Skip to content

Werewolf add new models to model proxy #573

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 3 commits into
base: master
Choose a base branch
from
+59 −42
Open

Werewolf add new models to model proxy #573

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
06aac78
add gpt-5.1 and grok-4-1-fast
hannw Nov 20, 2025
8595f90
enable cost analysis
hannw Nov 20, 2025
863411e
add gemini-3-pro-preview
hannw Nov 20, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
20 changes: 19 additions & 1 deletion kaggle_environments/envs/werewolf/harness/litellm_models.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -49,6 +49,9 @@ vertex_ai/gemini-2.5-flash:
vertex_ai/gemini-2.5-flash-lite:
input_cost_per_token: 1e-7
output_cost_per_token: 4e-7
vertex_ai/gemini-3-pro-preview:
input_cost_per_token: 2e-6
output_cost_per_token: 1.2e-5
openai/anthropic/claude-3-5-haiku@20241022:
input_cost_per_token: 8e-7
output_cost_per_token: 4e-6
Expand Down Expand Up @@ -91,6 +94,9 @@ openai/google/gemini-2.5-flash-lite:
openai/google/gemini-2.5-pro:
input_cost_per_token: 1.25e-6
output_cost_per_token: 1e-5
openai/google/gemini-3-pro-preview:
input_cost_per_token: 2e-6
output_cost_per_token: 1.2e-5
openai/meta/llama-3.1-70b:
input_cost_per_token: 1.25e-6
output_cost_per_token: 4e-6
Expand Down Expand Up @@ -118,6 +124,9 @@ openai/openai/gpt-5-mini-2025-08-07:
openai/openai/gpt-5-nano-2025-08-07:
input_cost_per_token: 5e-8
output_cost_per_token: 4e-7
openai/openai/gpt-5.1-2025-11-13:
input_cost_per_token: 1.25e-6
output_cost_per_token: 1e-5
openai/openai/gpt-oss-20b:
input_cost_per_token: 7.5e-8
output_cost_per_token: 3e-7
Expand All @@ -136,6 +145,15 @@ openai/qwen/qwen3-235b-a22b-instruct-2507:
openai/xai/grok-4-fast-reasoning:
input_cost_per_token: 2e-7
output_cost_per_token: 5e-7
openai/xai/grok-4-fast-non-reasoning:
input_cost_per_token: 2e-7
output_cost_per_token: 5e-7
openai/xai/grok-4-0709:
input_cost_per_token: 3e-6
output_cost_per_token: 1.5e-5
output_cost_per_token: 1.5e-5
openai/xai/grok-4-1-fast-reasoning:
input_cost_per_token: 2e-7
output_cost_per_token: 5e-7
openai/xai/grok-4-1-fast-non-reasoning:
input_cost_per_token: 2e-7
output_cost_per_token: 5e-7
81 changes: 40 additions & 41 deletions kaggle_environments/envs/werewolf/werewolf.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -24,7 +24,7 @@
from .game.records import WerewolfObservationModel, get_raw_observation, set_raw_observation
from .game.roles import create_players_from_agents_config
from .game.states import EventName, GameState, get_last_action_request
# from .harness.base import LLMCostTracker, LLMWerewolfAgent
from .harness.base import LLMCostTracker, LLMWerewolfAgent

logger = logging.getLogger(__name__)

Expand All @@ -34,24 +34,24 @@
DEFAULT_BIDDING_PROTOCOL_NAME = "UrgencyBiddingProtocol"


# class AgentCost(BaseModel):
# total_cost: float = 0.0
# prompt_tokens: int = 0
# completion_tokens: int = 0
class AgentCost(BaseModel):
total_cost: float = 0.0
prompt_tokens: int = 0
completion_tokens: int = 0


# class AgentCostSummary(BaseModel):
# agent_config: Dict
# costs: AgentCost = Field(default_factory=AgentCost)
# data: Optional[LLMCostTracker] = None
class AgentCostSummary(BaseModel):
agent_config: Dict
costs: AgentCost = Field(default_factory=AgentCost)
data: Optional[LLMCostTracker] = None


# class CostSummary(BaseModel):
# cost_per_agent: List[AgentCostSummary] = Field(default_factory=list)
# total_cost: float = 0.0
# total_prompt_tokens: int = 0
# total_completion_tokens: int = 0
# total_tokens: int = 0
class CostSummary(BaseModel):
cost_per_agent: List[AgentCostSummary] = Field(default_factory=list)
total_cost: float = 0.0
total_prompt_tokens: int = 0
total_completion_tokens: int = 0
total_tokens: int = 0


_PERCEIVED_THREAT_LEVELS = [item.value for item in PerceivedThreatLevel]
Expand Down Expand Up @@ -413,38 +413,37 @@ def interpreter(state, env):
return state


# Only applicable once the werewolf/harness/ folder is merged
# def collect_cost_summary(env) -> CostSummary:
# cost_summary = CostSummary()
def collect_cost_summary(env) -> CostSummary:
cost_summary = CostSummary()

# for agent_config in env.configuration.agents:
# player_id = agent_config["id"]
# agent_id = agent_config["agent_id"]
for agent_config in env.configuration.agents:
player_id = agent_config["id"]
agent_id = agent_config["agent_id"]

# agent_cost_summary = AgentCostSummary(agent_config=agent_config)
agent_cost_summary = AgentCostSummary(agent_config=agent_config)

# if isinstance(agents.get(agent_id), AgentFactoryWrapper) and issubclass(
# agents[agent_id].agent_class, LLMWerewolfAgent
# ):
# agent_instance = agents[agent_id].get_instance(player_id)
# if agent_instance:
# cost_tracker = agent_instance.cost_tracker
# agent_cost = AgentCost(
# total_cost=cost_tracker.query_token_cost.total_costs_usd,
# prompt_tokens=cost_tracker.prompt_token_cost.total_tokens,
# completion_tokens=cost_tracker.completion_token_cost.total_tokens,
# )
# agent_cost_summary.costs = agent_cost
# agent_cost_summary.data = cost_tracker
if isinstance(agents.get(agent_id), AgentFactoryWrapper) and issubclass(
agents[agent_id].agent_class, LLMWerewolfAgent
):
agent_instance = agents[agent_id].get_instance(player_id)
if agent_instance:
cost_tracker = agent_instance.cost_tracker
agent_cost = AgentCost(
total_cost=cost_tracker.query_token_cost.total_costs_usd,
prompt_tokens=cost_tracker.prompt_token_cost.total_tokens,
completion_tokens=cost_tracker.completion_token_cost.total_tokens,
)
agent_cost_summary.costs = agent_cost
agent_cost_summary.data = cost_tracker

# cost_summary.total_cost += agent_cost.total_cost
# cost_summary.total_prompt_tokens += agent_cost.prompt_tokens
# cost_summary.total_completion_tokens += agent_cost.completion_tokens
cost_summary.total_cost += agent_cost.total_cost
cost_summary.total_prompt_tokens += agent_cost.prompt_tokens
cost_summary.total_completion_tokens += agent_cost.completion_tokens

# cost_summary.cost_per_agent.append(agent_cost_summary)
cost_summary.cost_per_agent.append(agent_cost_summary)

# cost_summary.total_tokens = cost_summary.total_prompt_tokens + cost_summary.total_completion_tokens
# return cost_summary
cost_summary.total_tokens = cost_summary.total_prompt_tokens + cost_summary.total_completion_tokens
return cost_summary


def record_game_end(state, env, game_state, current_info, agent_error):
Expand Down