twitch_streaming / agents.py
Jofthomas's picture
Jofthomas HF Staff
Update agents.py
ba5a866 verified
import os
import json
import asyncio
import random
# --- OpenAI ---
from openai import AsyncOpenAI, APIError
# --- Google Gemini ---
from google import genai
from google.genai import types
# --- Mistral AI ---
from mistralai import Mistral
# --- Poke-Env ---
from poke_env.player import Player
from poke_env.environment.battle import Battle
from poke_env.environment.move import Move
from poke_env.environment.pokemon import Pokemon
from typing import Optional, Dict, Any, Union
# --- Helper Function & Base Class ---
def normalize_name(name: str) -> str:
"""Lowercase and remove non-alphanumeric characters."""
return "".join(filter(str.isalnum, name)).lower()
STANDARD_TOOL_SCHEMA = {
"choose_move": {
"name": "choose_move",
"description": "Selects and executes an available attacking or status move.",
"parameters": {
"type": "object",
"properties": {
"move_name": {
"type": "string",
"description": "The exact name or ID (e.g., 'thunderbolt', 'swordsdance') of the move to use. Must be one of the available moves.",
},
},
"required": ["move_name"],
},
},
"choose_switch": {
"name": "choose_switch",
"description": "Selects an available Pokémon from the bench to switch into.",
"parameters": {
"type": "object",
"properties": {
"pokemon_name": {
"type": "string",
"description": "The exact name of the Pokémon species to switch to (e.g., 'Pikachu', 'Charizard'). Must be one of the available switches.",
},
},
"required": ["pokemon_name"],
},
},
}
# --- OpenAI Tools Schema (with 'type' field) ---
OPENAI_TOOL_SCHEMA = {
"choose_move": {
"type": "function",
"function": {
"name": "choose_move",
"description": "Selects and executes an available attacking or status move.",
"parameters": {
"type": "object",
"properties": {
"move_name": {
"type": "string",
"description": "The exact name or ID (e.g., 'thunderbolt', 'swordsdance') of the move to use. Must be one of the available moves.",
},
},
"required": ["move_name"],
},
}
},
"choose_switch": {
"type": "function",
"function": {
"name": "choose_switch",
"description": "Selects an available Pokémon from the bench to switch into.",
"parameters": {
"type": "object",
"properties": {
"pokemon_name": {
"type": "string",
"description": "The exact name of the Pokémon species to switch to (e.g., 'Pikachu', 'Charizard'). Must be one of the available switches.",
},
},
"required": ["pokemon_name"],
},
}
},
}
class LLMAgentBase(Player):
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
self.standard_tools = STANDARD_TOOL_SCHEMA
self.battle_history = []
def _format_battle_state(self, battle: Battle) -> str:
active_pkmn = battle.active_pokemon
active_pkmn_info = f"Your active Pokemon: {active_pkmn.species} " \
f"(Type: {'/'.join(map(str, active_pkmn.types))}) " \
f"HP: {active_pkmn.current_hp_fraction * 100:.1f}% " \
f"Status: {active_pkmn.status.name if active_pkmn.status else 'None'} " \
f"Boosts: {active_pkmn.boosts}"
opponent_pkmn = battle.opponent_active_pokemon
opp_info_str = "Unknown"
if opponent_pkmn:
opp_info_str = f"{opponent_pkmn.species} " \
f"(Type: {'/'.join(map(str, opponent_pkmn.types))}) " \
f"HP: {opponent_pkmn.current_hp_fraction * 100:.1f}% " \
f"Status: {opponent_pkmn.status.name if opponent_pkmn.status else 'None'} " \
f"Boosts: {opponent_pkmn.boosts}"
opponent_pkmn_info = f"Opponent's active Pokemon: {opp_info_str}"
available_moves_info = "Available moves:\n"
if battle.available_moves:
available_moves_info += "\n".join(
[f"- {move.id} (Type: {move.type}, BP: {move.base_power}, Acc: {move.accuracy}, PP: {move.current_pp}/{move.max_pp}, Cat: {move.category.name})"
for move in battle.available_moves]
)
else:
available_moves_info += "- None (Must switch or Struggle)"
available_switches_info = "Available switches:\n"
if battle.available_switches:
available_switches_info += "\n".join(
[f"- {pkmn.species} (HP: {pkmn.current_hp_fraction * 100:.1f}%, Status: {pkmn.status.name if pkmn.status else 'None'})"
