scriptoriumadmin/litellm
1
0
Fork 0
mirror of https://github.com/BerriAI/litellm.git synced 2026-10-14 04:09:09 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 97

test(rust): support automatic Vertex tool selection

Browse source
This commit is contained in:
Yujong Lee 2026-10-10 12:16:57 -07:00
parent c35fa4427d
commit d5b5ebbf2f
2 changed files with 2 additions and 2 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
litellm-rust/crates/inference-messages/tests/live/vertex_ai.rs
View file

@ -62,7 +62,7 @@ async fn tool_round_trip(route: MessagesRoute, #[case] stream: bool) {
route,
PROVIDER,
stream,
json!({"tool_choice": {"type": "tool", "name": "echo"}}),
json!({"tool_choice": {"type": "auto"}}),
)
.await;
}

2
litellm-rust/crates/inference-testing/README.md
View file

@ -27,7 +27,7 @@ LITELLM_LIVE_MESSAGES_VERTEX_AI_MODEL=<claude-model-on-vertex> \
For Vertex AI, set `LITELLM_LIVE_MESSAGES_VERTEX_AI_MODEL` to a Claude model published in the project's location. Credentials, project and location resolve through the production Rust GCP authentication, from `GOOGLE_APPLICATION_CREDENTIALS` or `VERTEXAI_CREDENTIALS`, with `VERTEXAI_PROJECT` and `VERTEXAI_LOCATION`
Vertex AI cases live in `inference-messages/tests/live/vertex_ai.rs`. They cover plain and system messages, text streaming, and complete and streaming tool round trips
Vertex AI cases live in `inference-messages/tests/live/vertex_ai.rs`. They cover plain and system messages, text streaming, and complete and streaming tool round trips. Tool cases use automatic selection because current Claude models reject forced `tool_choice`. Each case still requires a tool call, checks its arguments, and sends the tool result back
Tool cases require a model that supports forced `tool_choice`, except where a provider's section says otherwise. A provider rejection fails the selected case instead of skipping it