Is "Upload for Text" broken?

[MemmoB]

What happened?

I'm trying to chat with an uploaded PDF-file in the "Upload for Text" mode. I have Mistral OCR configured.
The file is not added to the context and the attached files disappears from the chat.

I am using LiteLLM as a proxy - I have tried several different models, gpt-5-nano, gpt-4-omni, claude-sonnet-4.5. All fail.

Version Information

On main - commit: 20282f3

Steps to Reproduce

1. Attach a pdf using "Upload as file"
2. Ask a question about the content
3. See response

What browsers are you seeing the problem on?

Firefox

Relevant log output

....
2025-10-09T14:05:22.976Z debug: [AgentController] Request aborted on close
2025-10-09T14:05:22.977Z debug: [BaseClient] Loading history:
{
  conversationId: "605e95a5-85cc-477a-a4a8-d986da42985b",
  parentMessageId: "00000000-0000-0000-0000-000000000000",
}
2025-10-09T14:05:23.079Z debug: [BaseClient] Context Count (1/2)
{
  remainingContextTokens: 359985,
  maxContextTokens: 360000,
}
2025-10-09T14:05:23.079Z debug: [BaseClient] Context Count (2/2)
{
  remainingContextTokens: 359985,
  maxContextTokens: 360000,
}
2025-10-09T14:05:23.079Z debug: [BaseClient] tokenCountMap:
{
  663fb90f-8cab-480a-a401-c1d2e41da85e: 12,
}
2025-10-09T14:05:23.080Z debug: [BaseClient]
{
  promptTokens: 15,
  remainingContextTokens: 359985,
  payloadSize: 1,
  maxContextTokens: 360000,
}
2025-10-09T14:05:23.080Z debug: [BaseClient] tokenCountMap
{
  663fb90f-8cab-480a-a401-c1d2e41da85e: 12,
}
2025-10-09T14:05:23.080Z debug: [BaseClient] userMessage
{
  messageId: "663fb90f-8cab-480a-a401-c1d2e41da85e",
  parentMessageId: "00000000-0000-0000-0000-000000000000",
  conversationId: "605e95a5-85cc-477a-a4a8-d986da42985b",
  sender: "User",
  text: "Vilken adress för besiktningen?",
  isCreatedByUser: true,
  tokenCount: 12,
}
2025-10-09T14:05:23.131Z debug: [saveConvo] api/app/clients/BaseClient.js - saveMessageToDatabase #saveConvo
2025-10-09T14:05:29.918Z debug: [spendTokens] conversationId: 605e95a5-85cc-477a-a4a8-d986da42985b | Context: message | Token usage:
{
  promptTokens: 15,
  completionTokens: 142,
}
2025-10-09T14:05:29.937Z debug: [spendTokens] No transactions incurred against balance
2025-10-09T14:05:29.958Z debug: [saveConvo] api/app/clients/BaseClient.js - saveMessageToDatabase #saveConvo
2025-10-09T14:05:29.998Z debug: [AgentController] Request closed
2025-10-09T14:05:33.803Z debug: [spendTokens] conversationId: 605e95a5-85cc-477a-a4a8-d986da42985b | Context: title | Token usage:
{
  promptTokens: 183,
  completionTokens: 656,
}
2025-10-09T14:05:33.805Z debug: [saveConvo] api/server/services/Endpoints/agents/title.js
2025-10-09T14:05:33.836Z debug: [spendTokens] No transactions incurred against balance
2025-10-09T14:05:33.849Z debug: [AgentController] Title generation started
2025-10-09T14:05:33.849Z debug: [AgentController] Title generation completed
2025-10-09T14:05:33.849Z debug: [AgentController] Performing cleanup
2025-10-09T14:05:33.849Z debug: [AgentController] Cleaning up abort controller
2025-10-09T14:05:33.849Z debug: [AgentController] Cleanup completed

Screenshots

No response

Code of Conduct

• I agree to follow this project's Code of Conduct

[georgenmatthew]

I've also been experiencing inconsistent results with "Upload as Text", especially in the case of agents where the File Search capability is also enabled.

[danny-avila]

File Search uses a different upload option, "Upload for File Search."

