Enh/ref: Include notes on limit in config; simplify agent

Closes #27

Author: klieret

codeclash/agents/minisweagent.py

@@ -1,18 +1,13 @@
 import logging
-import os
-import platform
 import traceback
 from collections.abc import Callable
-from dataclasses import asdict
 
-from jinja2 import Template
 from minisweagent import Model
 from minisweagent.agents.default import AgentConfig, DefaultAgent
 from minisweagent.environments.docker import DockerEnvironment
 from minisweagent.models import get_model
 from minisweagent.models.test_models import DeterministicModel
 from minisweagent.run.utils.save import save_traj
-from rich.console import Console
 
 from codeclash.agents.abstract import Player
 from codeclash.agents.utils import GameContext
@@ -29,38 +24,17 @@ def __init__(
         self,
         model: Model,
         env: DockerEnvironment,
-        name: str,
-        game_context: GameContext,
         *,
         logger: logging.Logger,
         config_class: Callable = AgentConfig,
         **kwargs,
     ):
         super().__init__(model, env, config_class=config_class, **kwargs)
-        self.name = name
-        self.game_context = game_context
-        self.console = Console()
         self.logger = logger
 
     def add_message(self, role: str, content: str, **kwargs):
         super().add_message(role, content, **kwargs)
         self.logger.debug(f"[{role}] {content}", extra={"highlighter": None})
-        if role == "assistant":
-            self.logger.info(f"Step taken (step {self.model.n_calls}, cost {self.model.cost:.2f})")
-
-    def render_template(self, template: str, **kwargs) -> str:
-        cs = (
-            asdict(self.config)
-            | asdict(self.env.config)
-            | asdict(self.model.config)
-            | platform.uname()._asdict()
-            | self.game_context.to_template_vars()
-        )
-        return Template(template).render(**kwargs, **cs, **os.environ)
-
-    def run(self) -> tuple[str, str]:
-        """Run step() until agent is finished. Return exit status & message"""
-        return super().run(task="")
 
 
 class MiniSWEAgent(Player):
@@ -78,15 +52,13 @@ def run(self):
         self.agent = ClashAgent(
             model=model,
             env=self.environment,
-            name=self.name,
-            game_context=self.game_context,
             logger=self.logger,
             **self.config["config"]["agent"],
         )
         exit_status = None
         result = None
         try:
-            exit_status, result = self.agent.run()
+            exit_status, result = self.agent.run(task="", **self.game_context.to_template_vars())
         except Exception as e:
             exit_status = str(e)
             exc_message = traceback.format_exc()

configs/mini/default.yaml

@@ -32,7 +32,14 @@ instance_template: |
   - `trajs/`: History of your edits
   - and a lot more. It's up to you to explore and utilize these resources.
 
-  The game is played in rounds. You won't remember past rounds.
+  The game is played in rounds and you will be evaluated on the performance over all the rounds. You won't remember past rounds.
+
+  In every round, you have a limit of {{step_limit}} steps and a cost limit of {{cost_limit}} dollars.
+  We will show you the number of steps and cost used so far after every response in the `<limit_note>` tag.
+  After you've reached the step or cost limit, you cannot continue working on this task, and we will play the game with your codebase.
+  This means that it's fine to reach the step or cost limit while working on documentation or testing, but you shouldn't
+  reach the limit while working on the actual game logic to avoid submitting an invalid codebase.
+
   So if you want to carry knowledge forward — leave tools, notes, or strategies in the codebase.
   Good documentation means you (and others) can pick up right where you left off.
 
@@ -112,6 +119,7 @@ instance_template: |
   </example_response>
 
   If you need to run multiple commands, either:
+
   1. Combine them in one block using && or ||
   ```bash
   command1 && command2 || echo "Error occurred"
@@ -179,6 +187,7 @@ instance_template: |
   This command will submit your work.
   You cannot continue working (reading, editing, testing) in any way on this task after submitting.
 action_observation_template: |
+  <limit_note>This is the output of step {{n_model_calls}} ({{step_limit}} limit). You've used {{model_cost | round(2)}} USD ({{cost_limit}} USD limit).</limit_note>
   <returncode>{{output.returncode}}</returncode>
   {% if output.output | length < 10000 -%}
   <output>