You extract purchase receipt information from a single document. "
"Return one JSON object that matches the schema exactly. "
"Use null for missing values, empty arrays for missing line items, and numbers for amounts. "
"Prefer the receipt's own values over guesses. "
"Extract only these fields: fecha, fecha_vencimiento, cuit_proveedor, description_proveedor, moneda, subtotal, taxes, total. "
"If the document is not a purchase receipt, set document_type to unknown but still capture any purchase-related hints you can find. "
"Do not add markdown, prose, or extra keys."
)
def build_user_prompt(file_name: str, source_type: str) -> str:
return (
f"File name: {file_name}\n\n"
f"Source type: {source_type}\n"
"Important rules:\n"
"- Read the full document before answering.\n"
"- Copy amounts as numbers, not strings.\n"
"- Use ISO 8601 for dates when possible.\n"
"- Extract only: fecha, fecha_vencimiento, cuit_proveedor, description_proveedor, moneda, subtotal, taxes, total.\n"
"- If a field cannot be found confidently, set it to null.\n"
"- Keep output compact and do not include extra fields."
)
def build_schema() -> dict[str, Any]:
return {
"name": "purchase_receipt_extraction",
"schema": {
"type": "object",
"additionalProperties": False,
"properties": {
"source_file": {"type": "string"},
"document_type": {"type": "string", "enum": ["purchase_receipt", "unknown"]},
"fecha": {"type": ["string", "null"], "description": "ISO 8601 date when possible."},
"fecha_vencimiento": {"type": ["string", "null"], "description": "ISO 8601 date when possible."},
"cuit_proveedor": {"type": ["string", "null"]},
"description_proveedor": {"type": ["string", "null"]},
"moneda": {"type": ["string", "null"]},
"subtotal": {"type": ["number", "null"]},
"taxes": {"type": ["number", "null"]},
"total": {"type": ["number", "null"]},
"confidence": {"type": ["number", "null"], "minimum": 0, "maximum": 1},
},
"required": [
"source_file",
"document_type",
"fecha",
"fecha_vencimiento",
"cuit_proveedor",
"description_proveedor",
"moneda",
"subtotal",
"taxes",
"total",
"confidence",
],
},
"strict": True,
}
def extract_receipt_json(client: Any, model: str, file_path: Path, text: str) -> dict[str, Any]:
logger.debug("extract_receipt_json called: file=%s, model=%s, text_length=%s", file_path.name, model, len(text))
try:
logger.debug("Sending text extraction request to OpenAI for %s", file_path.name)
response = client.responses.create(
model=model,
instructions=EXTRACTION_INSTRUCTIONS,
input=f"{build_user_prompt(file_path.name, 'text')}\n\nDocument text:\n{text[:20000]}",
text={
"format": {
"type": "json_schema",
"name": "purchase_receipt_extraction",
"schema": build_schema()["schema"],
"strict": True,
}
},
)
logger.debug("Received response from OpenAI for %s", file_path.name)
output_text = getattr(response, "output_text", None)
if output_text:
logger.debug("Parsing output_text from response")
result = json.loads(output_text)
logger.debug("Successfully parsed JSON: document_type=%s, total=%s, confidence=%s",
result.get("document_type"), result.get("total"), result.get("confidence"))
return result
if response.output and response.output[0].content:
logger.debug("Parsing response.output content")
result = json.loads(response.output[0].content[0].text)
logger.debug("Successfully parsed JSON: document_type=%s, total=%s, confidence=%s",
result.get("document_type"), result.get("total"), result.get("confidence"))
return result
logger.error("No JSON returned for %s - no output_text and no response.output", file_path.name)
raise RuntimeError(f"No JSON returned for {file_path.name}")
except json.JSONDecodeError as e:
logger.error("JSON decode error for %s: %s", file_path.name, e)
raise
except Exception as e:
logger.error("Error in extract_receipt_json for %s: %s", file_path.name, e)
raise
def extract_receipt_json_from_image(client: Any, model: str, file_path: Path, image_data_uri: str) -> dict[str, Any]:
logger.debug("extract_receipt_json_from_image called: file=%s, model=%s, image_size=%s bytes",
file_path.name, model, len(image_data_uri))
try:
logger.debug("Sending image extraction request to OpenAI for %s", file_path.name)
response = client.responses.create(
model=model,
instructions=EXTRACTION_INSTRUCTIONS,
input=[
{
"role": "user",
"content": [
{
"type": "input_text",
"text": build_user_prompt(file_path.name, "image"),
},
{
"type": "input_image",
"image_url": image_data_uri,
"detail": "high",
},
],
}
],
text={
"format": {
"type": "json_schema",
"name": "purchase_receipt_extraction",
"schema": build_schema()["schema"],
"strict": True,
}
},
)
logger.debug("Received response from OpenAI for image %s", file_path.name)
output_text = getattr(response, "output_text", None)
if output_text:
logger.debug("Parsing output_text from image response")
result = json.loads(output_text)
logger.debug("Successfully parsed image JSON: document_type=%s, total=%s, confidence=%s",
result.get("document_type"), result.get("total"), result.get("confidence"))
return result
if response.output and response.output[0].content:
logger.debug("Parsing response.output content from image")
result = json.loads(response.output[0].content[0].text)
logger.debug("Successfully parsed image JSON: document_type=%s, total=%s, confidence=%s",
result.get("document_type"), result.get("total"), result.get("confidence"))
return result
logger.error("No JSON returned for image %s - no output_text and no response.output", file_path.name)
raise RuntimeError(f"No JSON returned for {file_path.name}")
except json.JSONDecodeError as e:
logger.error("JSON decode error for image %s: %s", file_path.name, e)
raise
except Exception as e:
logger.error("Error in extract_receipt_json_from_image for %s: %s", file_path.name, e)
raise
def extract_receipt_json_from_pdf(client: Any, model: str, file_path: Path) -> dict[str, Any]:
"""Upload PDF to OpenAI Files API and extract structured data."""
