Phospho News - Product Update #2
Phospho going open source, A/B tests and much more !!
Hello everyone ! ๐๐งช
Paul, Pierre and Nicolas from phospho here !
Here is a quick overview of some major features we shipped in the last few weeks
Phospho Open Source
Finally ! Phospho is going open source
Give the core analytics component of phospho, the phospho lab, a try, or deploy the whole platform on your own using Docker.
Leverage phospho analytics with more flexibility. With phospho labs, create and run your own jobs. Learn more about it in the documentation here.
Discover the repo and how to self serve it here
Give it a spin ! Contributions and feedback are much appreciated.
Use phospho With Your Own Models
Use your own model to run jobs. Utilize your OpenAI or Mistral API, or even local models with Ollama. Learn more with this cookbook.
A/B Tests
Leverage your own evaluation pipeline with A/B testing. The only thing you need to do is add the version ID (e.g., prompt or model version) in the log, and you're all set.
log = phospho.log( input="log this", output="and that", version_id="YOUR_VERSION_ID")
Get the average success rate and the standard deviation of your experiment directly in the platform or via API.
Learn more about it here
Tests From a Dataset
We've also added the possibility to test your LLM app using a historical or a custom dataset before going into production.
Learn more how to leverage this new feature here.
What's Next ?
Soon, you will be able to directly run analyses in notebooks and build robust data pipelines using phospho. Additionally, labeling using the platform will be easier ! Many exciting updates are coming next week!
If you have any feedback or feature requests on how to make Phospho better, send us an email or join our discord.
Paul-Louis from phospho
Send with ๐๐งช from SF