Add scripts for environment setup and Node.js installation; include AGENTS.md for Windows shell configuration

AGENTS.md

@@ -0,0 +1,2 @@
+[windows]
+shell = "cmd"

check_env.ps1

@@ -0,0 +1,15 @@
+$badVars = @()
+Get-ItemProperty -Path 'HKLM:\SYSTEM\CurrentControlSet\Control\Session Manager\Environment' | 
+    Get-Member -MemberType NoteProperty | 
+    ForEach-Object { 
+        $name = $_.Name
+        $val = (Get-ItemProperty -Path 'HKLM:\SYSTEM\CurrentControlSet\Control\Session Manager\Environment').$name
+        if ($val -match '[^\x09\x20-\x7E]') {
+            $badVars += "$name`: $val"
+        }
+    }
+if ($badVars.Count -eq 0) {
+    Write-Output "No corrupted env vars found in HKLM"
+} else {
+    $badVars | ForEach-Object { Write-Output $_ }
+}

check_env2.ps1

@@ -0,0 +1,15 @@
+$badVars = @()
+Get-ItemProperty -Path 'HKCU:\Environment' | 
+    Get-Member -MemberType NoteProperty | 
+    ForEach-Object { 
+        $name = $_.Name
+        $val = (Get-ItemProperty -Path 'HKCU:\Environment').$name
+        if ($val -match '[^\x09\x20-\x7E]') {
+            $badVars += "$name`: $val"
+        }
+    }
+if ($badVars.Count -eq 0) {
+    Write-Output "No corrupted env vars found in HKCU"
+} else {
+    $badVars | ForEach-Object { Write-Output $_ }
+}