for pkmn in battle.available_switches]
)
else:
available_switches_info += "- None"
state_str = f"{active_pkmn_info}\n" \
f"{opponent_pkmn_info}\n\n" \
f"{available_moves_info}\n\n" \
f"{available_switches_info}\n\n" \
f"Weather: {battle.weather}\n" \
f"Terrains: {battle.fields}\n" \
f"Your Side Conditions: {battle.side_conditions}\n" \
f"Opponent Side Conditions: {battle.opponent_side_conditions}"
return state_str.strip()
def _find_move_by_name(self, battle: Battle, move_name: str) -> Optional[Move]:
normalized_name = normalize_name(move_name)
# Prioritize exact ID match
for move in battle.available_moves:
if move.id == normalized_name:
return move
# Fallback: Check display name (less reliable)
for move in battle.available_moves:
if move.name.lower() == move_name.lower():
print(f"Warning: Matched move by display name '{move.name}' instead of ID '{move.id}'. Input was '{move_name}'.")
return move
return None
def _find_pokemon_by_name(self, battle: Battle, pokemon_name: str) -> Optional[Pokemon]:
normalized_name = normalize_name(pokemon_name)
for pkmn in battle.available_switches:
# Normalize the species name for comparison
if normalize_name(pkmn.species) == normalized_name:
return pkmn
return None
async def choose_move(self, battle: Battle) -> str:
battle_state_str = self._format_battle_state(battle)
decision_result = await self._get_llm_decision(battle_state_str)
print(decision_result)
decision = decision_result.get("decision")
error_message = decision_result.get("error")
action_taken = False
fallback_reason = ""
if decision:
function_name = decision.get("name")
args = decision.get("arguments", {})
if function_name == "choose_move":
move_name = args.get("move_name")
if move_name:
chosen_move = self._find_move_by_name(battle, move_name)
if chosen_move and chosen_move in battle.available_moves:
action_taken = True
chat_msg = f"AI Decision: Using move '{chosen_move.id}'."
print(chat_msg)
return self.create_order(chosen_move)
else:
fallback_reason = f"LLM chose unavailable/invalid move '{move_name}'."
else:
fallback_reason = "LLM 'choose_move' called without 'move_name'."
elif function_name == "choose_switch":
pokemon_name = args.get("pokemon_name")
if pokemon_name:
chosen_switch = self._find_pokemon_by_name(battle, pokemon_name)
if chosen_switch and chosen_switch in battle.available_switches:
action_taken = True
chat_msg = f"AI Decision: Switching to '{chosen_switch.species}'."
print(chat_msg)
return self.create_order(chosen_switch)
else:
fallback_reason = f"LLM chose unavailable/invalid switch '{pokemon_name}'."
else:
fallback_reason = "LLM 'choose_switch' called without 'pokemon_name'."
else:
fallback_reason = f"LLM called unknown function '{function_name}'."
if not action_taken:
if not fallback_reason:
if error_message:
fallback_reason = f"API Error: {error_message}"
elif decision is None:
fallback_reason = "LLM did not provide a valid function call."
else:
fallback_reason = "Unknown error processing LLM decision."
print(f"Warning: {fallback_reason} Choosing random action.")
if battle.available_moves or battle.available_switches:
return self.choose_random_move(battle)
else:
print("AI Fallback: No moves or switches available. Using Struggle/Default.")
return self.choose_default_move(battle)
async def _get_llm_decision(self, battle_state: str) -> Dict[str, Any]:
raise NotImplementedError("Subclasses must implement _get_llm_decision")
# --- Google Gemini Agent ---
class GeminiAgent(LLMAgentBase):
"""Uses Google Gemini API for decisions."""
def __init__(self, api_key: str = None, model: str = "gemini-2.5-pro-preview-03-25", avatar: str = "steven", *args, **kwargs):
# Set avatar before calling parent constructor
kwargs['avatar'] = avatar
kwargs['start_timer_on_battle_start'] = True
super().__init__(*args, **kwargs)
self.model_name = model
used_api_key = api_key or os.environ.get("GOOGLE_API_KEY")
if not used_api_key:
raise ValueError("Google API key not provided or found in GOOGLE_API_KEY env var.")
