What the data for this category looks like
Coking coal data sources cover public industry reports from a national coal industry association, traded quotes listed on futures exchanges, internal enterprise production and sales ledgers, and capacity and environmental protection policy documents released by regulatory authorities. Update rhythms vary: industry quality data is updated monthly, trading quotes are updated daily, enterprise marketing materials are updated as needed, and policy documents are released in line with regulatory requirements.
Documents are divided into four categories: quality parameter documents with fields including origin, dry basis total sulfur content, colloidal layer thickness, transaction price, and more; marketing script libraries containing scenario-based promotional copy and customer Q&A templates; compliance documents covering capacity control and environmental standards; weekly market reports including regional supply and demand and price fluctuations. Field units uniformly use non-percentage units such as yuan/ton, ten thousand tons, and millimeters.
What constraints do these characteristics impose on knowledge base retrieval and recall
The multi-type document structure and differing update rhythms of the coking coal category require the retrieval system to support both structured field exact matching and unstructured semantic recall. Unique quality parameter fields require recall results to include accurate indicator values, and prevent incorrectly matched fuzzy results. Recall weights must be configured based on timeliness for documents with different update frequencies, to avoid outdated market data being recalled first. Classified storage of marketing scripts and compliance documents requires retrieval to limit recall scope by business scenario, to avoid compliance risks. A document library mixing long and short documents requires an adaptive flexible segmentation strategy to balance retrieval accuracy and processing efficiency.
How to set configurations
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
PARSE_FILE_TIMEOUT_SECONDS | 1200 seconds | Coking coal compliance policy documents are usually lengthy, and single-file parsing takes a long time. This setting adapts to long document parsing requirements |
chunk_size | 800–1000 characters | Coking coal quality parameter documents contain both short fields and long text descriptions. This segmentation length balances field extraction and semantic retrieval needs |
RECALL_TOP_K | Top 8 entries | Coking coal marketing scenarios require a balance between accuracy and coverage, to avoid excessive redundant results interfering with business personnel judgment |
SIMILARITY_THRESHOLD | 0.75–0.85 | The exact matching requirement for coking coal quality parameters requires filtering irrelevant results with low similarity |
UPLOAD_FILE_PRIORITY | Ascending by file size | Most coking coal marketing material documents are short text. Large compliance files can be processed later, adapting to the need for rapid launch of marketing content |
DOCUMENT_COLLECTION_ID | Bound by business scenario | Coking coal marketing requires distinguishing three types of knowledge bases: quality data, script library, and compliance documents. Limit recall scope by collection ID |
The parameter values provided on this page are common recommended starting points for configuration. Actual values are affected by material form, data volume, and business rules. Specific issues require targeted analysis. It is recommended to conduct tests on relevant samples before finalizing settings.
Three common misconfiguration issues
- A 504 Gateway Timeout error occurs when uploading coking coal compliance policy documents. The root cause is failure to adjust the
PARSE_FILE_TIMEOUT_SECONDSparameter. The default timeout duration is insufficient for parsing long documents. - Outdated historical market data is recalled first when retrieving coking coal marketing scripts. The root cause is failure to configure recall weights based on document update time, and failure to distinguish priority between different knowledge base types.
- Marketing script documents are returned when retrieving compliance verification content. The root cause is failure to limit the recall scope via
DOCUMENT_COLLECTION_ID, and failure to isolate knowledge base collections by business scenario.
How to confirm configurations are correctly set
- Upload a single coking coal compliance policy document, check parsing status and duration, and confirm the
PARSE_FILE_TIMEOUT_SECONDSsetting meets document parsing requirements. - Initiate a retrieval with keywords for coking coal quality parameters, verify field matching degree and similarity score of returned results, and confirm the
SIMILARITY_THRESHOLDandRECALL_TOP_Kconfigurations are reasonable. - Initiate retrieval using different
DOCUMENT_COLLECTION_IDvalues, verify only documents from the corresponding knowledge base are returned, and confirm the business scenario isolation configuration takes effect. - Upload multiple coking coal documents of varying sizes, check the processing order of the upload queue, and confirm the
UPLOAD_FILE_PRIORITYconfiguration takes effect.
Question material comes from public community discussions. Configuration values are common starting points and should be measured against your own samples. Verified on 2026-09-14.