Datasets:

lyffseba
/

xai-studies

#!/usr/bin/env python3
"""Enforce monorepo package contract for every models/*/ directory with a Makefile."""
from __future__ import annotations

import sys
from pathlib import Path

ROOT = Path(__file__).resolve().parents[2]
MODELS = ROOT / "models"

REQUIRED_PATHS = [
    "README.md",
    "Makefile",
    "book",
    "code-study",
    "reference",
    "derived/summary.json",
    "scripts/generate_derived.py",
    "scripts/validate.py",
    # refresh is online-only; not required for offline contract
]

REQUIRED_SUMMARY_KEYS = [
    "model_id",
    "display_name",
    "status",
    "context_tokens",
    "generated_at_utc",
]


def main() -> int:
    import json

    errors: list[str] = []
    packages = sorted(p for p in MODELS.iterdir() if p.is_dir() and (p / "Makefile").exists())
    if not packages:
        print("ERROR: no model packages found")
        return 1

    for pkg in packages:
        mid = pkg.name
        for rel in REQUIRED_PATHS:
            if not (pkg / rel).exists():
                errors.append(f"{mid}: missing {rel}")
        summary_path = pkg / "derived" / "summary.json"
        if summary_path.exists():
            try:
                s = json.loads(summary_path.read_text())
            except json.JSONDecodeError as e:
                errors.append(f"{mid}: invalid summary.json ({e})")
                continue
            for k in REQUIRED_SUMMARY_KEYS:
                if k not in s:
                    errors.append(f"{mid}: summary.json missing key '{k}'")
            if s.get("model_id") not in (None, mid) and s.get("model_id") != mid:
                errors.append(f"{mid}: summary.model_id={s.get('model_id')!r} != dir name")

    # monorepo dashboard present after check
    if not (ROOT / "derived" / "DASHBOARD.md").exists():
        errors.append("monorepo: missing derived/DASHBOARD.md (run make dashboard)")

    # comparison notes for registered packages
    if (MODELS / "inkling").exists() and (MODELS / "kimi-k3").exists():
        cmp_path = ROOT / "shared" / "comparisons" / "inkling-vs-kimi-k3.md"
        if not cmp_path.exists():
            errors.append("missing shared/comparisons/inkling-vs-kimi-k3.md")
    if (
        (MODELS / "laguna-s-2.1").exists()
        and (MODELS / "inkling").exists()
        and (MODELS / "kimi-k3").exists()
    ):
        tri = ROOT / "shared" / "comparisons" / "laguna-s-vs-inkling-vs-kimi.md"
        if not tri.exists():
            errors.append("missing shared/comparisons/laguna-s-vs-inkling-vs-kimi.md")

    # Laguna-specific contract (weights + chat stack)
    lag = MODELS / "laguna-s-2.1"
    if lag.exists() and (lag / "Makefile").exists():
        for rel in (
            "reference/hub/config.json",
            "reference/hub/chat_template.jinja",
            "reference/hub/tokenizer_config.json",
            "reference/hub/generation_config.json",
            "reference/laguna-code/modeling_laguna.py",
            "derived/special_tokens.md",
            "derived/param_accounting.md",
            "code-study/04-chat-and-tools.md",
        ):
            if not (lag / rel).exists():
                errors.append(f"laguna-s-2.1: missing {rel}")

    # Monorepo root hygiene & cross-doc consistency
    monorepo = ROOT.parent
    forbidden_files = (
        ".cleanup-github-only.sh",
        ".finalize-remaining.sh",
        ".finalize-report.txt",
        ".finalize-workflow.sh",
        ".finish-cleanup.sh",
        ".gh-push-switch-request",
        "hf/.publish-request",
        "hf/_do_publish.py",
        "hf/_publish_inline.py",
        "hf/_run_publish_and_push.sh",
        "studies/models/registry.yaml",
        "studies/models/kimi-k3/reference/facts.yaml",
        "studies/.github/workflows/check.yml",
    )
    for rel in forbidden_files:
        if (monorepo / rel).exists():
            errors.append(f"hygiene: obsolete/scratch file present: {rel}")

    ci_wf = monorepo / ".github" / "workflows" / "check.yml"
    if not ci_wf.exists():
        errors.append("hygiene: missing root .github/workflows/check.yml")
    elif "pip install" in ci_wf.read_text(encoding="utf-8"):
        errors.append("hygiene: .github/workflows/check.yml must not use pip install")

