Overview
- Apache OzHera(incubating) Intelligence Platform is responsible for generating intelligent-related functionalities.
- Currently, features are under development.
- Planned features to be developed:
- Root cause analysis, including trace, alerts, and topology-based intelligent analysis at the application level.
Module Introduction
ozhera-intelligence-server
- Service to initiate intelligent functionalities.
- Access to the ozhera-intelligence-server requires the x-token to be included in the header for authorized access to ozhera's data.
ozhera-intelligence-service
- LogService: Retrieves log data
- MetricsService: Retrieves monitoring metric data
- TraceService: Retrieves trace data
Important Dependencies
It relies on a Xiaomi open-source Agent builder platform--m78
Intelligent Workflow (Sequence Diagram)
LLM Environment Variables
Currently, the LLM class is used for invoking large language models (LLMs). The required environment variables (env vars) are defined in the LLMProvider class.
Prerequisite:
A base LLM_PROVIDER env var must be set to specify the desired LLMProvider.
Example:
To use DeepSeek, follow these steps:
Set LLM_PROVIDER=DEEPSEEK.
Refer to the LLMProvider class for DeepSeek's specific requirements. In this case, DEEPSEEK_API_KEY is mandatory. Configure the API key:
DEEPSEEK_API_KEY=xxx-xxxxx
Recommended Models
Given that this use case involves large-scale data analysis with significant logical reasoning and analytical processing, it is advisable to leverage high-performance inference-optimized models or models with strong code-generation capabilities available on the market.
Suggested Criteria:
- High accuracy in complex reasoning tasks.
- Robust contextual understanding for analytical workflows.
- Efficient handling of long-context prompts (if applicable).
Examples include models fine-tuned for reasoning-intensive workloads or those excelling in programmatic problem-solving.
