Subscribe
Sign in
CodiumAI โ Generating meaningful tests for busy devs
CodiumAI analyzes your Python, JS, or TS code and generates meaningful tests to catch bugs before you ship. With CodiumAI, you can easily and quickly generate comprehensive test suites that help you ensure the reliability and correctness of your software.
122
Replies
Best
@dean_soc , definitely. Probably next in line. Although, recently, on our discord server we got more requests for Ruby..
@benln , thank you Ben!
We are lucky to have such an excellent Hunter!
We are just Gathering feedback here :)
Hey guys, I'm with the CodiumAI team,
I made a few videos to introduce some of our features. Just a taste of what CodiumAI can do :)
Basic showcase: https://www.loom.com/share/658d0...
Basic configuration options: https://www.loom.com/share/22f2d...
Reference tests: https://www.loom.com/share/a607c...
@shai_g thanks! Great share!
One of our users also created a video showing with an excellent example of using CodiumAI to generate tests for the Gilded Rose kata (typescript):
Congrats on the launch!
This looks super useful as we have a growing Python codebase and I'm not always doing the best job when it comes to writing tests due to lack of time...
Have installed Codium and will give it a go ๐๐ฟ
What are the next steps for this project?
@ed_forson
Hi Eddie.
In a week or two, we plan to upgrade the Pycharm GUI, and add capabilities that are currently only in our VS-Code extension. For example, auto-mocks, example input, example test, give-me-more,...
So if you are a Python-PyCharm user, stay tuned :-)
Under the hood, a lot is going on. We will definitely continue to improve the quality of our tests, enable them to have a larger context, and add capabilities and features.
@ed_forson Thank you, Eddie! I hope our product will benefit you. Next, we are planning to support more languages, and more IDEs, we will, of course, keep improving the test results (getting feedback from users will certainly help!) and we are planning to add Team's plan with more advanced capabilities.
I invite you to register for our newsletter and stay tuned https://www.codium.ai/#home-wait ๐ป
@ed_forson , thank you for asking!
1> We wish to make sure that almost any type of code (different frameworks, different languages) is well supported by CodiumAI.
2> We want to improve our UX/UI according to feedback we hear from our dear users.
These are the top urgent items in the roadmap.
In our seed funding announcement, we share a longer roadmap of our plans!
https://www.codium.ai/blog/codiu...
Congratulations @maritamar on your launch! Brilliant idea to leverage TestGPT to generate test cases, catch bugs, and check code. Looking forward to putting this through the paces!
@sfali789 , thank you! We really want to help developer to make code logic testing a less tedious and frustrating task... it could even be a fun experience.
Congrats Codium team! Testing code is very tedious and absolutely necessary. Simplifying the process reducing the time to deploy code in production is a real need in every start up out there.๐ฅ
@mehdi_rifai , wow, I really appreciate your message, it exactly describes what we are trying to help with. Please do send over any additional feedback and suggestions, thank you!
Congratulations on the launch of Codium! This tool seems to be very useful for developers to streamline their test cases, detect errors, and verify code. I will put it to the test with my team.
@yenire_leal Thank you Yenire! I would love to hear about your experience with us and get your feedback :)
@nevoalva Hi Nevo, Thank you for your question. We constantly benchmark on HumanEval Test Generation Task among other benchmarks internal benchmarks. On HumanEval Test Generation, CodiumAI's TestGPT delivers tests with higher Branch Coverage compared to simple Test completion with GPT3.5-Turbo (97.3% vs 92%) and a higher test pass rate of 97% vs 61%.
Congratulations @maritamar on launching Codium! ๐ It seems like an amazing tool that can help busy developers generate meaningful tests effortlessly. I love the idea of getting non-trivial tests suggested right inside the IDE, making it easier for developers to code smart, create more value and stay confident when pushing their code. The fact that Codium analyzes code, docstring, and comments, and interacts with you to suggest tests is really impressive. It's great to see that it provides full visibility of how the code behaves and how the changes affect the rest of the code.
And by the way, I like the landing page.๐ช
@officialexaking , thank you Jim! We hope to meet your expectations, let us know about anything you have in mind: suggestions, issues, requests..
Congratulations on the launch of CodiumAI! It sounds like a very promising tool that can save developers a lot of time and frustration when it comes to testing their code.
@tajul_islam1 , thank you! Exactly, we want to help getting rid of this frustration. As developers, we get it -- we also suffered from it for years!
Hey all, you must see this funny video.
Old Coder Guy uses and teases CodiumAI to test his code logic ๐คฃ
He also makes fun of me, but I'm totally fine with that ๐
Writing tests could be a frustrating task, so pouring in some fun and artificial intelligence is the way to go!
