verifiers 0.1.15.dev13__tar.gz → 0.1.15.dev14__tar.gz

{verifiers-0.1.15.dev13 → verifiers-0.1.15.dev14}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: verifiers
-Version: 0.1.15.dev13
+Version: 0.1.15.dev14
 Summary: Verifiers: Environments for LLM Reinforcement Learning
 Project-URL: Homepage, https://github.com/primeintellect-ai/verifiers
 Project-URL: Documentation, https://github.com/primeintellect-ai/verifiers

{verifiers-0.1.15.dev13 → verifiers-0.1.15.dev14}/tests/test_v1_textarena_taskset.py

@@ -198,6 +198,50 @@ async def test_textarena_user_steps_env_and_stops_when_game_finishes(fake_textar
     assert state["stop_condition"] == "textarena_done"
 
 
+@pytest.mark.asyncio
+async def test_textarena_user_accepts_structured_assistant_content(fake_textarena):
+    _, fake_ta = fake_textarena
+    taskset = textarena.TextArenaTaskset(
+        config=textarena.TextArenaTasksetConfig(
+            game="FakeWordle-v0",
+            answer_state_key="secret_word",
+            num_train_examples=1,
+            num_eval_examples=0,
+        )
+    )
+    task = taskset.to_task(
+        vf.Task(
+            {
+                "example_id": 0,
+                "prompt": [],
+                "answer": "apple",
+                "textarena": {
+                    "game": "FakeWordle-v0",
+                    "answer_state_key": "secret_word",
+                },
+            }
+        )
+    )
+    state = vf.State.for_task(task)
+    state["completion"] = [
+        vf.AssistantMessage(
+            content=[
+                vf.TextContentPart(text="I will guess "),
+                vf.TextContentPart(text="<guess>[apple]</guess>."),
+            ]
+        )
+    ]
+
+    env = vf.Env(taskset=taskset, harness=vf.Harness(config=vf.HarnessConfig()))
+    state = await env.harness.setup_state(task, state)
+    messages = await env.harness.runtime.user_messages(task, state)
+    ta_env = fake_ta.envs[-1]
+
+    assert ta_env.guesses == ["[apple]"]
+    assert messages == [{"role": "user", "content": "Solved."}]
+    assert state["stop_condition"] == "textarena_done"
+
+
 @pytest.mark.asyncio
 async def test_textarena_user_returns_wordle_feedback_for_unfinished_game(
     fake_textarena,

{verifiers-0.1.15.dev13 → verifiers-0.1.15.dev14}/verifiers/__init__.py

@@ -1,4 +1,4 @@
-__version__ = "0.1.15.dev13"
+__version__ = "0.1.15.dev14"
 
 import importlib
 import os