ci: test safety with starter

We are now testing the safety capability with the starter image. This
includes a few changes:

* Enable the safety integration test
* Relax the shield model requirements from llama-guard to make it work
  with llama-guard3:8b coming from Ollama
* Expose a shield for each inference provider in the starter distro. The
  shield will only be registered if the provider is enabled. Shields
  will be added if the provider claims to support a safety model
* Missing providers models have been added too
* Pointers to official documentation pages for provider models support
  have been added

Closes: https://github.com/meta-llama/llama-stack/issues/2528
Signed-off-by: Sébastien Han <seb@redhat.com>
This commit is contained in:
Sébastien Han 2025-07-04 17:07:11 +02:00
parent cd0ad21111
commit 11c912da0a
No known key found for this signature in database
20 changed files with 621 additions and 126 deletions

View file

@ -44,6 +44,7 @@ def build_hf_repo_model_entry(
]
if additional_aliases:
aliases.extend(additional_aliases)
aliases = [alias for alias in aliases if alias is not None]
return ProviderModelEntry(
provider_model_id=provider_model_id,
aliases=aliases,
@ -90,7 +91,7 @@ class ModelRegistryHelper(ModelsProtocolPrivate):
# embedding models are always registered by their provider model id and does not need to be mapped to a llama model
provider_resource_id = model.provider_resource_id
if provider_resource_id:
if provider_resource_id != supported_model_id: # be idemopotent, only reject differences
if provider_resource_id != supported_model_id: # be idempotent, only reject differences
raise ValueError(
f"Model id '{model.model_id}' is already registered. Please use a different id or unregister it first."
)