Show HN: Codel – Autonomous Open Source AI Developer Agent(github.com)
github.com
Show HN: Codel – Autonomous Open Source AI Developer Agent
https://github.com/semanser/codel
16 comments
For anyone looking for the prompts: https://github.com/semanser/codel/tree/main/backend/template...
I am having trouble getting it to work properly on macOS, I think due to Docker API version mismatch between the Go client and Docker Desktop.
I added a comment to one of your existing issues on GH. Happy to connect and help debug. Do you have a Discord server or anything?
I added a comment to one of your existing issues on GH. Happy to connect and help debug. Do you have a Discord server or anything?
Interesting. I saw one more person reporting a similar issue so I will prioritize fixing this. I just created a new discord server for everyone who would like to join: https://discord.gg/uMaGSHNjzc
This looks fantastic! Thanks for building.
Given that single projects are not that large, is there a reason you chose Postgres? I imagine using sqlite would reduce the installation complexity of the solution.
Given that single projects are not that large, is there a reason you chose Postgres? I imagine using sqlite would reduce the installation complexity of the solution.
Thanks!
The project is using gorm (gorm.io) so it should be easy to be able to support sqlite as well since all the queries should be db agnostic. I want to make installation/usage as easy as possible so wrapping everything into a docker container and/or using another DB is an option for sure. I will check what is possible to do :)
It's not listed in the prerequisites but this requires an openai account. If they are doing all the work, what exactly does the rest of this code do?
“OpenAI do all the work” is a very wrong claim in this case.
It’s true that most of the complexity are solved by using LLM, but it’s not everything. There are still a good amount of work needed to be done if you want to build an agent (or even an AI wrapper, if you’re implying it).
It’s true that most of the complexity are solved by using LLM, but it’s not everything. There are still a good amount of work needed to be done if you want to build an agent (or even an AI wrapper, if you’re implying it).
Have you eval'd SWEbench yet?
Not yet, but it's definitely in the short term plans.
Is it possible to use it with a local model via ollama f.ex?
Not yet. I added a ticket to track progress for that: https://github.com/semanser/codel/issues/16.
doesn't seem to work as per readme instructions.
yeah, I already pushed a fix for this. Thanks!
https://github.com/semanser/codel/issues/1
https://github.com/semanser/codel/issues/1