agents/task2.py

"""Agents for Task2: Property Discovery"""

import random as _random
from typing import Any, Dict, List

from server import Task2Environment
from env.schemas import Action, ActionType
from data.data_loader import load_contracts, get_function_by_name

# ─────────────────────────────────────────────────────────────────────────────
# Task 2 agents
# ─────────────────────────────────────────────────────────────────────────────

def oracle_t2(env: Task2Environment, seed: int, verbose: bool = False) -> Dict[str, Any]:
    """Submits ground-truth in natural langugage (English) → score ≥ 0.70."""
    r        = env.reset(seed=seed)
    obs      = r.observation
    fn_name  = obs.extra["target_function"]
    contract = obs.contract_name
    contracts = load_contracts()
    gt_text = ""
    for c in contracts:
        if c["contract_name"] == contract:
            fn = get_function_by_name(c, fn_name)
            if fn and fn.get("property"):
                gt_text = fn["property"]
            break
    if verbose:
        print(f"    {contract}.{fn_name}()")
    env.step(Action(action_type=ActionType.GET_FUNCTION_CODE))
    result = env.step(Action(action_type=ActionType.SUBMIT_PROPERTY,
                              params={"property": gt_text}))
    return {"seed": seed, "contract": contract, "function": fn_name,
            "grader_score": result.reward.value }


def partial_t2(env: Task2Environment, seed: int) -> Dict[str, Any]:
    """Submits only the function's short NatSpec comment — partial credit."""
    r   = env.reset(seed=seed)
    obs = r.observation
    contracts = load_contracts()
    comment = ""
    for c in contracts:
        if c["contract_name"] == obs.contract_name:
            fn = get_function_by_name(c, obs.extra["target_function"])
            if fn:
                comment = fn.get("comment", "")
            break
    result = env.step(Action(action_type=ActionType.SUBMIT_PROPERTY,
                              params={"property": comment}))
    return {"seed": seed, "grader_score": result.reward.value}


def random_t2(env: Task2Environment, seed: int) -> Dict[str, Any]:
    """Genuine random agent: random browse then submits a generic property template.

    The submitted text contains high-frequency words that are unlikely to match
    task-specific key phrases. Expected score: near 0 (coincidental matches only).
    Uses a seeded RNG for reproducibility.
    """
    rng = _random.Random(seed ^ 0xBEEF1)

    r   = env.reset(seed=seed)
    obs = r.observation
    fn_name = obs.extra.get("target_function", "this function")

    # Random browse: pick 1–2 actions at random
    browse_pool = [
        ActionType.GET_FILE_NATSPEC,
        ActionType.GET_RELATED_FUNCTIONS,
        ActionType.GET_SIGNATURE,
    ]
    rng.shuffle(browse_pool)
    for at in browse_pool[:rng.randint(1, 2)]:
        env.step(Action(action_type=at))

    # Submit a randomly assembled generic property (won't match specific key phrases)
    templates = [
        f"The {fn_name} operation completes the intended computation on the input data.",
        f"When {fn_name} executes, it processes the provided arguments and updates the contract.",
        f"The {fn_name} function validates inputs and performs the expected operation.",
        f"Calling {fn_name} causes the contract to execute its designated logic.",
        f"{fn_name} runs when invoked and modifies internal state as designed.",
    ]
    prop = rng.choice(templates)
    result = env.step(Action(action_type=ActionType.SUBMIT_PROPERTY,
                              params={"property": prop}))
    return {"seed": seed, "grader_score": result.reward.value,
            "submitted": prop[:60]}


def floor_t2(env: Task2Environment, seed: int) -> Dict[str, Any]:
    """Submits empty string → score = 0.0 guaranteed."""
    env.reset(seed=seed)
    result = env.step(Action(action_type=ActionType.SUBMIT_PROPERTY,
                              params={"property": ""}))
    return {"seed": seed, "grader_score": 0.001}