fix: migrate to @grafana/llm package for OpenAI functionality #1052
Conversation
|
Use the following command to run this PR with Docker at http://localhost:3000: |
adamyeats
force-pushed
the
use-grafana-llm-for-openai
branch
from
2ef0d07
to
8e51cb8
Compare
|
|
The @grafana/experimental package is being deprecated so development of LLM-related functionality has been moved to a separate package. The API is exactly the same so this _should_ just work. I've had to fix the resolution of some Grafana packages which were otherwise being duplicated because the @grafana/llm package has quite a loose dependency range; without this, we had both version 10.4.1 and 11.3.0 of the various Grafana packages (and all of their subdependencies) which was causing tests to fail. Fixes #1051.
sd2k
force-pushed
the
use-grafana-llm-for-openai
branch
from
8e51cb8
to
d9427ac
Compare
|
I've had to fix the resolution of some Grafana packages which were otherwise being duplicated because the @grafana/llm package has quite a loose dependency range; without this, we had both version 10.4.1 and 11.3.0 of the various Grafana packages (and all of their subdependencies) which was causing tests to fail. |
|
@sd2k, I'm not overly keen on adding in the resolutions. Can the dependencies for the LLM package be made stricter? Alternatively, we can look at updating the |
|
Yeah I'm not really keen on it either, I was surprised that the relaxed restrictions caused the resolver to choose two separate versions rather than preferring a common version 😕 Upgrading to 11+ would probably do the trick and feels like the right call, especially given those packages are injected by Grafana anyway. |
I think this is fine now that the correct versions are in the yarn.lock...
The @grafana/experimental package is being deprecated so development
of LLM-related functionality has been moved to a separate package.
The API is exactly the same so this should just work.
Fixes #1051.