This document was translated by ChatGPT

1. Prerequisites

Community edition of DeepFlow has been deployed in K8s.

2. Configuring Session Models

Currently, the service supports the following models, which can be enabled as needed through values-custom.yaml configuration:

stella-agent-ce:
  configmap:
    df-llm-agent.yaml:
      ai:
        enable: False
        platforms:
          - platform: 'azure' # https://learn.microsoft.com/zh-cn/azure/ai-services/openai/
            enable: False
            model: 'gpt'
            api_type: 'azure'
            api_key: 'xxx' # FIXME
            api_base: 'xxx' # FIXME
            api_version: 'xxx' # FIXME
            engine_name:
              - 'xxx' # FIXME (model deployment name)
          - platform: 'aliyun' # https://help.aliyun.com/zh/dashscope/create-a-chat-foundation-model
            enable: False
            model: 'dashscope'
            api_key: 'xxx' # FIXME
            engine_name:
              - 'qwen-turbo'
              - 'qwen-plus'
          - platform: 'baidu' # https://cloud.baidu.com/doc/WENXINWORKSHOP/index.html
            enable: False
            model: 'qianfan'
            api_key: 'xxx' # FIXME
            api_secret: 'xxx' # FIXME
            engine_name:
              - 'ERNIE-Bot'
              - 'ERNIE-Bot-turbo'
          - platform: 'zhipu' # https://open.bigmodel.cn/
            enable: False
            model: 'zhipuai'
            api_key: 'xxx' # FIXME
            engine_name:
              - 'chatglm_turbo'

Update DeepFlow:

helm upgrade deepflow -n deepflow -f values-custom.yaml deepflow/deepflow

3. Using in Grafana

The AI model interpretation feature (alpha version) is currently available in DeepFlow Topo Panel and DeepFlow Tracing Panel:

DeepFlow Topo Panel

DeepFlow Tracing Panel

4. Using the API

Method: POST

URL:

• http[s]:😕/{ip}:{port}/v1/ai/stream/{platform_name}?engine={engine_name}
• Parameter Description
  • ip/port: K8s Service of stella-agent, default port is 30831.
  • platform_name: Name of the platform where the model is located, e.g., azure.
  • engine_name: Name of the engine, e.g., DF-GPT4-32K.

Header:

content-type: application/json

Request JSON Body:

{
  "system_content": "You are an expert in web development", # Model role positioning
  "user_content": "How to optimize a web page" # Problem description
}

Response:

Web page optimization refers to improving the performance of a web page through various means, including load speed, response speed, and user experience.

Web page optimization is a complex and comprehensive task that involves multiple aspects, including front-end, back-end, and network. It requires analysis and optimization based on specific situations.

Here are some common web page optimization methods:

Front-end Optimization

Reduce HTTP requests: Reducing HTTP requests can decrease network latency and browser parsing time. Methods to reduce HTTP requests include:
Combining CSS and JavaScript files.
Using sprite images.
Implementing lazy loading.
Optimize resources: Optimizing resources can reduce their size, thereby improving load speed. Methods to optimize resources include:
Compressing images.
Using appropriate image formats.
Minifying CSS and JavaScript files.
Proper use of caching: Caching can reduce repeated network requests, thereby improving load speed. Methods to use caching properly include:
Setting appropriate cache headers.
Using browser caching.
Optimize DOM structure: A reasonable DOM structure can reduce browser parsing and rendering time. Methods to optimize DOM structure include:
Reducing the number and levels of DOM elements.
Avoiding complex CSS selectors.
Optimize JavaScript code: Optimizing JavaScript code can improve code execution efficiency. Methods to optimize JavaScript code include:
Avoiding global variables.
Using caching.
Using appropriate algorithms and data structures.
Back-end Optimization

Optimize database queries: Optimizing database queries can reduce the load on the database server, thereby improving page response speed. Methods to optimize database queries include:
Using appropriate indexes.
Avoiding unnecessary queries.
Using caching.
Optimize application server: Optimizing the application server can improve its performance. Methods to optimize the application server include:
Using appropriate load balancing strategies.
Using caching.
Optimizing code.
Network Optimization

Choose an appropriate CDN: A CDN can distribute content to servers closer to the user, reducing network latency.
Optimize DNS resolution: Optimizing DNS resolution can improve DNS resolution speed. Methods to optimize DNS resolution include:
Using a CDN.
Configuring DNS records.
Using Gzip compression: Gzip compression can reduce the amount of data transmitted, thereby improving load speed.
Tools

Various tools can be used to test and analyze the performance of web pages, such as:

Google PageSpeed Insights
Lighthouse
WebPageTest
By using these tools, you can identify performance bottlenecks in web pages and perform corresponding optimizations.

Web page optimization is a continuous process that requires constant testing and optimization to achieve the best performance.

Here are some additional suggestions:

Always consider performance when developing web pages.
Use performance testing tools to test the performance of web pages.
Monitor the performance of web pages and optimize them regularly.
I hope this information is helpful to you.