Feature/rag metering (#75)

* Add metering-rag for RAG token metrics

* Make metering executable

* Add metering-rag to Prom metrics

* Metricate the right queue

* Update templates to include metering-rag
This commit is contained in:
cybermaggedon 2024-09-29 20:15:03 +01:00 committed by GitHub
parent 6d57b7a6f3
commit 90127dfecb
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
19 changed files with 305 additions and 0 deletions

View file

@ -132,5 +132,37 @@ local prompt = import "prompt-template.jsonnet";
},
"metering-rag" +: {
create:: function(engine)
local container =
engine.container("metering-rag")
.with_image(images.trustgraph)
.with_command([
"metering",
"-p",
url.pulsar,
"-i",
"non-persistent://tg/response/text-completion-rag-response",
])
.with_limits("0.5", "128M")
.with_reservations("0.1", "128M");
local containerSet = engine.containers(
"metering-rag", [ container ]
);
local service =
engine.internalService(containerSet)
.with_port(8000, 8000, "metrics");
engine.resources([
containerSet,
service,
])
},
} + prompt