What This Category's Data Looks Like
Carbon steel financial report data primarily comes from exchange public announcements, industry news platforms, and third-party data terminals. The core update cycle follows quarterly and annual reports. Quarterly reports are released within 45 days after the end of each quarter. Annual reports must be disclosed by the end of April of the following year. Temporary announcements for production volume or cost adjustments are released as needed.
The document structure includes two main parts: core financial indicators and production and operation data. Core fields include crude steel output, comprehensive cost per ton of steel, operating revenue, and attributable net profit, with units of ten thousand tons, yuan per ton, hundred million yuan, and hundred million yuan respectively.
Constraints Imposed on Workflow Orchestration
Differences in formats across multiple data sources require configuring multiple types of document parsing nodes in the workflow. These nodes support plain text extraction from PDF announcements and field mapping for third-party structured data.
Quarterly batch data triggers require setting scheduled scheduling nodes. These nodes limit processing to only the latest quarterly data to avoid reloading historical reports.
Uniformity of units for industry-specific fields requires adding a field validation step in the workflow. This step intercepts abnormal data where ton steel cost is listed in yuan, and uniformly converts it to the industry-standard unit of thousand yuan.
Unstructured content in temporary announcements requires configuring industry-specific term dictionaries to improve the accuracy of keyword extraction.
Configuration Settings
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
multi_knowledge_base_enable | true | Carbon steel financial report data is scattered across three independent knowledge bases: announcements, industry news, and financial databases, requiring simultaneous recall of relevant content |
document_parse_timeout | 600 seconds | A single annual PDF announcement has over 150 pages, and the default parsing duration is insufficient to complete full extraction |
keyword_extract_threshold | 0.75 | The extraction confidence of industry-specific terms for carbon steel such as "comprehensive cost per ton of steel" and "crude steel capacity utilization rate" must exceed the general threshold |
global_variable_pass_mode | all_module | Variables extracted in the workflow such as crude steel output and gross profit per ton of steel must be passed to subsequent MCP tool nodes |
rag_recall_top_k | Top 8 entries | Core indicators of carbon steel financial reports are scattered across 3-5 paragraphs, requiring sufficient recall of context fragments to ensure analysis completeness |
workflow_trigger_cron | 0 0 10 5,20 * * | Matches the regular windows of 5 and 20 days after quarterly report disclosure, triggering workflow execution on a scheduled basis |
The parameter values provided on this page are common recommendations for starting point configuration. Actual values are affected by material form, data volume, and business rules. Specific issues require case-by-case analysis, and it is recommended to test on your own samples before finalizing settings.
Three Common Configuration Mistakes
- Issue: The workflow returns an "undefined global variable" error after execution, and subsequent nodes cannot read the extracted crude steel output field. Cause:
global_variable_pass_modeis not configured asall_module, only allowing partial nodes to access global variables. - Issue: The token field passed during MCP tool invocation is empty, preventing completion of interface requests. Cause:
global_variable_pass_modeis not configured to allow cross-module passing, or the token variable is not bound to the corresponding tool parameter position. - Issue: Extracted keywords include "steel price" but do not include "carbon steel price", resulting in insufficient keyword matching accuracy. Cause: No carbon steel industry-specific term dictionary is loaded, and the keyword extraction threshold is set too low, causing general vocabulary to take priority in matching.
How to Verify Correct Configuration
- Navigate to the workflow configuration page, check if the
multi_knowledge_base_enableswitch is turned on, and confirm that the three knowledge bases of announcements, industry news, and financial databases have been associated. - Test the parsing task for a single annual PDF announcement, check the parsing log to confirm that the duration does not exceed the configured
document_parse_timeoutvalue, and there are no parsing failure errors. - Trigger a test workflow, check the global variable panel to confirm that fields such as extracted crude steel output and gross profit per ton of steel have been generated, and can be selected for reference in subsequent MCP tool nodes.
- Check the scheduled trigger configuration, confirm that the
workflow_trigger_cronexpression matches the regular time window for quarterly report disclosure, and there are no conflicting settings for repeated triggers.
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.