What the Data for This Category Looks Like
Power grid enterprise business process guidance data mainly comes from internal business procedures, user service guides, Feishu shared documents, and paper scans. Updates trigger when national power policies adjust or new businesses launch, with no fixed schedule. Most documents are multi-level structured text, with fields including business number, processing time limit, required materials, and processing channels. Units include workdays, kilovolts, kilowatt-hours, and other professional measurement and time units. Some documents contain embedded doc attachments and image resources.
Constraints on the Document Parsing and Chunking Link
Multi-level structured document structures require parsing processes to retain hierarchical relationships, to avoid losing business process logical order after splitting. Long-text business procedures require chunk length to match complete business link information density, to avoid context breaks from chunks being too short or semantic ambiguity from chunks being too long. Professional terms and fixed units require parsers to adapt to power industry-specific expressions, to avoid incorrect unit and term splitting. Scenarios with embedded non-PDF attachments and images require parsing processes to support full multi-format file parsing, to avoid missing resources. Documents with no fixed update cycle require parsing configurations to flexibly adapt to different business document types.
How to Configure the Settings
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Most power grid enterprise business documents contain complete business processing links. This length retains full context for a single link, avoiding split breaks. |
chunkSplitter | Split by title hierarchy | Adapts to business document multi-level directory structures, retains business process hierarchical relationships, avoiding logical confusion. |
customPdfParse | Enable power term adaptation mode | Adapts to professional terms and unit formats in grid business documents, avoiding parsing errors. |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Covers upload requirements for single large business procedures or batch work order files, avoiding 413 errors. |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Adapts to long document parsing time requirements, avoiding parsing process timeout interruptions. |
chunkSettingMode | Manual configuration | Flexibly adjust parameters for different business document types such as procedures and work orders, adapting to personalized parsing needs. |
Three Common Misconfigurations
- A 413 error appears when uploading a file, and the interface shows file upload failed. The cause is failure to adjust the
UPLOAD_FILE_MAX_SIZEparameter, exceeding the platform's default file size limit. - Some content in Feishu multi-level directories is not parsed and retrieved, and doc files and images in the document cannot be retrieved normally. The cause is failure to enable the
chunkSplitterconfiguration for splitting by title hierarchy, and failure to enable multi-format document parsing adaptation, causing nested levels and non-PDF files to be skipped. - Calling the API to parse a file shows a successful call but no returned results, while manual upload parses normally. The cause is incorrect configuration of external parser key parameters, or failure to carry correct parsing result callback configuration during API calling, causing parsing results to not return to the specified interface.
How to Confirm Proper Configuration
- Upload a single large business procedure file within 1000 MB, confirm no 413 error appears, and parsing progress completes normally without abnormal interruptions.
- Upload a Feishu business document with a multi-level directory, confirm parsed retrieval results include all levels of content, and doc files and embedded images can be retrieved and displayed normally.
- Call the API to upload a business document, confirm returned results include parsed chunked text, with no empty results or timeout prompts.
- Adjust the
chunkSizeparameter, view parsed chunked text, confirm a single chunk retains complete business processing links, with no cross-link splitting.
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.