# Initialize Gemini client using the correct API
self.genai_client = genai.Client(api_key=used_api_key)
# Configure the tools for function calling
self.function_declarations = list(self.standard_tools.values())
async def _get_llm_decision(self, battle_state: str) -> Dict[str, Any]:
"""Sends state to the Gemini API and gets back the function call decision."""
prompt = (
"Based on the current battle state, decide the best action: either use an available move or switch to an available Pokémon. "
"Consider type matchups, HP, status conditions, field effects, entry hazards, and potential opponent actions. "
"Only choose actions listed as available using their exact ID (for moves) or species name (for switches). "
"Use the provided functions to indicate your choice.\n\n"
f"Current Battle State:\n{battle_state}\n\n"
"Choose the best action by calling the appropriate function ('choose_move' or 'choose_switch')."
)
try:
# Configure tools using the Gemini API format
tools = genai.types.Tool(function_declarations=self.function_declarations)
config = genai.types.GenerateContentConfig(tools=[tools],automatic_function_calling=types.AutomaticFunctionCallingConfig(disable=True))
# Send request to the model
response = self.genai_client.models.generate_content(
model=self.model_name,
contents=prompt,
config=config
)
try:
function_calls = response.function_calls
if function_calls:
function_name = function_calls[0].name
arguments = function_calls[0].args
return {"decision": {"name": function_name, "arguments": arguments}}
else:
return {"error": "Gemini did not return a function call."}
except Exception as e:
return {"error": f"Model called unknown @function '{function_name}'."}
# No function call found
return {"error": "Gemini did not return a function call."}
except Exception as e:
print(f"Unexpected error during Gemini processing: {e}")
import traceback
traceback.print_exc()
return {"error": f"Unexpected error: {str(e)}"}
# --- OpenAI Agent ---
class OpenAIAgent(LLMAgentBase):
"""Uses OpenAI API for decisions."""
def __init__(self, api_key: str = None, model: str = "gpt-4.1", avatar: str = "giovanni", *args, **kwargs):
# Set avatar before calling parent constructor
kwargs['avatar'] = avatar
kwargs['start_timer_on_battle_start'] = True
super().__init__(*args, **kwargs)
self.model = model
used_api_key = api_key or os.environ.get("OPENAI_API_KEY")
if not used_api_key:
raise ValueError("OpenAI API key not provided or found in OPENAI_API_KEY env var.")
self.openai_client = AsyncOpenAI(api_key=used_api_key)
# Use the OpenAI-specific schema with type field
self.openai_tools = list(OPENAI_TOOL_SCHEMA.values())
async def _get_llm_decision(self, battle_state: str) -> Dict[str, Any]:
system_prompt = (
"You are a skilled Pokemon battle AI. Your goal is to win the battle. "
"Based on the current battle state, decide the best action: either use an available move or switch to an available Pokémon. "
"Consider type matchups, HP, status conditions, field effects, entry hazards, and potential opponent actions. "
"Only choose actions listed as available using their exact ID (for moves) or species name (for switches). "
"Use the provided functions to indicate your choice."
