What the data for this category looks like
Data related to service inquiries and responses in medical product distribution mainly comes from pharmaceutical business qualification documents, purchase and sales contracts, cold chain transportation records, drug package inserts, and government service inquiry and response documents. Qualification documents are updated annually. Purchase and sales contracts are generated in real time with business transactions. Cold chain transportation records are updated daily. Drug package inserts are updated synchronously with changes to drug approval documents. Documents fall into three categories: encrypted PDFs with digital signatures, structured spreadsheet ledgers, and long-text service guidelines. Fields include generic drug name, specification, batch number, transportation temperature, inquiry response content, and others. Units include box, vial, case, ℃, yuan, and others.
Constraints on document parsing and chunking
Encrypted PDFs with digital signatures prevent default parsing engines from reading text content. Parsing configurations must be adjusted to support encrypted documents. Structured spreadsheet ledger cells have tightly linked content. Fixed-length chunking easily splits business data across cells, breaking semantic integrity. Long-text cold chain records and service guidelines contain continuous timelines or question-and-answer logic. Chunking must preserve contextual connections. Large batches of qualification documents increase parsing time. Timeout configurations must be adjusted to avoid task interruptions.
How to set configurations
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
customPdfParse | Enabled | Adapts to encrypted PDF documents with digital signatures common in medical product distribution scenarios, preventing content recognition failures |
chunkSize | 800–1200 characters | Matches the semantic length of a single business inquiry or single purchase and sales record in medical documents, avoiding splitting complete information blocks |
chunkSplitter | Mixed paragraph and fixed-length splitting | Balances structured spreadsheets and long-text documents, avoiding splitting linked content within table cells |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Supports processing large batches of cold chain transportation records or qualification documents, preventing parsing tasks from timing out mid-process |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Supports uploading large batches of medical qualification documents, meeting the document storage needs of state-owned enterprises and public sector entities in medical product distribution |
chunkSettingMode | Custom mode | Allows manual adjustment of chunking rules for structured fields in medical documents, adapting to the layout of ledger-style documents |
Three common mistakes
- Phenomenon: After uploading a digitally signed pharmaceutical business qualification PDF, the knowledge base returns empty text content. Cause: The
customPdfParseconfiguration is not enabled, and the default parsing engine cannot recognize the content of encrypted PDFs with digital signatures. - Phenomenon: After uploading a structured purchase and sales ledger spreadsheet, the chunking result splits the drug batch number and quantity information from the same row. Cause: A pure fixed-length chunking rule is used, without adapting to the structured layout of spreadsheet documents.
- Phenomenon: Calling the knowledge base interface returns a file reading link, and clicking the link prompts that only .txt and .m format files are supported. Cause: The
customPdfParseconfiguration is not enabled, causing the format of non-standard medical documents to not be correctly recognized, and the returned link is associated with an incorrect parsing format.
How to confirm the configuration is correct
- Upload a digitally signed pharmaceutical business license PDF, and check if the parsed text content includes core fields such as the license number and business scope.
- Upload a structured purchase and sales ledger spreadsheet, and check if the chunking result retains complete row data without splitting linked content within cells.
- Adjust the
chunkSizeparameter, test chunking of long-text service guidelines, and confirm that each chunk's length falls within the set range. - View the parsing task logs, and confirm that the
PARSE_FILE_TIMEOUT_SECONDSconfiguration does not trigger timeout errors.
The values provided are common starting points and should be measured against the reader's own samples.
Question material comes from public community discussions. Configuration values are common starting points and should be measured against your own samples. Verified on 2026-09-28.