fix: ensure enough tokens for structured output in vLLM test (#591) (#595)

planetf1 · web-flow · commit ac6a4cf1e2b8 · 2026-03-06T17:16:43.000Z
diff --git a/test/backends/test_openai_vllm.py b/test/backends/test_openai_vllm.py
@@ -222,6 +222,7 @@ class Answer(pydantic.BaseModel):
         actions=[CBlock(value=prompt) for prompt in prompts],
         format=Answer,
         ctx=m_session.ctx,
+        model_options={ModelOption.MAX_NEW_TOKENS: 256},
     )
 
     assert len(results) == len(prompts)

Original file line number	Diff line number	Diff line change
`@@ -222,6 +222,7 @@ class Answer(pydantic.BaseModel):`
`222`	`222`	`actions=[CBlock(value=prompt) for prompt in prompts],`
`223`	`223`	`format=Answer,`
`224`	`224`	`ctx=m_session.ctx,`
	`225`	`+ model_options={ModelOption.MAX_NEW_TOKENS: 256},`
`225`	`226`	`)`
`226`	`227`
`227`	`228`	`assert len(results) == len(prompts)`