Good job!
https://twitter.com/itamar_mar/s...
https://www.tiktok.com/@oldcoder...
Generating test cases using AI sounds interesting, writing test cases is a challenge. Congratulations team codium.
@karan_bhansali , agree! Generating meaningful tests is a challenging task, even with the latest AI capabilities! And we are serious about it.
We're so excited to announce our first milestone today, and we'll keep adding capabilities and features, and improving our product accordingly.
Great launch, as a startup we tend to make the mistake of feeling too busy to code in tests so this will be helpful
@david_glass , agree!! we hope that the experience of using CodiumAI is so ... empowering, that developers won't wait for the last moment, if it all, to write and run code-level testing.
That's interesting! I'm not a tech person, but as Product Manager I know that `testing` is one the most crucial steps in shipping a product/feature. Glad if there's a tool to help the engineers (and testers) to prevent a `hotfix/rollback` required once a bug/defect found in the production env. Good luck!
@razii_abraham , thank you!
By the way, in the long term, we hope to introduce the concept of "executable specification". In such a future, Product Managers will also be able to test code behavior. You can read a bit about it here:
https://www.codium.ai/blog/codiu...
@maritamar - Congrats on the launch! It is exciting to see AI-driven test generation. Can you share more about how CodiumAI handles edge cases and unique scenarios that might not be immediately obvious from the code, docstrings, or comments?
@salil_sethi , thank you!
With CodiumAI, which is already available in VS Code for example:
1) The algorithm runs in an iterative process, like a chain of thoughts, aiming to think about happy/general cases as well as edge cases. You will see a tag in the test case description.
2) CodiumAI various interfaces enable the developer to advise the algorithm by providing input data (e.g. input for class parameters), test examples, or general instructions. With these interfaces, the developer can steer CodiumAI to generate more edge cases.
Check it out, and let us know what you think!
In the future, we will integrate with various additional sources of data, e.g. to Jira.
Amazing stuff guys! taking off the hassle of actually writing tests (and way more). love it. @maritamar
@tal_peretz , thank you!!
Exactly! Enough with hating tests, it is time to love code logic testing.
Well, it was quite frustrating to write component tests, but with CodiumAI the process becomes interactive and pleasant. And we just started, more features and capabilities around the corner :)
Congratulations on an incredible launch today on Product Hunt! Your hard work and dedication have paid off, and it's inspiring to see your product gain such recognition. I'm thrilled to see your vision come to life and can't wait to see what the future holds for your product. Wishing you all the best on this exciting journey!
It is tough to write tests for your own code. I am too lazy to do it most of the time. Love the product.
Congratulations on the launch!
@isha_varma07 , thank you Isha. We hope to make it much faster and more enjoyable for you to generate a meaningful test suite. Let us know how it worked!
@realmaxrush , thank you! We work closely with our partner OpenAI. Our system and model TestGPT-1 is based on a combination of gpt-3.5-turbo (a.k.a ChatGPT-Turbo) and GPT-4.
We will share more technical details later this year, when we release TestGPT-2
Great product and congratulations on the launch. Some questions:
1. Does it have built in support for in memory DBs to mimic the actual calls ?
2. Are there any plans to support writing integration tests by chaining calls together ? As a developer, I would love to see something which only needs the chaining instructions (in a yaml, json) and have it produce end to end tests
@kushagra_sharma3 Hi Kushagra,
Thank you! Very good questions.
1. The way CodiumAI can currently help with this is by learning from your existing tests. If you already have tests that work this way, you can provide them as reference tests and Codium will pick up the initialization and style. Please see this video for an example (with JavaScript. Python behaves the same way): https://www.loom.com/share/a607c.... Does that help in your use case?
2. At this point in time we don't have specific plans for such a feature, but I would be very happy to hear more about the use-case! What type of integration tests were thinking about? Tests that run through a standard test runner like pytest / jest, and make HTTP request to a configured backend?
Thanks,
Shai
@shai_codium
1. Alright, makes sense
2. Historically I have seen integration tests where calls are either chained between modules/files(like Unit tests) or chained as http calls to the backend sitting outside the service(s). We have seen that CodiumAI can create tests for something like create an order, cancel of an order, refund an order.
An integration test would typically be something like Create an order -> Pick inventory(validate) -> Generate shipping label(validate) -> Ship an order(validate). The orchestration of calls is the key here and there can be a number of such call flows. Looking for a low code/no code solution for this
@kushagra_sharma3
Hi, sorry for the delay,
Thank you for the info!
I think in this case CodiumAI might be able to help - but we are currently focused on test-code generation, so not as a low code/no code solution, only as standard code-based tests.
Thank you again,
Shai
Qodo (formerly CodiumAI)