Environment Inspection and Confirmation

---

Feature Description

After deployment on the SERVICEME platform is completed, the administrator needs to perform environment inspection and confirmation operations to ensure that model configurations, system dependencies, and authorizations are all in a valid state.
This step is a key part of ensuring the stable operation of system functions (such as document recognition, speech recognition, translation, RAG, etc.).

Inspection Scope

Environment inspection mainly includes the following modules:

Inspection Item	Description	Required
Model Set	Check whether it includes models within the standard supported scope (GPT, Embedding, OCR, STT, etc.).	Yes
Model Group	Check whether each model group is configured correctly and available.	Yes
Default Model Setting	Check whether the binding relationship between default models and scenarios is correct.	Yes
System / ENV Environment Variables	Check key system variables and model connection status (such as the availability of OCR, Whisper, and Embedding models).	Yes

Inspection Steps

Open Model Management

Go to Management > Model Management and check the following items one by one:

Model Set

• Confirm whether the following standard models exist:
  • LLM
  • Embedding
• If any are missing, please contact the system administrator to re-import the model set.

Model Group

• Check whether model groups have been configured according to business scenarios, for example:
  • Chat / RAG / Translation / PDF Parsing / OCR, etc.
• Confirm that the models referenced in each model group are consistent with the actual supported scope.

Default Model Setting

• Go to the "Default Model Setting" page and confirm the default bound models item by item (as shown in the example below):
  • translate → gpt-4.1-mini
  • gallery rednote → gpt-4.1-mini
  • recommend config → gpt-4.1-mini
  • gallery chat lead → gpt-4.1
  • optimize prompt → gpt-4.1
  • rag → gpt-4.1
  • i18n translation → gpt-4.1-mini
  • gallery mindmap → gpt-4.1-mini

Note: For tasks with high computational load or higher reasoning requirements (such as knowledge retrieval, complex problem analysis, Prompt optimization, etc.), models with stronger performance should be prioritized;

Note: For lightweight scenarios (such as text translation, summary generation, daily copywriting processing, etc.), models with faster response speed and lower cost can be selected to balance performance and efficiency.

Common Issues and Handling

Issue	Possible Cause	Solution
OCR call failed	API Key is invalid or not configured correctly	Update the key again in the environment variables
Whisper not responding	Model is not enabled or the server side is not deployed	Check the model group configuration and deployment status
Default model setting is empty	License is incomplete or import failed	Confirm the License file and authorization scope
Call latency is too high	Network access to external APIs is unstable	It is recommended to use a model service in the same region as the deployment location