What Data for This Category Looks Like
This content typically covers daily documents for process, equipment, quality, and production departments of crystalline silicon cell manufacturers. Examples include process cards, diffusion and coating parameters, printing and sintering records, EL inspection images, IV grading data, and yield reports. Data sources include silicon wafer incoming materials, MES systems, diffusion furnaces, coating equipment, printing presses, EL and IV testing equipment, and laboratory data. Updates follow each batch of materials, each wafer inspection result, and daily production reports. Documents include fields such as process parameters, inspection values, and grading standards. Some data uses specific units, such as degrees Celsius for diffusion temperature and percentage for conversion efficiency.
Constraints on Model Access and Configuration
Data sources are scattered and update frequencies vary. Model access must support multi-format process documents and inspection data. This prevents retrieval failures caused by format mismatches. The accuracy of process parameters and inspection results directly impacts production quality. Model-retrieved content must be complete and unambiguous. Configuration must balance context length and response rigor. Wafer-level inspection data has a large volume. Reasonably control the number and length of retrieved documents to avoid exceeding the model's processing limit. Parameter windows for different processes must be accurately linked. Configuration must support targeted retrieval of corresponding documents based on business scenarios.
How to Configure Settings
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
maxContext | 8000–12000 token | Adapts to the overall length of photovoltaic process documents, avoiding truncation of critical process parameters and inspection records |
temperature | 0.1–0.3 | Process standard Q&A requires rigor. Lower temperature values reduce ambiguous or non-standard outputs |
quoteMaxToken | 2000–3000 characters | Fully retrieves core content of single documents such as EL inspection image descriptions and IV grading data |
maxTokens | 4000–6000 token | Meets the length requirements for process question responses, covering parameter explanations, process descriptions, and other content |
AIPROXY_API_ENDPOINT | Fill in according to the official access address of the selected model provider | Establishes the basic network connection for model calls, ensuring requests can be delivered normally |
AIPROXY_API_TOKEN | Fill in according to the key assigned by the provider | Completes identity verification for model calls, ensuring interface access permissions |
Three Common Misconfigurations
- The symptom is that the knowledge base can retrieve data, but there is no output after passing it to the large model. The cause is incorrect configuration of the
maxContextparameter. The retrieved process documents exceed the context limit and are truncated, so they cannot participate in model inference. - The symptom is that the index model cannot be called normally after being enabled. The cause is incorrect entry of
AIPROXY_API_ENDPOINTorAIPROXY_API_TOKEN, or the parameters contain extra spaces or format errors. - The symptom is that offline configuration reranking model fails after upgrading to FastGPT v4.17.0. The cause is failure to load the reranking model dependency package for the corresponding version, or incorrect configuration of the model file path during offline deployment.
How to Verify Proper Configuration
- Call the test interface to verify the validity of
AIPROXY_API_ENDPOINTandAIPROXY_API_TOKEN. Confirm there are no identity verification or network access errors. - Upload a single photovoltaic process card document, trigger a Q&A test related to process parameters, and check whether the returned results include complete process requirements and standard values.
- Test query requirements involving EL inspection and IV grading. Confirm that the number and length of returned reference documents meet the configuration limits of
quoteMaxToken. - Check the system operation logs to confirm there are no error messages such as model call timeouts or parameter format errors, verifying the stability of the configuration.
The values provided here are common starting points and should be measured against the reader's own samples.