tinymind-native-8b-remote-handoff/bundle/evaluation/ai_devtools.py

"""TinyMind AI DevTools Suite.

Project-native developer tools for evidence-driven model development.  The
suite scans source assets, hashes critical files, runs allowlisted checks, and
builds a training/readiness report for TinyMind without making unsupported
world-best claims.
"""

from __future__ import annotations

from datetime import datetime, timezone
import hashlib
import json
from pathlib import Path
import subprocess
from typing import Iterable


SOURCE_EXTENSIONS = {".py", ".rs", ".toml", ".md", ".json", ".jsonl", ".ps1", ".cu", ".h", ".cpp"}
EXCLUDED_DIRS = {".git", "__pycache__", ".pytest_cache", "target", "reports", "checkpoints", ".venv", "venv"}
DEFAULT_SMOKE_COMMANDS = [
    ["python", "-m", "pytest", "tests\\test_grounded_answer.py", "tests\\test_universal_context.py", "-q"],
]


def _sha256(path: Path) -> str:
    h = hashlib.sha256()
    with path.open("rb") as f:
        for chunk in iter(lambda: f.read(1024 * 1024), b""):
            h.update(chunk)
    return h.hexdigest()


def _is_excluded(path: Path) -> bool:
    return bool(set(path.parts) & EXCLUDED_DIRS)


def scan_project(root: str | Path) -> dict:
    root_path = Path(root).resolve()
    files = []
    counts: dict[str, int] = {}
    for path in sorted(root_path.rglob("*")):
        if not path.is_file() or _is_excluded(path):
            continue
        if path.suffix.lower() not in SOURCE_EXTENSIONS:
            continue
        rel = str(path.relative_to(root_path))
        ext = path.suffix.lower() or "<none>"
        counts[ext] = counts.get(ext, 0) + 1
        files.append(
            {
                "path": rel,
                "extension": ext,
                "bytes": path.stat().st_size,
                "sha256": _sha256(path),
            }
        )
    return {
        "root": str(root_path),
        "file_count": len(files),
        "extension_counts": counts,
        "files": files,
        "project_sha256": hashlib.sha256(
            "\n".join(f"{row['path']}:{row['sha256']}" for row in files).encode("utf-8")
        ).hexdigest(),
    }


def _allowed_command(argv: list[str]) -> bool:
    if len(argv) >= 3 and argv[0].lower() == "python" and argv[1:3] == ["-m", "pytest"]:
        return True
    if len(argv) >= 2 and argv[0].lower() == "cargo" and argv[1] in {"test", "build"}:
        return True
    return False


def run_allowlisted_command(argv: list[str], cwd: str | Path, timeout_s: int = 180) -> dict:
    if not _allowed_command(argv):
        return {"argv": argv, "ok": False, "skipped": True, "reason": "command_not_allowlisted"}
    try:
        proc = subprocess.run(argv, cwd=cwd, capture_output=True, text=True, shell=False, timeout=timeout_s)
        return {
            "argv": argv,
            "ok": proc.returncode == 0,
            "returncode": proc.returncode,
            "stdout_tail": proc.stdout[-4000:],
            "stderr_tail": proc.stderr[-4000:],
        }
    except subprocess.TimeoutExpired as exc:
        return {"argv": argv, "ok": False, "returncode": None, "stdout_tail": exc.stdout or "", "stderr_tail": "timeout"}


def training_readiness(root: str | Path) -> dict:
    root_path = Path(root)
    required = {
        "knowledge_full_cycle": root_path / "reports" / "knowledge_full_cycle" / "knowledge_full_cycle_report.json",
        "context_ledger": root_path / "reports" / "context_ledger_tinymind_clean" / "universal_context_manifest.json",
        "ai_tools_dll": root_path / "runtime" / "ai_tools_dll" / "target" / "release" / "tinymind_ai_tools.dll",
        "grounded_answer_guard": root_path / "evaluation" / "grounded_answer.py",
        "universal_context": root_path / "data" / "universal_context.py",
    }
    rows = []
    for name, path in required.items():
        rows.append({"name": name, "path": str(path), "exists": path.exists(), "bytes": path.stat().st_size if path.exists() else 0})
    passed = sum(1 for row in rows if row["exists"])
    return {
        "required_count": len(rows),
        "passed_count": passed,
        "coverage_percent": 100.0 * passed / max(len(rows), 1),
        "passed": passed == len(rows),
        "checks": rows,
        "definition": "100% means all local DevTools/training support artifacts required by this suite exist and are hashable.",
    }


def run_ai_devtools(
    root: str | Path,
    out_dir: str | Path,
    run_smoke: bool = False,
    extra_commands: Iterable[list[str]] = (),
) -> dict:
    root_path = Path(root).resolve()
    out = Path(out_dir)
    out.mkdir(parents=True, exist_ok=True)
    scan = scan_project(root_path)
    readiness = training_readiness(root_path)
    commands = []
    if run_smoke:
        for argv in list(DEFAULT_SMOKE_COMMANDS) + list(extra_commands):
            commands.append(run_allowlisted_command(argv, root_path))
    command_gate = {"passed": all(row.get("ok", False) for row in commands) if commands else True, "commands": commands}
    report = {
        "schema_version": "tinymind-ai-devtools-suite-v1",
        "created_at": datetime.now(timezone.utc).isoformat(),
        "claim": "AI-native DevTools for TinyMind evidence, training readiness, and source-grounded tool use.",
        "world_best_claim_allowed": False,
        "scan": scan,
        "training_readiness": readiness,
        "command_gate": command_gate,
        "devtools_gate": {
            "passed": readiness["passed"] and command_gate["passed"] and scan["file_count"] > 0,
            "notes": "This is a local completeness gate, not an official global ranking.",
        },
    }
    json_path = out / "ai_devtools_report.json"
    md_path = out / "ai_devtools_report.md"
    report["json_path"] = str(json_path)
    report["markdown_path"] = str(md_path)
    json_path.write_text(json.dumps(report, ensure_ascii=False, indent=2, sort_keys=True), encoding="utf-8")
    md_path.write_text(_markdown(report), encoding="utf-8")
    return report


def _markdown(report: dict) -> str:
    lines = [
        "# TinyMind AI DevTools Suite",
        "",
        f"- Files scanned: {report['scan']['file_count']}",
        f"- Project hash: `{report['scan']['project_sha256']}`",
        f"- Training readiness: {report['training_readiness']['coverage_percent']:.1f}%",
        f"- Command gate: {report['command_gate']['passed']}",
        f"- DevTools gate: {report['devtools_gate']['passed']}",
        "- World-best claim allowed: false",
        "",
        "## Readiness Checks",
        "",
    ]
    for row in report["training_readiness"]["checks"]:
        lines.append(f"- {row['name']}: {row['exists']} ({row['path']})")
    return "\n".join(lines) + "\n"