logger.debug("extract_receipt_json_from_pdf called: file=%s, model=%s", file_path.name, model)
file_id = None
try:
# 1. Upload PDF to Files API
logger.debug("Uploading PDF to OpenAI Files API: %s", file_path.name)
with open(file_path, "rb") as pdf_file:
uploaded_file = client.files.create(
file=pdf_file,
purpose="user_data"
)
file_id = uploaded_file.id
logger.debug("PDF uploaded successfully, file_id: %s", file_id)
# 2. Send file_id to model for processing
logger.debug("Sending PDF to OpenAI via Files API for %s", file_path.name)
response = client.responses.create(
model=model,
instructions=EXTRACTION_INSTRUCTIONS,
input=[
{
"role": "user",
"content": [
{
"type": "input_file",
"file_id": file_id
},
{
"type": "input_text",
"text": build_user_prompt(file_path.name, "pdf")
}
]
}
],
text={
"format": {
"type": "json_schema",
"name": "purchase_receipt_extraction",
"schema": build_schema()["schema"],
"strict": True,
}
},
)
logger.debug("Received response from OpenAI for PDF %s", file_path.name)
output_text = getattr(response, "output_text", None)
if output_text:
logger.debug("Parsing output_text from PDF response")
result = json.loads(output_text)
logger.debug("Successfully parsed PDF JSON: document_type=%s, total=%s, confidence=%s",
result.get("document_type"), result.get("total"), result.get("confidence"))
return result
if response.output and response.output[0].content:
logger.debug("Parsing response.output content from PDF")
result = json.loads(response.output[0].content[0].text)
logger.debug("Successfully parsed PDF JSON: document_type=%s, total=%s, confidence=%s",
result.get("document_type"), result.get("total"), result.get("confidence"))
return result
logger.error("No JSON returned for PDF %s - no output_text and no response.output", file_path.name)
raise RuntimeError(f"No JSON returned for {file_path.name}")
except json.JSONDecodeError as e:
logger.error("JSON decode error for PDF %s: %s", file_path.name, e)
raise
except Exception as e:
logger.error("Error in extract_receipt_json_from_pdf for %s: %s", file_path.name, e)
raise
finally:
# Clean up uploaded file
if file_id:
try:
logger.debug("Deleting uploaded file: %s", file_id)
client.files.delete(file_id)
except Exception as e:
logger.warning("Failed to delete file %s: %s", file_id, e)
def extract_receipt_json_from_document(client: Any, model: str, file_path: Path) -> dict[str, Any]:
"""Send document type to OpenAI by extracting text (DOCX, TXT, etc)."""
logger.debug("extract_receipt_json_from_document called: file=%s, model=%s", file_path.name, model)
try:
logger.debug("Extracting text from document: %s", file_path.name)
# Extract text based on file type
if file_path.suffix.lower() == ".txt":
text = file_path.read_text(encoding="utf-8", errors="ignore")
logger.debug("Read TXT file: %s characters", len(text))
elif file_path.suffix.lower() == ".docx":
from docx import Document
document = Document(str(file_path))
paragraphs = [paragraph.text for paragraph in document.paragraphs]
text = "\n".join(paragraphs)
logger.debug("Extracted DOCX file: %s characters", len(text))
else:
logger.warning("Unsupported document format: %s, attempting generic text read", file_path.suffix)
text = file_path.read_text(encoding="utf-8", errors="ignore")
logger.debug("Read document file: %s characters", len(text))
# Send extracted text to OpenAI
logger.debug("Sending extracted document text to OpenAI for %s", file_path.name)
response = client.responses.create(
model=model,
instructions=EXTRACTION_INSTRUCTIONS,
input=f"{build_user_prompt(file_path.name, 'document')}\n\nDocument text:\n{text[:20000]}",
text={
"format": {
"type": "json_schema",
"name": "purchase_receipt_extraction",
"schema": build_schema()["schema"],
"strict": True,
}
},
)
logger.debug("Received response from OpenAI for document %s", file_path.name)
output_text = getattr(response, "output_text", None)
if output_text:
logger.debug("Parsing output_text from document response")
result = json.loads(output_text)
logger.debug("Successfully parsed document JSON: document_type=%s, total=%s, confidence=%s",
result.get("document_type"), result.get("total"), result.get("confidence"))
return result
if response.output and response.output[0].content:
logger.debug("Parsing response.output content from document")
result = json.loads(response.output[0].content[0].text)
logger.debug("Successfully parsed document JSON: document_type=%s, total=%s, confidence=%s",
result.get("document_type"), result.get("total"), result.get("confidence"))
return result
logger.error("No JSON returned for document %s - no output_text and no response.output", file_path.name)
raise RuntimeError(f"No JSON returned for {file_path.name}")
except json.JSONDecodeError as e:
logger.error("JSON decode error for document %s: %s", file_path.name, e)
raise
except Exception as e:
logger.error("Error in extract_receipt_json_from_document for %s: %s", file_path.name, e)
raise |