What data looks like for this category
Typical business data sources for the general mechanical equipment assembly category include PLM/ERP systems, part inspection data, assembly tool collection terminals, test benches, MES systems, and customer technical agreements. Data is updated once per device or per order. Daily documents include assembly process cards, multi-level BOMs, torque records, clearance adjustment records, commissioning reports, factory inspection certificates, and packing lists. Data fields include BOM hierarchy numbers, torque values (unit N·m), clearance values (unit mm), concentricity tolerances, test load parameters, and more. There is no unified standardized field format, so multi-type documents must be associated by device.
What constraints these characteristics impose on deployment and upgrade
Deployments must support incremental synchronization configuration. This avoids excessive resource usage from full synchronization, as data is updated per device or per order and associated with multiple source systems.
Multi-level BOMs require association with multiple types of inspection records. This necessitates configuring field mapping rules for multi-source data.
Document formats vary widely, and individual document sizes differ greatly. Parsing thresholds for different document types must be adapted.
Data fields include specific engineering units. Unit validation rules must be configured during deployment to prevent unit mismatch issues during retrieval.
When upgrading versions, database index compatibility must be checked. This prevents full-text search functionality failure caused by version updates.
How to set configurations
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
MONGODB_URI | mongodb://fastgpt:pass123@mongodb-host:27017/assembly_qms?authSource=admin | Stores structured data such as multi-level BOMs and torque records, supports full-text search and associated queries |
PG_URL | postgresql://fastgpt:pass456@rds-host:5432/assembly_qms?sslmode=require | Stores unstructured document metadata such as assembly process cards and factory inspection certificates, supports complex conditional filtering |
REDIS_URL | redis://:redis-pass@redis-host:6379/0 | Caches frequently accessed device BOM snapshots and assembly process parameters to reduce database query load |
STORAGE_S3_ENDPOINT | http://oss-cn-shanghai.aliyuncs.com | Stores large-volume documents such as commissioning reports and packing lists, compatible with Alibaba Cloud OSS or local MinIO storage |
| Single document parsing limit | 10–20 pages per run | Adapts to batch parsing of multi-page documents such as assembly process cards and multi-level BOMs, avoids parsing timeouts |
Three common mistakes
- Phenomenon: A
text index required for $text queryerror may occur during knowledge base retrieval after FastGPT is upgraded to v4.17.0. Cause: The full-text index of MongoDB was not manually rebuilt during the upgrade process, and the index initialization command was not configured in the deployment script. - Phenomenon: FastGPT deployed on Alibaba Cloud ECS cannot connect to Alibaba Cloud RDS database, with a connection timeout prompt. Cause: The VPC network segment to which the ECS belongs was not added to the RDS whitelist, and the intranet access address was not configured in
PG_URL. Using a public network address resulted in permission interception. - Phenomenon: When calling the FastGPT API for supplier audit, the returned results do not associate the inspection records of the corresponding device. Cause: The access permission scope of
ROOT_KEYwas not configured, causing the API to fail to obtain the metadata of the corresponding device, or the field mapping rule did not associate the association fields between BOM and inspection records.
How to confirm correct configuration
- Upload an assembly process card to the FastGPT knowledge base, confirm that fields such as BOM hierarchy, torque values, and clearance values are correctly extracted after parsing, with no parsing failure prompts.
- Perform a supplier audit retrieval, enter "coaxiality inspection records for reducer assembly", confirm that the returned results include test data of the corresponding device, with no index errors or empty results.
- Check system logs, confirm that the connection status of
MONGODB_URI,PG_URL, andREDIS_URLis normal, with no authentication failure or timeout logs. - Call the FastGPT API interface, pass the device number and audit query instruction, confirm that the returned results are associated with the correct assembly documents and inspection data, with no permission errors.
The values provided are common starting points and should be measured against the reader's own samples.