What this category's data looks like
Daily business documents for this category include sales orders, master production schedules, material requirement plans, weekly production schedules, work orders, material shortage lists, and delivery change notices. Data sources cover ERP orders and inventory, BOMs and process routes, equipment capacity, staff shifts, work in progress, and supplier arrivals. Update cadences include daily, weekly, or triggered by orders or exceptional events. Document fields include order number, material code, process labor hours, equipment number, number of shifts, and others. Units include hours, pieces, dates, and more. This document is written for FastGPT v4.17.0.
What constraints data characteristics impose on model access and configuration
Dispersed data sources and inconsistent update frequencies require model access configurations to support dynamic synchronization of multi-source business data. Diverse document structures and specialized fields require clear field mapping rules in configurations to ensure Q&A accuracy. Scheduling tasks involve multi-dimensional resource constraints, so single-round Q&A must cover extensive context information, which imposes requirements on the model's context window and token limits. Data updates triggered by exceptional events require configurations to have fast-response calling mechanisms to avoid delays in scheduling adjustments.
How to set configurations
| Configuration Item | Recommended Value Range | Rationale |
|---|---|---|
maxContext | 8000–12000 token | This category's data includes multiple work orders, production schedules and other documents. Single-round process Q&A needs to cover at least 3 complete business documents to support accurate judgments |
maxTokens | 1500–2000 token | Process specification and technical standard Q&A needs to output detailed scheduling adjustment suggestions to avoid truncation of key parameters and operation steps |
quoteMaxToken | 3000–4000 token | Long document fragments such as BOMs and process routes need to be cited as Q&A basis to ensure complete cited content |
temperature | 0.1–0.3 | Process standard Q&A needs to strictly match established specifications to avoid generating content that deviates from standards |
AIPROXY_API_ENDPOINT | Fill in according to the official endpoint of the accessed model | Ensure the routing path of model call requests is correct, avoid requests failing to reach the target model |
AIPROXY_API_TOKEN | Fill in the authorization key of the corresponding model | Complete identity verification for model access, ensure valid call permissions |
Three common misconfigurations
- Symptom: Models configured on third-party channels do not appear in the model selection dropdown menu of the FastGPT Q&A module. Cause: The corresponding channel has not been bound in FastGPT's model access configuration, or the business permissions for the production planning and shop scheduling category have not been enabled.
- Symptom: Q&A response results are truncated, with key scheduling parameters or process steps missing. Cause: The value of
maxTokensis too small, and does not adapt to the overall length requirements of process specifications and scheduling data. - Symptom: Model calls return prompts that the channel does not exist or authentication has failed. Cause: The
AIPROXY_API_TOKENconfiguration is incorrect, or the channel permission activation process for the target model has not been completed.
How to confirm configuration is complete
- Navigate to the FastGPT model management page, check whether the status of the accessed model is shown as normal and available.
- Import a real weekly production schedule as context, launch a test Q&A that includes material requirement queries, check whether the returned results can accurately extract information from the corresponding fields.
- Adjust the
temperatureparameter to 0.2 and launch a test, confirm that the rigor of the returned results meets the requirements of process standards. - Trigger a simulated order change request, check whether the model can normally load updated business data and complete Q&A.
The values provided here are common starting points, and should be evaluated against the user's own samples.