See the docs for detailed explanations:

• https://www.librechat.ai/docs/features/upload_as_text#overview
• https://www.librechat.ai/docs/features/agents#file-search

[georgenmatthew]

Thanks -- if I can reliably reproduce the error behavior I'm seeing I will open a separate issue.

[danny-avila]

Is "Upload for Text" broken?

It is not, but it was changed to make OCR optional and not required. What you may be experiencing is PDFs that may have only been possible to parse via OCR (likely image-based, as whole pages of text can sometimes be "images" with PDFs)

See more info for customizing what gets used for OCR vs. not if you have it configured.

https://www.librechat.ai/docs/features/ocr#ocr-processing-configuration

[MemmoB]

Using the same config and same pdf-file it works on commit: 9c77f53.

These are my settings in librechat.yaml for fileConfig:

ocr:
  mistralModel: "mistral-ocr-2503"                 # Should match your Azure deployment name
  apiKey: "${AZURE_MISTRAL_OCR_API_KEY}"                 # Reference to your Azure API key in .env
  baseURL: "https://.....l.models.ai.azure.com/v1/ocr?api-version=2024-05-01-preview"
  strategy: "azure_mistral_ocr"                          # Use Azure strategy

fileConfig:
  # Maximum tokens from text files before truncation
  fileTokenLimit: 100000

  # Files matching these patterns use OCR (if configured)
  ocr:
    supportedMimeTypes:
      - "^image/(jpeg|gif|png|webp|heic|heif)$"
      - "^application/pdf$"
      - "^application/vnd\\.openxmlformats-officedocument\\.(wordprocessingml\\.document|presentationml\\.presentation|spreadsheetml\\.sheet)$"
      - "^application/vnd\\.ms-(word|powerpoint|excel)$"
      - "^application/epub\\+zip$"

  # Files matching these patterns use text parsing
  text:
    supportedMimeTypes:
      - "^text/(plain|markdown|csv|json|xml|html|css|javascript|typescript|x-python|x-java|x-csharp|x-php|x-ruby|x-go|x-rust|x-kotlin|x-swift|x-scala|x-perl|x-lua|x-shell|x-sql|x-yaml|x-toml)$"

I have also tested just a simple txt-file and it shows the same behavior.
I will run git bisect and see which commit introduces the behavior I'm seeing.

[danny-avila]

I can already see something got removed erroneously here: #9994

Will fix it soon

FYI @dustinhealy

[MemmoB]

Thank you! Yeah - was just about to write that the error was introduced in commit: bcd97aa

Added some logging - seems like the attachment count is 1 but documentCount is 0 after processing.

