Roadmap

[aaa006bd]

This issue addresses the general development plan for the dev-addy aka codecaptain.ai which is a tool that allows a user to interact with a code base and analyze and document in a conversational manner. Mentioned below are the areas that should be addressed:

Roadmap

• [x] #6
• [x] Update the UI
• [x] #10
• [x] allow the user to add private forges
• [x] allow users to add private repos
• [ ] limit repos / messages-per-minute => #43
• [x] #4
• [ ] Improve answering
  • [x] include issue comments as embeddings
  • [x] use chat-history https://python.langchain.com/docs/modules/memory/chat_messages/
  • [x] try gpt-4
  • [ ] output related file if question has a specific answer
  • [ ] use intelligent parsing of code files
  • [ ] include PRs (title and description) & comments
  • [ ] include PRs as embeddings
  • [ ] include PR comments as embeddings
• [ ] handle expired tokens
  • [x] allow users to re-authenticate with forge
• [x] add chats #46
  • [x] show chat list instead of repos list
  • [x] store chat in db
  • [x] open new chat by selecting a configured forge and repo
  • [ ] show summary as title of chat
  • [ ] fork chat
• [ ] #54
  • [ ] manage members
  • [ ] share repos
  • [ ] share openai token
  • [ ] can share chats
  • [ ] share configured forges
• [x] delete repo
• [x] delete forge
  • [x] and all repos of that forge
• [ ] auto-update repos on push using webhooks
• [ ] use incremental repo updates
  • [x] pull if repo already exists
  • [ ] use default branch for pulling (currently always main)
  • [x] only get updated issues

[aaa006bd]

@anbraten please check and edit if needed.

[anbraten]

@aaa006bd Those are some nice ideas. But before we start engineering sth we should think about what nice functions codecaptian should provider for a user. Exp: #3 is about adding a database for chat-history, but how would that be beneficial for a user? Would it for example show old conversations, would it allow auto-completion using old text messages, ...? So I would suggest what we note down what the user gets and then think about the how when dealing with the individual steps.

[anbraten]

Some long-term features I had in mind:

• Automatic code-reviews using the knowledge / embeddings of the repo
• Sending newletter-ish email / create an issue with things the AI suggests you to improve in your project. Somehow like an inspection by some expert consultant, maybe also sth like an security audit / automatic pen-test.
• Auto generated FAQs / Knowledge docs (auto generate system architecture diagram)

[aaa006bd]

@anbraten I have updated the MR. Please take a look and update if you have anything more to address. We can later address the long-term features you mentioned.

[anbraten]

• [ ] fix indexing
• [ ] chats #46
  • [ ] stored in db
  • [ ] open new chat for a configured repo
  • [ ] fork chat
  • [ ] summarize chat text and show as title
• [ ] orgs
  • [ ] share configured forge
  • [ ] share openai key
  • [ ] share chat