What the data for this category looks like
Refinery category financial report data mainly comes from exchange public periodic disclosures, enterprise investor relations section announcements, and third-party structured financial report databases. Update cadence follows regulatory requirements: annual reports are updated once per year, semi-annual reports every six months, and quarterly reports every quarter. Document structures include core fields such as refining segment revenue, crude oil processing volume, refined oil and chemical product output, and unit processing cost. Units for crude oil processing volume and output fields are mostly ten thousand tons, revenue fields use hundred million yuan, and unit processing cost uses yuan per ton. Some enterprises provide supplementary monthly production operation data.
What constraints do these characteristics impose on deployment and upgrade?
The fixed update cycle and bulk document properties of refinery financial reports require configuring timeout and concurrency parameters for batch parsing during deployment, to avoid large document parsing timeouts. Specific industry fields and unit requirements need preset field extraction rules in knowledge base configuration. During upgrade, retain old field mapping logic to prevent historical configuration from becoming invalid. Temporary supplementary monthly operation data needs flexible upload support; do not overwrite configured custom parsing templates during upgrade. Segmentation processing for long-text annual reports needs adjustment of segmentation parameters to match the average length of industry documents, to avoid content truncation that impacts analysis results.
How to set configurations
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
PARSE_FILE_TIMEOUT_SECONDS | 600–900 seconds | Refinery financial report documents are mostly long-text, single-file parsing takes longer. This parameter prevents large file parsing interruptions |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Single annual financial report documents have large file sizes, this adapts to large file upload requirements |
RECALL_CHUNK_NUM | Top 8–12 entries | Core fields of refinery financial reports are concentrated in the first few matched document segments. Excessive recall increases context redundancy |
PARSE_SEGMENT_LENGTH | 1200–1500 characters | Financial report documents contain technical terms and long sentences. This segmentation length preserves semantic integrity and avoids splitting technical terms |
INDEXING_BATCH_SIZE | 50–80 documents/batch | When importing financial report documents in bulk, this parameter balances indexing efficiency and server load, preventing indexing stagnation in docker deployments |
HTTP_TOOL_OUTPUT_SHOW | Enabled | Refinery financial report analysis workflows often call industry data interfaces via HTTP tools. Enabling this displays intermediate results from interface returns in conversations |
The parameter values provided on this page are common starting points for configuration. Actual values are affected by material format, data volume and business rules. Specific issues require case-by-case analysis. It is recommended to test on your own samples before finalizing settings.
Three common misconfigurations
- Phenomenon: When using HTTP requests for orchestration, tool return output content is not displayed in conversation results. Cause: The
HTTP_TOOL_OUTPUT_SHOWconfiguration item is not enabled, so intermediate results returned by the tool are not included in the conversation context. - Phenomenon: In docker deployment environments, knowledge base indexing tasks remain in a running state with no progress updates. Cause: The
INDEXING_BATCH_SIZEconfiguration is too large, exceeding server memory and CPU load thresholds, causing indexing process blocking. - Phenomenon: After upgrading to version v4.8.17, charts generated by financial report analysis cannot be displayed normally, and the interface returns an error prompt. Cause: Old field mapping configurations were not retained after upgrade, and dedicated fields for refinery financial reports were not correctly identified, resulting in missing data sources required for chart rendering.
How to confirm configuration is correct
- Upload a test refinery financial report document, check if the parsed fields include industry-specific core fields, to confirm parsing rules are effective.
- Initiate a financial report analysis conversation, check if intermediate results from HTTP tool calls are displayed normally, to confirm configuration items are effective.
- Upload multiple financial report documents in bulk, observe progress updates of indexing tasks, to confirm indexing parameters do not exceed server load thresholds.
- Try importing old workflow files, check if workflow nodes load completely, to confirm workflow-compatible configurations are effective.
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.