    # Every registered model must be referenced in top-level docs
    model_ids = [p.name for p in packages]
    for doc_rel in (
        "README.md",
        "START_HERE.md",
        "OFFLINE.md",
        "CITATION.cff",
        "CHECKLIST.md",
        "index.html",
        "hf/dataset/README.md",
        "hf/space/index.html",
    ):
        doc_path = monorepo / doc_rel
        if not doc_path.exists():
            errors.append(f"hygiene: missing {doc_rel}")
            continue
        content = doc_path.read_text(encoding="utf-8").lower()
        for mid in model_ids:
            short = mid.split("-")[0]
            if mid not in content and short not in content:
                errors.append(f"{doc_rel}: missing model reference for '{mid}'")

    ds_readme = (monorepo / "hf" / "dataset" / "README.md").read_text(encoding="utf-8")
    if "studies/models/inkling|kimi-k3/" in ds_readme:
        errors.append("hf/dataset/README.md: contents tree omits laguna-s-2.1")
    if "Poolside" not in ds_readme.split("## limitations")[-1]:
        errors.append("hf/dataset/README.md: limitations section omits Poolside")

    # Check relative links in authored markdown and HTML files
    import re

    for md_path in sorted(monorepo.rglob("*.md")):
        if any(part in md_path.parts for part in (".git", ".hf-stage", "reference")):
            continue
        text = md_path.read_text(encoding="utf-8", errors="ignore")
        for m in re.finditer(r"\[([^\]]*)\]\(([^)]+)\)", text):
            target = m.group(2).split("#")[0].strip()
            if not target or target.startswith(("http://", "https://", "mailto:")):
                continue
            if not (md_path.parent / target).resolve().exists():
                rel_src = md_path.relative_to(monorepo)
                errors.append(f"broken link in {rel_src}: {target}")

    for html_path in sorted(monorepo.rglob("*.html")):
        if any(part in html_path.parts for part in (".git", ".hf-stage")):
            continue
        text = html_path.read_text(encoding="utf-8", errors="ignore")
        for m in re.finditer(r'href="([^"]+)"', text):
            target = m.group(1).split("#")[0].strip()
            if not target or target.startswith(("http://", "https://", "mailto:")):
                continue
            if not (html_path.parent / target).resolve().exists():
                rel_src = html_path.relative_to(monorepo)
                errors.append(f"broken link in {rel_src}: {target}")

    # Check STUDY.md ordered list numbering in Laguna section
    study_md = ROOT / "STUDY.md"
    if study_md.exists():
        s_txt = study_md.read_text(encoding="utf-8")
        if "8. [code-study/04-chat-and-tools.md]" in s_txt and "\n7. benches + limits:" in s_txt:
            errors.append("STUDY.md: out-of-order list item '7. benches + limits' after item 8")

    if errors:
        for e in errors:
            print(f"ERROR: {e}")
        print(f"\npackage check FAILED ({len(errors)} errors)")
        return 1

    print(f"package check OK ({len(packages)} models)")
    return 0


if __name__ == "__main__":
    raise SystemExit(main())

	#!/usr/bin/env python3
	"""Enforce monorepo package contract for every models/*/ directory with a Makefile."""
	from __future__ import annotations

	import sys
	from pathlib import Path

	ROOT = Path(__file__).resolve().parents[2]
	MODELS = ROOT / "models"

	REQUIRED_PATHS = [
	"README.md",
	"Makefile",
	"book",
	"code-study",
	"reference",
	"derived/summary.json",
	"scripts/generate_derived.py",
	"scripts/validate.py",
	# refresh is online-only; not required for offline contract
	]

	REQUIRED_SUMMARY_KEYS = [
	"model_id",
	"display_name",
	"status",
	"context_tokens",
	"generated_at_utc",
	]


	def main() -> int:
	import json

	errors: list[str] = []
	packages = sorted(p for p in MODELS.iterdir() if p.is_dir() and (p / "Makefile").exists())
	if not packages:
	print("ERROR: no model packages found")
	return 1

	for pkg in packages:
	mid = pkg.name
	for rel in REQUIRED_PATHS:
	if not (pkg / rel).exists():
	errors.append(f"{mid}: missing {rel}")
	summary_path = pkg / "derived" / "summary.json"
	if summary_path.exists():
	try:
	s = json.loads(summary_path.read_text())
	except json.JSONDecodeError as e:
	errors.append(f"{mid}: invalid summary.json ({e})")
	continue
	for k in REQUIRED_SUMMARY_KEYS:
	if k not in s:
	errors.append(f"{mid}: summary.json missing key '{k}'")
	if s.get("model_id") not in (None, mid) and s.get("model_id") != mid:
	errors.append(f"{mid}: summary.model_id={s.get('model_id')!r} != dir name")

