Estaoy usando la version trial de TMSAIstudio, y probando como usar la estraccion de datos desde un pdf.
El archivo pdf es este
GUEMES 2782 01 2025 RM 1868-1869-1890-1906-1924-1925-1942-1954-1969-1970-2002-2005-2017-2018-2054-2055-2056.pdf (317.8 KB)
el prompt que le envio es:
extraer del pdf los datos de "Fecha","Nro Comprobante o nro.cpbte o numero","Tipo Comprobante",
"codigo comprobante"(no poner la letra solo el codigo), si no exieste poner vacio,"Nombre del Emisor", "Cuit del Emisor",
y el arreglo de items compuestos por :
"codigo o id ","Descripcion o Detalle","Cantidad o cant","Precio Unitario","Descuento","Total".
Terminar con el "Neto Comprobante","Descuento General",Total del Comprobante", sin agregado de carcaters especiales.
Los items representarlos separado por punto y coma entre sus partes, si no hay Descuento por items poner 0.
Si codigo no existe poner "********'.
En el detalle poner los datos y no detallar la descripcion.
pero me da error de carga:
[Window Title]
Invoiceextractor
[Content]
Error processing: 415#13#10{
"error": {
"message": "Invalid Content-Type header (application/json,multipart/form-data; boundary=AaB03x), expected multipart/form-data. (HINT: If you're using curl, you can pass -H 'Content-Type: multipart/form-data')",
"type": "invalid_request_error",
"param": null,
"code": null
}
}
[OK]
Me pueden ayudar?
Gracias
Estoy usando el ejemplo de Carga de pdf
What LLM do you use? How do you submit the PDF?
Not all LLMs offer at this time automatic data extraction from PDF files.
See this blog
that shows at this time, these LLMs support it: OpenAI, Gemini or Claude.
Hello
I use OpenAI, the thing is that sometimes it takes it right at once, but other times I have to insist, that's why I save the file.
this one did me good
It is then hard to judge whether it is an issue with the LLM or our interface if it doesn't work well.
If you have evidence that it is our interface, please provide some sample with which we can reproduce this here so we can investigate.
Sorry, but I'm not saying it's the platform. The example is made with your example. Just adjust the extracted data to my comfort. And I sent you the file I created as a pdf. There are times when I have to request a PDF extraction 2 or 3 times. Then bring the info. If you want to try it, look at reading several pdfs and you will see the error. Thank you.
Luis Roldan
Do you have reproducible steps / PDF files with which this can be reproduced?
Hello Bruno,
We are also facing this issue, 100% reproducible, and just using the invoice file available in your delphi project zip.
For what it worth:
- tested with and without a proxy
- tested with both user account and service account API key
All attemps led to this error.
Kind regards,
Stéphane
Try with the newest TMS AI Studio v1.1.0.0
I'm sorry Bruno but i'm getting another error (415):
Isn't not only one content type allowed per request body? Maybe mixing application/json and multipart/form-data in the same header makes the server reject it?
Again, is this TMS AI Studio v1.1.0.0?
If so and the problem persists, please enable logging and send the log file so we can inspect.
Yes, this is. Sorry not writing it.
How to enable logs?
TMSMCPCloudAI.Logging = true
TMSMCPCloudAI.LogFileName = 'yourfile.log'
Logging was already activated. Please see pdf.log file (I have now exceeded my current quota: Im using the free plan)
pdf.log (5.0 KB)
We have done an improvement to avoid this potential header issue.
This improvement is in v1.1.0.1
