Skip to content

Agent Zero AI framework

fix: keep gateway model ids intact for OpenAI-compatible embeddings

Md. Asiqur Rahman Khan655056985602
diff --git a/models.py b/models.py
index e1fa5c423..62f1db31d 100644
--- a/models.py
+++ b/models.py
@@ -795,7 +795,7 @@ class LiteLLMEmbeddingWrapper(Embeddings):
model_config: Optional[ModelConfig] = None,
**kwargs: Any,
):
- self.model_name = f"{provider}/{model}" if provider != "openai" else model
+ self.model_name = f"{provider}/{model}"
self.kwargs = kwargs
self.a0_model_conf = model_config
diff --git a/tests/test_model_config_api_keys.py b/tests/test_model_config_api_keys.py
index eb779aba1..99e25a799 100644
--- a/tests/test_model_config_api_keys.py
+++ b/tests/test_model_config_api_keys.py
@@ -723,6 +723,52 @@ def test_local_provider_runtime_defaults_and_overrides(monkeypatch):
assert custom_vllm_chat.kwargs["api_key"] == "real-local-key"
+def test_openai_compatible_embedding_keeps_gateway_model_string(monkeypatch):
+ """Gateway model ids must reach LiteLLM with an explicit `openai/` provider.
+
+ An OpenAI-compatible gateway owns its own model namespace, so ids such as
+ `nvidia/...` or `auto/embedding` are model names, not provider prefixes.
+ Without the prefix LiteLLM re-parses the first segment as a provider and
+ either raises "LLM Provider NOT provided" or routes to the wrong provider,
+ mangling the model id before the gateway ever sees it.
+ """
+ monkeypatch.setattr(models, "get_api_key", lambda provider: "None")
+
+ gateway = "https://gateway.example/v1"
+ for model in (
+ "nvidia/llama-nemotron-embed-vl-1b-v2:free",
+ "openrouter/openai/text-embedding-3-small",
+ "auto/embedding",
+ ):
+ embedding = models.get_embedding_model(
+ "other", model, api_base=gateway, api_key="gateway-key"
+ )
+ assert embedding.model_name == f"openai/{model}"
+ assert embedding.kwargs["api_base"] == gateway
+
+ # The bundled OpenRouter embedding provider is affected the same way: it
+ # resolves to litellm_provider `openai` against OpenRouter's api_base, so an
+ # OpenRouter-style id has to survive intact. Previously `openai/<model>` was
+ # handed to LiteLLM bare, which consumed the `openai/` segment and forwarded
+ # only `<model>` to OpenRouter.
+ for model in (
+ "openai/text-embedding-3-small",
+ "nvidia/llama-nemotron-embed-vl-1b-v2:free",
+ ):
+ openrouter_embedding = models.get_embedding_model("openrouter", model)
+ assert openrouter_embedding.model_name == f"openai/{model}"
+ assert openrouter_embedding.kwargs["api_base"] == "https://openrouter.ai/api/v1"
+
+ # Plain OpenAI behaviour is unchanged: LiteLLM strips the `openai/` prefix
+ # and forwards the bare model id, exactly as it did without a prefix.
+ openai_embedding = models.get_embedding_model("openai", "text-embedding-3-small")
+ assert openai_embedding.model_name == "openai/text-embedding-3-small"
+
+ # Providers that do not resolve to `openai` keep their existing prefix.
+ ollama_embedding = models.get_embedding_model("ollama", "nomic-embed-text")
+ assert ollama_embedding.model_name == "ollama/nomic-embed-text"
+
+
def test_embedding_config_repairs_sentence_transformer_aliases(monkeypatch):
from plugins._model_config.helpers import model_config