What data for this category looks like
Data for this category originates from daily documents produced by process and quality departments at foundries and die casting workshops. These documents include casting process cards, batching lists, heat records, mold histories, first-piece inspection records, material test reports, and nonconforming product review forms. Data sources cover ERP material master data, melting furnace thermometers, spectrometers, MES heat records, mold management systems, and customer drawings and technical specifications. Updates occur per heat, per batch, or daily. Core fields include alloy batching ratios, melting temperatures, pouring parameters, defect types, and corresponding heat and mold identification numbers.
Constraints on deployment and upgrade workflows
High update frequency and dense per-document fields require adaptation to high-frequency, small-batch data write requirements during deployment, with sufficient database connection and cache resources reserved. Documents from diverse sources have significant format differences, requiring compatibility with multi-format data parsing during upgrade. Strict unit and precision requirements apply to core business fields. Unified field mapping rules must be implemented during deployment to avoid parameter conversion errors. Data association relies on unique identifiers such as heat and mold numbers, so association indexes must be configured during deployment to ensure query efficiency for cross-review.
Configuration guidelines
This configuration is based on FastGPT v4.17.0. Specific configuration items and recommendations are listed below:
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
DB_MAX_LINK | 30–50 | Adapts to concurrent demands of high-frequency per-heat data writes, and prevents database connection exhaustion |
MONGODB_URI | Points to a standalone sharded MongoDB instance | Stores frequently updated unstructured documents such as heat records and mold histories, to ensure read and write performance |
REDIS_URL | Configured as a locally intranet-deployed Redis service | Caches frequently queried heat parameters and mold information, to reduce database access pressure |
STORAGE_S3_ENDPOINT | Points to an enterprise intranet object storage service address | Stores large files such as material test reports and customer drawings, to support scenarios without public network deployment |
FE_DOMAIN | Fill in the access domain name after intranet penetration | Adapts to configuration requirements for accessing FastGPT deployed on Feiniu NAS via FRP penetration |
ROOT_KEY | Generate a random string in accordance with enterprise internal permission rules | Ensures access permission security for the system backend, and prevents unauthorized operations |
Three common configuration errors
- Symptom: The knowledge base fails to recognize Markdown first- and second-level titles during local deployment, but parses them correctly during public network deployment. Cause:
FILE_TOKEN_KEYis not configured correctly, leading to insufficient file parsing permissions, or the local deployment file storage path lacks full read-write permissions, which disrupts structured extraction of Markdown text. - Symptom: The error "Reached the max retries per request limit" appears when connecting to an external vector database. Cause: The
DB_MAX_LINKconfiguration value is too low, or no intranet access whitelist is configured, resulting in exhausted request retry attempts. - Symptom: FastGPT accessed via FRP penetration cannot call external document review services. Cause: The penetration domain name configured for
FE_DOMAINdoes not have corresponding port permissions enabled, or the intranet penetration configuration does not map the port for the document parsing service.
How to confirm successful configuration
- Run a single-heat data import test, review database connection logs, and confirm that the concurrent connection count for
DB_MAX_LINKdoes not trigger the upper limit. - Upload a process card document that includes Markdown titles, and verify that the knowledge base correctly extracts title hierarchy information.
- After configuring intranet penetration, access
FE_DOMAINvia a local browser, and confirm that the knowledge base management interface loads normally. - Initiate a cross-review query for alloy batching ratios, and verify that the returned field precision matches the original document.
The values provided here are common starting points, and should be evaluated against the reader’s own samples.