File size: 7,357 Bytes
3af908f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
#!/usr/bin/env python3
"""Integration eval: fixtures, course API, or single task."""

from __future__ import annotations

import argparse
import json
import os
import sys
import tempfile
from pathlib import Path

from dotenv import load_dotenv

load_dotenv()

from agent import GaiaAgent
from agent.evolve import record_run, update_strategy_deltas
from eval.course_client import fetch_course_questions, submit_answers
from eval.report import build_report, print_report_summary
from file_resolver import resolve_task_attachment
from scoring import course_answer_matches, passes_course_threshold

FIXTURES_DIR = Path(__file__).resolve().parent / "fixtures"
RESULTS_DIR = Path(__file__).resolve().parent / "results"
DEFAULT_API_URL = os.getenv(
    "COURSE_API_URL", "https://agents-course-unit4-scoring.hf.space"
)


def load_fixtures() -> list[dict]:
    items: list[dict] = []
    for path in sorted(FIXTURES_DIR.glob("*.json")):
        data = json.loads(path.read_text(encoding="utf-8"))
        if isinstance(data, list):
            items.extend(data)
    return items


def run_fixture(agent: GaiaAgent, fixture: dict) -> dict:
    answer = agent(
        fixture["question"],
        file_path=fixture.get("file_path"),
        task_id=fixture.get("task_id"),
    )
    expected = fixture["expected"]
    correct = course_answer_matches(answer, expected)
    return {
        "task_id": fixture.get("task_id", "fixture"),
        "question": fixture["question"],
        "submitted": answer,
        "expected": expected,
        "correct": correct,
        "question_tags": fixture.get("question_tags", []),
        "strategy": fixture.get("strategy", "direct"),
    }


def run_course(agent: GaiaAgent, api_url: str) -> list[dict]:
    questions = fetch_course_questions(api_url)
    results: list[dict] = []
    answers_payload: list[dict] = []

    with tempfile.TemporaryDirectory(prefix="gaia_eval_") as temp_dir:
        download_dir = Path(temp_dir)
        for item in questions:
            task_id = item["task_id"]
            file_path = None
            file_error = None
            file_name = item.get("file_name") or ""
            if file_name:
                file_path, file_error = resolve_task_attachment(
                    api_url, task_id, file_name, download_dir
                )
            try:
                submitted = agent(
                    item["question"],
                    file_path=file_path,
                    file_error=file_error,
                    task_id=task_id,
                )
            except Exception as error:
                submitted = f"ERROR: {error}"
            answers_payload.append(
                {"task_id": task_id, "submitted_answer": submitted}
            )
            results.append(
                {
                    "task_id": task_id,
                    "question": item["question"],
                    "submitted": submitted,
                    "correct": False,
                }
            )

    username = os.getenv("HF_USERNAME", "").strip()
    if username:
        submit_result = submit_answers(
            api_url,
            username,
            os.getenv("AGENT_CODE", "local-eval"),
            answers_payload,
        )
        detail_map = {
            row.get("task_id"): row
            for row in submit_result.get("details") or submit_result.get("results") or []
        }
        correct_count = submit_result.get("correct_count", 0)
        for row in results:
            detail = detail_map.get(row["task_id"], {})
            row["correct"] = detail.get("correct", False)
            row["expected"] = detail.get("expected_answer", "?")
        print(f"Course submit score: {submit_result.get('score', '?')}% ({correct_count} correct)")
    else:
        print("HF_USERNAME not set — answers collected but not graded via course API.")

    return results


def main() -> None:
    parser = argparse.ArgumentParser(description="Run GAIA eval harness.")
    parser.add_argument(
        "--mode",
        choices=("fixtures", "course", "single"),
        default="fixtures",
    )
    parser.add_argument("--task-id", help="Fixture or course task id for single mode.")
    parser.add_argument(
        "--min-score",
        type=float,
        default=float(os.getenv("EVAL_FIXTURE_MIN_SCORE", "80")),
        help="Minimum percent for fixtures mode pass gate.",
    )
    args = parser.parse_args()

    agent = GaiaAgent()
    results: list[dict] = []

    if args.mode == "fixtures":
        fixtures = load_fixtures()
        if args.task_id:
            fixtures = [f for f in fixtures if f.get("task_id") == args.task_id]
        for fixture in fixtures:
            print(f"Running fixture: {fixture.get('task_id')}")
            row = run_fixture(agent, fixture)
            results.append(row)
            record_run(
                row["task_id"],
                row["question"],
                row.get("strategy", "direct"),
                [],
                row["correct"],
            )
    elif args.mode == "course":
        results = run_course(agent, DEFAULT_API_URL)
    elif args.mode == "single":
        if not args.task_id:
            print("--task-id required for single mode", file=sys.stderr)
            sys.exit(1)
        fixtures = [f for f in load_fixtures() if f.get("task_id") == args.task_id]
        if fixtures:
            results = [run_fixture(agent, fixtures[0])]
        else:
            questions = fetch_course_questions(DEFAULT_API_URL)
            match = next((q for q in questions if q["task_id"] == args.task_id), None)
            if not match:
                print(f"Task {args.task_id} not found", file=sys.stderr)
                sys.exit(1)
            with tempfile.TemporaryDirectory(prefix="gaia_single_") as temp_dir:
                file_path = None
                file_error = None
                if match.get("file_name"):
                    file_path, file_error = resolve_task_attachment(
                        DEFAULT_API_URL,
                        match["task_id"],
                        match["file_name"],
                        Path(temp_dir),
                    )
                answer = agent(
                    match["question"],
                    file_path=file_path,
                    file_error=file_error,
                    task_id=match["task_id"],
                )
            results = [
                {
                    "task_id": match["task_id"],
                    "question": match["question"],
                    "submitted": answer,
                    "correct": False,
                }
            ]

    report_path = build_report(args.mode, results, RESULTS_DIR)
    update_strategy_deltas(report_path)
    correct = sum(1 for row in results if row.get("correct"))
    total = len(results)
    report = {
        "correct": correct,
        "total": total,
        "score_percent": round((correct / total) * 100, 2) if total else 0.0,
        "passed_threshold": passes_course_threshold(correct, total),
    }
    print_report_summary(report)

    if args.mode == "fixtures" and total:
        score = (correct / total) * 100
        if score < args.min_score:
            print(f"Fixture gate failed: {score:.1f}% < {args.min_score}%")
            sys.exit(1)


if __name__ == "__main__":
    main()