Update model path in pytest configuration to use gpt-oss-120b (#61)

* add --port arg to ep logs

* Fix WebSocketManager to reset broadcast task after cancellation

* simple tests work

* TODO: TestLogsServer

* TODO: TestLogsServerIntegration

* TODO: test HTML injection
- also test TestAsyncWebSocketOperations

* add logs server tests

* add port parameter testes

* use gpt-oss-120b to avoid rate limits

* point to port 8000 for dev

* woops

* fix "uvicorn eval_protocol.utils.logs_server:create_app --factory --reload"

* use gpt-oss-120b since less rate limiting (#57)

* Aggregated metrics part 7 (#58)

* use gpt-oss-120b for less rate limits and faster tests

* fix typeerror

* Refactor LogsServer event handling and improve integration tests

- Moved event_bus.start_listening() to the correct location in LogsServer to ensure it starts listening during the broadcast loop.
- Updated integration tests to use multiprocessing for server startup and improved health check validation.
- Enhanced test_create_app_factory to be asynchronous and added necessary imports for better clarity.

* Enhance test_create_app_factory to verify LogsServer start_loops call

- Updated the test_create_app_factory to mock and assert that the start_loops method of LogsServer is called during app creation.
- Ensured the test remains asynchronous and maintains clarity in its assertions.

* fix

* use active logger

* cohort -> experiment

* vite build

* Update model path in pytest configuration to use gpt-oss-120b for improved performance

Author: Dylan Huang

tests/pytest/test_pytest_mcp_config.py

@@ -20,7 +20,7 @@
         ]
     ],
     rollout_processor=default_agent_rollout_processor,
-    model=["fireworks_ai/accounts/fireworks/models/kimi-k2-instruct"],
+    model=["fireworks_ai/accounts/fireworks/models/gpt-oss-120b"],
     mode="pointwise",
     mcp_config_path="tests/pytest/mcp_configurations/mock_discord_mcp_config.json",
 )