2025-10-10T07:54:14.684Z debug: [parseText] RAG_API_URL not defined, falling back to native text parsing
2025-10-10T07:54:43.020Z debug: [AgentController] Request aborted on close
2025-10-10T07:54:43.021Z debug: [BaseClient] Loading history:
{
  conversationId: "d4d44a84-2f4e-4616-ae70-e096fec0fedc",
  parentMessageId: "00000000-0000-0000-0000-000000000000",
}
2025-10-10T07:54:43.031Z debug: [AgentClient buildMessages] Processing attachments
{
  attachmentCount: 1,
  messageId: "8b33b5a6-9da3-468f-afa5-d2cbfef45ae0",
  // 1 attachment(s)
  attachments: [{"file_id":"abf4052a-2e80-417a-83c5-de544bb38d79","filename":"test.txt","source":"text","hasText":tr... [truncated]],
    0.file_id: "abf4052a-2e80-417a-83c5-de544bb38d79",
    0.filename: "test.txt",
    0.source: "text",
    0.hasText: true,
    0.textLength: 1618,
    0.embedded: undefined,
}
2025-10-10T07:54:43.031Z debug: [BaseClient processAttachments] Processing attachments
{
  attachmentCount: 1,
  // 1 attachment(s)
  attachments: [{"file_id":"abf4052a-2e80-417a-83c5-de544bb38d79","filename":"test.txt","type":"text/plain","source"... [truncated]],
    0.file_id: "abf4052a-2e80-417a-83c5-de544bb38d79",
    0.filename: "test.txt",
    0.type: "text/plain",
    0.source: "text",
    0.hasText: true,
    0.textLength: 1618,
}
2025-10-10T07:54:43.031Z debug: [BaseClient processAttachments] Files after processing
{
  imageCount: 0,
  documentCount: 0,
  videoCount: 0,
  audioCount: 0,
}
2025-10-10T07:54:43.031Z debug: [BaseClient processAttachments] Returning unique files
{
  uniqueFileCount: 0,
  files: ,
}
2025-10-10T07:54:43.031Z debug: [AgentClient buildMessages] Attachments processed
{
  processedFileCount: 0,
  messageId: "8b33b5a6-9da3-468f-afa5-d2cbfef45ae0",
}
2025-10-10T07:54:43.125Z debug: [AgentClient buildMessages] Processing files for message
{
  messageId: "8b33b5a6-9da3-468f-afa5-d2cbfef45ae0",
  messageIndex: 0,
  isLastMessage: true,
  attachmentCount: 1,
  // 1 file(s)
  files: [{"file_id":"abf4052a-2e80-417a-83c5-de544bb38d79","filename":"test.txt","source":"text","hasText":tr... [truncated]],
    0.file_id: "abf4052a-2e80-417a-83c5-de544bb38d79",
    0.filename: "test.txt",
    0.source: "text",
    0.embedded: undefined,
    0.hasText: true,
    0.textLength: 1618,
    0.hasFileIdentifier: false,
}
2025-10-10T07:54:43.125Z debug: [BaseClient] Context Count (1/2)
{
  remainingContextTokens: 359987,
  maxContextTokens: 360000,
}
2025-10-10T07:54:43.126Z debug: [BaseClient] Context Count (2/2)
{
  remainingContextTokens: 359987,
  maxContextTokens: 360000,
}
2025-10-10T07:54:43.126Z debug: [BaseClient] tokenCountMap:
{
  8b33b5a6-9da3-468f-afa5-d2cbfef45ae0: 10,
}
2025-10-10T07:54:43.126Z debug: [BaseClient]
{
  promptTokens: 13,
  remainingContextTokens: 359987,
  payloadSize: 1,
  maxContextTokens: 360000,
}
2025-10-10T07:54:43.126Z debug: [BaseClient] tokenCountMap
{
  8b33b5a6-9da3-468f-afa5-d2cbfef45ae0: 10,
}
2025-10-10T07:54:43.126Z debug: [BaseClient] userMessage
{
  messageId: "8b33b5a6-9da3-468f-afa5-d2cbfef45ae0",
  parentMessageId: "00000000-0000-0000-0000-000000000000",
  conversationId: "d4d44a84-2f4e-4616-ae70-e096fec0fedc",
  sender: "User",
  text: "Vad handlar filen om?",
  isCreatedByUser: true,
  tokenCount: 10,
}
2025-10-10T07:54:43.170Z debug: [saveConvo] api/app/clients/BaseClient.js - saveMessageToDatabase #saveConvo
2025-10-10T07:54:47.805Z debug: [spendTokens] conversationId: d4d44a84-2f4e-4616-ae70-e096fec0fedc | Context: message | Token usage:
{
  promptTokens: 13,
  completionTokens: 215,
}
2025-10-10T07:54:47.846Z debug: [spendTokens] No transactions incurred against balance
2025-10-10T07:54:47.871Z debug: [saveConvo] api/app/clients/BaseClient.js - saveMessageToDatabase #saveConvo
2025-10-10T07:54:47.916Z debug: [AgentController] Request closed
2025-10-10T07:54:53.741Z debug: [spendTokens] conversationId: d4d44a84-2f4e-4616-ae70-e096fec0fedc | Context: title | Token usage:
{
  promptTokens: 254,
  completionTokens: 1171,
}
2025-10-10T07:54:53.744Z debug: [saveConvo] api/server/services/Endpoints/agents/title.js
2025-10-10T07:54:53.786Z debug: [spendTokens] No transactions incurred against balance
2025-10-10T07:54:53.804Z debug: [AgentController] Title generation started
2025-10-10T07:54:53.804Z debug: [AgentController] Title generation completed
2025-10-10T07:54:53.804Z debug: [AgentController] Performing cleanup
2025-10-10T07:54:53.804Z debug: [AgentController] Cleaning up abort controller
2025-10-10T07:54:53.805Z debug: [AgentController] Cleanup completed

[danny-avila]

fixing here: #10069