)
user_prompt = f"Current Battle State:\n{battle_state}\n\nChoose the best action by calling the appropriate function ('choose_move' or 'choose_switch')."
try:
response = await self.openai_client.chat.completions.create(
model=self.model,
messages=[
{"role": "system", "content": system_prompt},
{"role": "user", "content": user_prompt},
],
tools=self.openai_tools,
tool_choice="auto", # Let the model choose
temperature=0.5,
)
message = response.choices[0].message
print("OPENAI RESPONSE : ",response)
# Check for tool calls in the response
if message.tool_calls:
tool_call = message.tool_calls[0] # Get the first tool call
function_name = tool_call.function.name
try:
arguments = json.loads(tool_call.function.arguments or '{}')
if function_name in self.standard_tools:
return {"decision": {"name": function_name, "arguments": arguments}}
else:
return {"error": f"Model called unknown function '{function_name}'."}
except json.JSONDecodeError:
return {"error": f"Error decoding function arguments: {tool_call.function.arguments}"}
else:
# Model decided not to call a function
return {"error": f"OpenAI did not return a function call. Response: {message.content}"}
except APIError as e:
print(f"Error during OpenAI API call: {e}")
return {"error": f"OpenAI API Error: {e.status_code} - {e.message}"}
except Exception as e:
print(f"Unexpected error during OpenAI API call: {e}")
return {"error": f"Unexpected error: {e}"}
# --- Mistral Agent ---
class MistralAgent(LLMAgentBase):
"""Uses Mistral AI API for decisions."""
def __init__(self, api_key: str = None, model: str = "mistral-large-latest", avatar: str = "alder", *args, **kwargs):
# Set avatar before calling parent constructor
kwargs['avatar'] = avatar
kwargs['start_timer_on_battle_start'] = True
super().__init__(*args, **kwargs)
self.model = model
used_api_key = api_key or os.environ.get("MISTRAL_API_KEY")
if not used_api_key:
raise ValueError("Mistral API key not provided or found in MISTRAL_API_KEY env var.")
self.mistral_client = Mistral(api_key=used_api_key)
# Convert standard schema to Mistral's tool format with "function" wrapper
self.mistral_tools = []
for tool_name, tool_schema in self.standard_tools.items():
self.mistral_tools.append({
"type": "function",
"function": {
"name": tool_schema["name"],
"description": tool_schema["description"],
"parameters": tool_schema["parameters"]
}
})
async def _get_llm_decision(self, battle_state: str) -> Dict[str, Any]:
system_prompt = (
"You are a skilled Pokemon battle AI. Your goal is to win the battle. "
"Based on the current battle state, decide the best action: either use an available move or switch to an available Pokémon. "
"Consider type matchups, HP, status conditions, field effects, entry hazards, and potential opponent actions. "
"Only choose actions listed as available using their exact ID (for moves) or species name (for switches). "
"Use the provided tools to indicate your choice."
)
user_prompt = f"Current Battle State:\n{battle_state}\n\nChoose the best action by calling the appropriate function ('choose_move' or 'choose_switch')."
try:
# Create the messages array
messages = [
{"role": "system", "content": system_prompt},
{"role": "user", "content": user_prompt}
]
# Call the Mistral API with tool_choice set to "any" to force tool usage
response = self.mistral_client.chat.complete(
model=self.model,
messages=messages,
tools=self.mistral_tools,
tool_choice="any", # Force the model to use a tool
temperature=0.3,
)
print("Mistral RESPONSE : ", response)
# Check for tool calls in the response
message = response.choices[0].message
if hasattr(message, 'tool_calls') and message.tool_calls:
tool_call = message.tool_calls[0] # Get the first tool call
function_name = tool_call.function.name
try:
# Parse the function arguments from JSON string
arguments = json.loads(tool_call.function.arguments or '{}')
if function_name in self.standard_tools:
return {"decision": {"name": function_name, "arguments": arguments}}
else:
return {"error": f"Model called unknown function '{function_name}'."}
except json.JSONDecodeError:
return {"error": f"Error decoding function arguments: {tool_call.function.arguments}"}
else:
# Model did not return a tool call
return {"error": f"Mistral did not return a tool call. Response: {message.content}"}
except Exception as e:
print(f"Error during Mistral API call: {e}")
import traceback
traceback.print_exc()
return {"error": f"Unexpected error: {str(e)}"}