bb7cf6ba20
Summary: This tweaks the prompting so that the user's message is given on its own instead of as a docstring within Python. This is so that the prompt makes sense when: - the user asks a question such as "Can you write me a formula which does ...?" rather than describing their formula as a docstring would, or - the user sends a message that doesn't ask for a formula at all (https://grist.slack.com/archives/C0234CPPXPA/p1687699944315069?thread_ts=1687698078.832209&cid=C0234CPPXPA) Also added wording for the model to refuse when the user asks for something that the model cannot do. Because the code (and maybe in some cases the model) for non-ChatGPT models relies on the prompt consisting entirely of Python code produced by the data engine (which no longer contains the user's message) those code paths have been disabled for now. Updating them now seems like undesirable drag, I think it'd be better to revisit this when iteration/experimentation has slowed down and stabilised. Test Plan: Added entries to the formula dataset where the response shouldn't contain a formula, indicated by the value `1` for the new column `no_formula`. This is somewhat successful, as the model does refuse to help in some of the new test cases, but not all. Performance on existing entries also seems a bit worse, but it's hard to distinguish this from random noise. Hopefully this can be remedied in the future with more work, e.g. automatic followup messages containing example inputs and outputs. Reviewers: paulfitz Reviewed By: paulfitz Subscribers: dsagal Differential Revision: https://phab.getgrist.com/D3936
1.6 KiB
Using Large Language Models with Grist
In this experimental Grist feature, originally developed by Alex Hall, you can hook up OpenAI's ChatGPT to write formulas for you. Here's how.
First, you need an API key. Visit https://openai.com/api/ and prepare a key, then
store it in an environment variable OPENAI_API_KEY.
That's all the configuration needed!
Currently it is only a backend feature, we are still working on the UI for it.
Hugging Face and other OpenAI models (deactivated)
Not currently available, needs some work to revive. These notes are only preserved as a reminder to ourselves of how this worked.
To use a different OpenAI model such as code-davinci-002 or text-davinci-003,
set the environment variable COMPLETION_MODEL to the name of the model.
Alternatively, there are many non-proprietary models hosted on Hugging Face.
At the time of writing, none can compare with OpenAI for use with Grist.
Things can change quickly in the world of AI though. So instead of OpenAI,
you can visit https://huggingface.co/ and prepare a key, then
store it in an environment variable HUGGINGFACE_API_KEY.
The model used will default to NovelAI/genji-python-6B for
Hugging Face. There's no particularly great model for this application,
but you can try other models by setting an environment variable
COMPLETION_MODEL to codeparrot/codeparrot or
NinedayWang/PolyCoder-2.7B or similar.
If you are hosting a model yourself, host it as Hugging Face does,
and use COMPLETION_URL rather than COMPLETION_MODEL to
point to the model on your own server rather than Hugging Face.