fix(litellm): explicitly pass encoding_format="float" (#144)

LiteLLM documents encoding_format as defaulting to float, but in
practice some providers (e.g. nvidia_nim) error out when it's not
provided. Pass it explicitly as a workaround.

Author: georgeh0

src/cocoindex_code/litellm_embedder.py

@@ -85,6 +85,7 @@ async def run_embedding_request(self, *, input: list[str], **kwargs: Any) -> Any
             response = await self._aembedding_with_rate_limit_retries(
                 model=self._model,
                 input=input,
+                encoding_format="float",
                 **kwargs,
             )
 

tests/test_litellm_embedder.py

@@ -25,7 +25,7 @@ async def fake_aembedding(*, model: str, input: list[str], **kwargs: Any) -> Any
         attempts += 1
         assert model == "text-embedding-3-small"
         assert input == ["hello"]
-        assert kwargs == {}
+        assert kwargs == {"encoding_format": "float"}
         if attempts == 1:
             raise Exception("Rate limit exceeded. Please try again in 250ms")
         return SimpleNamespace(data=[{"embedding": [1.0, 2.0]}])
@@ -54,7 +54,7 @@ async def fake_sleep(delay: float) -> None:
 
     async def fake_aembedding(*, model: str, input: list[str], **kwargs: Any) -> Any:
         assert model == "text-embedding-3-small"
-        assert kwargs == {}
+        assert kwargs == {"encoding_format": "float"}
         inputs_seen.append(input)
         return SimpleNamespace(data=[{"embedding": [1.0, 2.0]}])