AI Core
AI Core is the shared AI service layer for QTO ERP.
It is backend-mounted and supports provider configuration, prompts, inference, logging, and response caching. No dedicated top-level React module was verified in the current shell.
Providers
Section titled “Providers”AI provider configuration supports:
- OpenAI
- Google Gemini
- Anthropic Claude
- Azure OpenAI
- AWS Bedrock
- vLLM
- Ollama
- DeepSeek
- Custom OpenAI-compatible endpoints
Provider records include name, provider type, encrypted API key, base URL, default model name, active state, priority, default temperature, default token limit, and provider-specific extra settings.
Prompt templates
Section titled “Prompt templates”Prompt templates include:
- Name
- Description
- System prompt
- User prompt
- Version
- Active state
- Default provider
- Metadata
When a new prompt version is created for an existing prompt name, the previous version is deactivated.
Inference
Section titled “Inference”AI Core exposes asynchronous and synchronous inference endpoints plus a connection test endpoint.
Logs and cache
Section titled “Logs and cache”AI request logs store provider, prompt, user, originating module, request and response data, token usage, latency, status, error message, cache flag, and task ID.
AI response cache stores hashed request content, module, prompt name, request text, response text, hit count, and access timing.
Statuses
Section titled “Statuses”AI request statuses include:
| Status | Meaning |
|---|---|
| Success | The request completed successfully. |
| Error | The request failed. |
| Timeout | The provider request timed out. |
| Rate Limited | The provider rejected or delayed the request because of limits. |
| Parsing Failed | The response could not be parsed as expected. |