haolingdong-msft / azure-dataplane-sdk-helper Goto Github PK

Tasks:

1. Push code - Xiaofei
2. Apply embedding models to JS and promopt enigneering - Mary
3. Apply embedding models to Java and promopt enigneering - Xiaofei

• Issues:

1. Where to put the code? pr
   pr: preferred solution, user does not need to deal with local environment issues.

• User Story:
  bot: Welcome!
  user: I want to generate js sdk.
  -> bot will need to identify this is a generation task. And it will ask the user to provide generation related information.
  bot: Could you provide spec repo link for your service? e.g. https://github.com/Azure/azure-rest-api-specs/tree/main/specification/cognitiveservices/OpenAI.Inference
  user: https://xxxx
  bot: Let me generate sdk for you, please wait 1 min...
  -> bot will call tool to generate sdk and create pr, and create codespace link. make sure user can commit the code to pr using codespace.
  bot: Here is the generated sdk: https://xxx/pr/123, you can tryout and edit the code here: http://codespace.com/xxx/123
  bot: Your next steps will be: 1. enhance readme.md 2. add samples 3. add tests
  user: How to enhance readme?
  -> bot will need to identify this is Q/A task, and provide information.
  bot: readme requirmenet doc link: xxx
  [user open codespace and commit code]
  user: How to add sampels?
  bot: sample link: https://xxx
  user: How to add tests?
  bot: test ref doc : https://xxx
  [user open codespace and commit code]
  user: I've added samples and tests. Could you check?
  bot: Let me review your pr. Please wait 2 min
  -> pr review. engineering + AI.
  bot: I added several comments in the pr. please take a look.
  user: I've updated according to your comments.
  bot: Great, it looks good to me. Feel free to make it as ready to review, our reviewers will review.

• Tasks:

1. bot will need to identify this is a generation task. And it will ask the user to provide generation related information. (P0)
2. bot will need to identify this is Q/A task, and provide information. (P0)
3. bot will call tool to generate sdk and create pr, and create codespace link. make sure user can commit the code to pr using codespace. (P0)
4. pr review. engineering + AI. (P1)

• Todo:

1. chatgpt classification -> Xiaofei
2. investigation: codespace, create pr -> Mary
3. chatgpt Q/A enhancement
4. implementation on generate, create pr url and codespace url
5. pr review. engineering + AI. (P1) -> Haoling and Jiahao to investigate tools

Todos:

1. Can we build codespace on top of the VM that generate sdk, codespace environment set up? - Mary
2. chatgpt code and call generate tool - Xiaofei
3. create pr to commit local/servier side code - Xiaofei
4. chatgpt code and add code review comments for readme - Haoling
5. Q/A prompt --> build index and embedding - Mary
6. use bot to create pr and add review comments

Next steps:

1. Have a VM that installs all the environments and able to generate sdk
2. Deploy teams chatbot to VM and able to run generate tool
3. Review other files: tests, samples.

Suppose we are from service team. We are new to the SDK code generator and we want to generate SDK from our existing swagger file.

1. How to generate Java(.net, js) SDK? -> autorest or typespec(default) command
2. What's the next step? -> add test case, run test, generate sample, draft PR, etc.
3. Specific questions for next steps, e.g. How to add test case, how to run test, etc.
4. How to solve issues? tsp command not found? (use doc prerequisite to answer).

Demo scenario:

1. Code generation tool, executable instructions(or execute for user)
2. PR review (integrated with github pipeline, @bot Please help review)

Todos:

1. Can we build codespace on top of the VM that generate sdk, codespace environment set up? -> Done
2. chatgpt code and call generate tool - Xiaofei
3. create pr to commit local/servier side code - Xiaofei
4. chatgpt code and add code review comments for readme - Haoling
5. Q/A prompt --> build index and embedding - Mary
6. use bot to create pr and add review comments

Next steps:

1. Have a VM that installs all the environments and able to generate sdk
2. Deploy teams chatbot to VM and able to run generate tool
3. Review other files: tests, samples.

You are an AI assistant that helps to review markdown files. I will give you the markdown file. You will need to give code review comments. The markdown file contains descriptions of an Azure service, code snippets and reference links. You need to verify the structure of the file, whether the description is fluent in natural language perspective, the code snippet provided in the file is correct and easy to understand. You will need to list the review comments with the code line.

Bot requirement:

1. guide user how to create sdk step by step
2. error guider
3. Sample / test quality

Tasks:

1. Teams bot/app development - Xiaofei
2. Investigation on prompt - Mary
3. Investigation on embedding: - Jiahao, Xiaofei

• https://github.com/LGDoor/asa-wiki-teams-bot
• https://tprompt-docs.azurewebsites.net/gettingstarted/index.html
  user -> bot -> search engine -> search result -> bot -> chatgpt -> answer

4. Integrate all languages

ARD Hackthon (July 28th) Delivery

1. Bot demo
2. Slides