✨ Add support for google gemini model.

[Myrausman]

Fixes #304

[jwmatthews]

@Myrausman thank you!

Quick look at the PR and looked great. I'll help run through a manual test, but I don't expect to have time to help until Wednesday.

[fabianvf]

@Myrausman Would you be able to start the server in demo mode (DEMO_MODE=true), which should dump a bunch of cached responses into kai/data/vcr/{app}/{model}, then commit those files? That will let us run some verification on our end without needing to have a valid API key

[Myrausman]

I started server in demo mode but I'm not getting any cache new response for that particular model (pro-gemini).

[jwmatthews]

I started server in demo mode but I'm not getting any cache new response for that particular model (pro-gemini).

@Myrausman did you run:
https://github.com/konveyor/kai/blob/main/example/run_demo.py

To get cached responses you need to send requests to the API, so that it can cache the response.
To send requests, easiest way is to use the run_demo.py against coolstore.

You can read this to learn more of how to run:
https://github.com/konveyor/kai/blob/main/example/README.md

[jwmatthews]

Code changes look good...
Waiting on:

1. Update of cached responses
2. A maintainer to run a manual test to confirm working with Gemini

I am willing to test with a Gemini API key, I won't have time to help with that today, will try for helping tomorrow.

[jwmatthews]

@Myrausman this will help to give more context of updating cached responses

https://github.com/konveyor/kai/blob/main/docs/contrib/Demo_Mode.md#demo_mode-updating-cached-responses

[Myrausman]

followin this file Contributing.md, when I'm running make load-data I am getting these error:

(.venv) myra@myra-ThinkPad-T14-Gen-2i:~/extra/lfx/kai$ make load-data
PYTHONPATH="/home/myra/extra/lfx/kai/kai:" python ./kai/service/incident_store/psql.py  --config_filepath ./kai/config.toml --drop_tables False
Console logging for 'kai' is set to level 'INFO'
File logging for 'kai' is set to level 'DEBUG' writing to file: '/home/myra/extra/lfx/kai/kai/service/incident_store/../../../logs/kai_psql.log'
INFO - 2024-08-22 02:58:32,264 - kai.service.kai_application.kai_application - [  kai_application.py:54   -             __init__()] - Tracing enabled.
modelid ############################### gemini-pro
INFO - 2024-08-22 02:58:32,281 - kai.service.kai_application.kai_application - [  kai_application.py:63   -             __init__()] - Selected provider: ChatGoogleGenerativeAI
INFO - 2024-08-22 02:58:32,282 - kai.service.kai_application.kai_application - [  kai_application.py:64   -             __init__()] - Selected model: gemini-pro
INFO - 2024-08-22 02:58:32,389 - kai.service.kai_application.kai_application - [  kai_application.py:84   -             __init__()] - Selected incident store: postgresql
INFO - 2024-08-22 02:58:32,460 - kai.service.incident_store.incident_store - [   incident_store.py:91   - load_reports_from_directory()] - Loading incident store with applications: ['ejb-security', 'ejb-remote', 'kitchensink', 'bmt', 'helloworld-mdb', 'cmt', 'greeter', 'tasks-qute', 'ticket-monster']

INFO - 2024-08-22 02:58:32,461 - kai.service.incident_store.incident_store - [   incident_store.py:95   - load_reports_from_directory()] - Loading application ejb-security

Traceback (most recent call last):
  File "/home/myra/extra/lfx/kai/./kai/service/incident_store/psql.py", line 4, in <module>
    cmd("postgresql")
  File "/home/myra/extra/lfx/kai/kai/service/incident_store/incident_store.py", line 475, in cmd
    load_reports_from_directory(incident_store, args.analysis_dir_path)
  File "/home/myra/extra/lfx/kai/kai/service/incident_store/incident_store.py", line 114, in load_reports_from_directory
    repo = Repo(repo_path)
           ^^^^^^^^^^^^^^^
  File "/home/myra/extra/lfx/kai/.venv/lib/python3.11/site-packages/git/repo/base.py", line 236, in __init__
    raise NoSuchPathError(epath)
git.exc.NoSuchPathError: /home/myra/extra/lfx/kai/samples/sample_repos/ejb-security
make: *** [Makefile:22: load-data] Error 1

[jwmatthews]

I think you just need to clone a few sample repos paths

Look at this link and run the fetch_apps.py it mentions:

https://github.com/konveyor/kai/blob/main/docs/contrib/Dev_Environment.md#steps

[Myrausman]

After running python3 run_demo.py, I am getting the following output: https://gist.github.com/Myrausman/3b9112e1342baa5a9accb8cb31d913d6.
I also don't see any new files from the model in this location: kai/data/vcr/{app}/{model}.

[jwmatthews]

The requirements.txt update looks incorrect.

@Myrausman did you run pip-compile, you should have more dependencies added to requirements.txt to account for the langchain-google-genai package and it's dependencies.

Let's update requirements.txt and then I think we'll be in shape to merge this.

I confirmed that when I ran pip-compile to update requirements.txt I had success running run_example.py.

I also confirmed that cached data is NOT being captured. I am less sure of why that is happening.

My thought is we move forward to merge this PR after the requirements.txt update is made, we accept we do not have cached data for gemini and work that as a separate issue, tracked as #315

[fabianvf]

interesting, I wonder if the google library is using a different library that isn't intercepted by VCR for managing the requests

[jwmatthews]

@Myrausman please take look at the merge conflicts we will need to resolve before we can merge. We merged another PR yesterday that altered the requirements a little, so you will need to reexamine requirements.in, add your changes in, rerun pip-compile and test, then update this PR with the changes.

[jwmatthews]

LGTM

Ran and verified functional on Mac with arm64

[JonahSussman]

@Myrausman Thank you for your contribution!