gaia_matcher.py

@@ -0,0 +1,160 @@
+import os
+import json
+import shutil
+import requests
+import pandas as pd
+from huggingface_hub import hf_hub_download
+
+QUESTIONS_URL = "https://agents-course-unit4-scoring.hf.space/questions"
+GAIA_REPO_ID = "gaia-benchmark/GAIA"
+GAIA_VAL_FILENAME = "2023/validation/metadata.parquet"
+CACHE_DIR = os.path.join(os.path.dirname(__file__), "data")
+CACHE_PATH = os.path.join(CACHE_DIR, "gaia_metadata.parquet")
+OUT_CSV = os.path.join(os.path.dirname(__file__), "gaia_ground_truth.csv")
+OUT_JSON = os.path.join(os.path.dirname(__file__), "gaia_ground_truth.json")
+
+
+def get_hf_token():
+    # Precedence of env vars and helpful message
+    for k in ("HUGGINGFACEHUB_API_TOKEN", "HF_TOKEN", "HUGGINGFACE_TOKEN"):
+        v = os.getenv(k)
+        if v:
+            return v
+    return None
+
+
+def download_parquet(token: str, dest_path: str) -> str:
+    # Download via hf_hub_download to a temporary location then move to dest
+    tmp = hf_hub_download(
+        repo_id=GAIA_REPO_ID,
+        filename=GAIA_VAL_FILENAME,
+        repo_type="dataset",
+        token=token,
+    )
+    os.makedirs(os.path.dirname(dest_path), exist_ok=True)
+    shutil.copy(tmp, dest_path)
+    return dest_path
+
+
+def detect_columns(df: pd.DataFrame):
+    cols = [c for c in df.columns]
+    # task id candidates
+    task_candidates = ["task_id", "id", "Task ID", "TaskID", "taskid"]
+    answer_candidates = ["Final answer", "final_answer", "answer", "Final Answer", "final answer", "Final_answer"]
+
+    task_col = next((c for c in cols if c in task_candidates), None)
+    answer_col = next((c for c in cols if c in answer_candidates), None)
+
+    # Fallback: case-insensitive match
+    if not task_col:
+        lowered = {c.lower(): c for c in cols}
+        for cand in task_candidates:
+            if cand.lower() in lowered:
+                task_col = lowered[cand.lower()]
+                break
+    if not answer_col:
+        lowered = {c.lower(): c for c in cols}
+        for cand in answer_candidates:
+            if cand.lower() in lowered:
+                answer_col = lowered[cand.lower()]
+                break
+
+    return task_col, answer_col
+
+
+def safe_read_parquet(path: str) -> pd.DataFrame:
+    # Try pandas default, then pyarrow engine if needed
+    try:
+        return pd.read_parquet(path)
+    except Exception:
+        try:
+            return pd.read_parquet(path, engine="pyarrow")
+        except Exception as e:
+            raise
+
+
+def main():
+    print(f"Fetching questions from {QUESTIONS_URL}...")
+    try:
+        resp = requests.get(QUESTIONS_URL, timeout=10)
+        resp.raise_for_status()
+        current_questions = resp.json()
+    except Exception as e:
+        print(f"Error fetching questions: {e}")
+        current_questions = []
+
+    token = get_hf_token()
+    if not token:
+        print("Warning: No HF token found in env (HUGGINGFACEHUB_API_TOKEN/HF_TOKEN). Trying public access; gated datasets may fail.")
+
+    # Ensure parquet is cached locally
+    if os.path.exists(CACHE_PATH):
+        print(f"Using cached GAIA parquet at {CACHE_PATH}")
+        parquet_path = CACHE_PATH
+    else:
+        try:
+            print("Downloading GAIA validation metadata (this may require a HF token)...")
+            parquet_path = download_parquet(token, CACHE_PATH)
+            print(f"Downloaded and cached to {parquet_path}")
+        except Exception as e:
+            print(f"Error downloading parquet: {e}")
+            print("Abort: could not obtain GAIA metadata. Consider setting HF_TOKEN or using an offline parquet in data/")
+            return
+
+    try:
+        df = safe_read_parquet(parquet_path)
+    except Exception as e:
+        print(f"Error reading parquet: {e}")
+        return
+
+    task_col, answer_col = detect_columns(df)
+    if not task_col or not answer_col:
+        print("Could not detect task_id or answer column. Available columns:\n", df.columns.tolist())
+        return
+
+    # Normalize to strings
+    df[task_col] = df[task_col].astype(str).str.strip()
+    df[answer_col] = df[answer_col].astype(str).str.strip()
+    answer_map = dict(zip(df[task_col], df[answer_col]))
+
+    results = []
+    found = 0
+    total = len(current_questions)
+    for i, q in enumerate(current_questions):
+        task_id = q.get("task_id")
+        task_id_str = str(task_id).strip() if task_id is not None else ""
+        answer = None
+        if task_id_str and task_id_str in answer_map:
+            answer = answer_map[task_id_str]
+        else:
+            # Try relaxed matching: maybe numeric vs string formatting
+            # Check without leading zeros and as int
+            try:
+                tid_int = str(int(task_id_str))
+                answer = answer_map.get(tid_int)
+            except Exception:
+                answer = None
+
+        ok = answer is not None and answer.lower() != "nan"
+        if ok:
+            found += 1
+        results.append({
+            "index": i + 1,
+            "task_id": task_id_str,
+            "question": (q.get("question") or "")[:1000],
+            "answer": answer if ok else None,
+            "found": bool(ok),
+        })
+
+    # Save outputs
+    out_df = pd.DataFrame(results)
+    out_df.to_csv(OUT_CSV, index=False)
+    with open(OUT_JSON, "w", encoding="utf-8") as f:
+        json.dump(results, f, ensure_ascii=False, indent=2)
+
+    print(f"Matched answers: {found}/{total}")
+    print(f"Results saved to: {OUT_CSV} and {OUT_JSON}")
+
+
+if __name__ == "__main__":
+    main()

install_node.sh

@@ -0,0 +1,8 @@
+#!/bin/bash
+cd /tmp
+tar -xf node-v20.18.0-linux-x64.tar.xz
+sudo cp -r node-v20.18.0-linux-x64/* /usr/local/
+/usr/local/bin/node --version
+/usr/local/bin/npm --version
+sudo /usr/local/bin/npm install -g @openai/codex
+codex --version

setupext.sh

@@ -0,0 +1,7 @@
+#!/bin/bash
+curl -fsSL https://deb.nodesource.com/setup_20.x | sudo -E bash -
+sudo apt-get install -y nodejs
+npm -v
+node -v
+npm install -g @openai/codex
+codex --version