Inconsistent environment variable name usage?

[eero-t]

GenAIComps project files consistently use LLM_SERVICE_* environment variable names:

GenAIComps$ git grep '\bLLM_SERV'
README.md:LLM_SERVICE_HOST_IP = os.getenv("LLM_SERVICE_HOST_IP", "0.0.0.0")
README.md:LLM_SERVICE_PORT = os.getenv("LLM_SERVICE_PORT", 9000)
README.md:            host=LLM_SERVICE_HOST_IP,
README.md:            port=LLM_SERVICE_PORT,
comps/cores/mega/exporter.py:            "LLM_SERVICE_HOST_IP": "llm-svc",
tests/cores/mega/mega.yaml:      - LLM_SERVER_HOST_IP: llm-server
tests/cores/mega/mega.yaml:      - LLM_SERVER_PORT: 8005

Whereas half of GenAIExamples files use LLM_SERVICE_*, and half use LLM_SERVER_* env variable names.

GenAIExamples$ git grep -l '\bLLM_SERVICE_' | wc -l
53
GenAIExamples$ git grep -l '\bLLM_SERVER_' | wc -l
59

(There's also similar split for other *_SERV* variables.)

E.g. chatqna.py and chatqna_wrapper.py (added in #1144) use different ones, but nothing in this repo uses/references latter one???

GenAIExamples$ git grep chatqna_wrapper
GenAIExamples$

Why env variable naming is not more consistent, and why repo has code that is not referenced anywhere (tests, Dockerfiles etc)?

[eero-t]

Also, why it's named *_IP instead of just LLM_SERVICE_HOST? While with Docker somebody might use IP addresses, with Kubernetes it's going to be a host name, so the name looks a bit silly to end users.

[eero-t]

Ah, I see. At least in ChatQnA , the *_SERVER_* variants are for the real infererencing servers, and *_SERVICE_* variables are for their (AFAIK nowadays / unused) wrapper services.

DocSum still uses LLM (always) through wrapper service, so it uses LLM_SERVICE_* name prefix. However, also env vars for its other services use this prefix, although they are not wrappers (MEGA_SERVICE_*, ASR_SERVICE_*).

=> While I assume wrapper and the actual inferencing services to have somewhat APIs, I'm not sure how useful the env var distinction is in practice, as no OPEA application sems to use both the wrapper and real service, they use either of them?

[CICD-at-OPEA]

This issue is stale because it has been open 30 days with no activity. Remove stale label or comment or this will be closed in 7 days.

[CICD-at-OPEA]

This issue was closed because it has been stalled for 7 days with no activity.