	# monorepo dashboard present after check
	if not (ROOT / "derived" / "DASHBOARD.md").exists():
	errors.append("monorepo: missing derived/DASHBOARD.md (run make dashboard)")

	# comparison notes for registered packages
	if (MODELS / "inkling").exists() and (MODELS / "kimi-k3").exists():
	cmp_path = ROOT / "shared" / "comparisons" / "inkling-vs-kimi-k3.md"
	if not cmp_path.exists():
	errors.append("missing shared/comparisons/inkling-vs-kimi-k3.md")
	if (
	(MODELS / "laguna-s-2.1").exists()
	and (MODELS / "inkling").exists()
	and (MODELS / "kimi-k3").exists()
	):
	tri = ROOT / "shared" / "comparisons" / "laguna-s-vs-inkling-vs-kimi.md"
	if not tri.exists():
	errors.append("missing shared/comparisons/laguna-s-vs-inkling-vs-kimi.md")

	# Laguna-specific contract (weights + chat stack)
	lag = MODELS / "laguna-s-2.1"
	if lag.exists() and (lag / "Makefile").exists():
	for rel in (
	"reference/hub/config.json",
	"reference/hub/chat_template.jinja",
	"reference/hub/tokenizer_config.json",
	"reference/hub/generation_config.json",
	"reference/laguna-code/modeling_laguna.py",
	"derived/special_tokens.md",
	"derived/param_accounting.md",
	"code-study/04-chat-and-tools.md",
	):
	if not (lag / rel).exists():
	errors.append(f"laguna-s-2.1: missing {rel}")

	# Monorepo root hygiene & cross-doc consistency
	monorepo = ROOT.parent
	forbidden_files = (
	".cleanup-github-only.sh",
	".finalize-remaining.sh",
	".finalize-report.txt",
	".finalize-workflow.sh",
	".finish-cleanup.sh",
	".gh-push-switch-request",
	"hf/.publish-request",
	"hf/_do_publish.py",
	"hf/_publish_inline.py",
	"hf/_run_publish_and_push.sh",
	"studies/models/registry.yaml",
	"studies/models/kimi-k3/reference/facts.yaml",
	"studies/.github/workflows/check.yml",
	)
	for rel in forbidden_files:
	if (monorepo / rel).exists():
	errors.append(f"hygiene: obsolete/scratch file present: {rel}")

	ci_wf = monorepo / ".github" / "workflows" / "check.yml"
	if not ci_wf.exists():
	errors.append("hygiene: missing root .github/workflows/check.yml")
	elif "pip install" in ci_wf.read_text(encoding="utf-8"):
	errors.append("hygiene: .github/workflows/check.yml must not use pip install")

	# Every registered model must be referenced in top-level docs
	model_ids = [p.name for p in packages]
	for doc_rel in (
	"README.md",
	"START_HERE.md",
	"OFFLINE.md",
	"CITATION.cff",
	"CHECKLIST.md",
	"index.html",
	"hf/dataset/README.md",
	"hf/space/index.html",
	):
	doc_path = monorepo / doc_rel
	if not doc_path.exists():
	errors.append(f"hygiene: missing {doc_rel}")
	continue
	content = doc_path.read_text(encoding="utf-8").lower()
	for mid in model_ids:
	short = mid.split("-")[0]
	if mid not in content and short not in content:
	errors.append(f"{doc_rel}: missing model reference for '{mid}'")

	ds_readme = (monorepo / "hf" / "dataset" / "README.md").read_text(encoding="utf-8")
	if "studies/models/inkling\|kimi-k3/" in ds_readme:
	errors.append("hf/dataset/README.md: contents tree omits laguna-s-2.1")
	if "Poolside" not in ds_readme.split("## limitations")[-1]:
	errors.append("hf/dataset/README.md: limitations section omits Poolside")

	# Check relative links in authored markdown and HTML files
	import re

	for md_path in sorted(monorepo.rglob("*.md")):
	if any(part in md_path.parts for part in (".git", ".hf-stage", "reference")):
	continue
	text = md_path.read_text(encoding="utf-8", errors="ignore")
	for m in re.finditer(r"\[([^\]]*)\]\(([^)]+)\)", text):
	target = m.group(2).split("#")[0].strip()
	if not target or target.startswith(("http://", "https://", "mailto:")):
	continue
	if not (md_path.parent / target).resolve().exists():
	rel_src = md_path.relative_to(monorepo)
	errors.append(f"broken link in {rel_src}: {target}")

	for html_path in sorted(monorepo.rglob("*.html")):
	if any(part in html_path.parts for part in (".git", ".hf-stage")):
	continue
	text = html_path.read_text(encoding="utf-8", errors="ignore")
	for m in re.finditer(r'href="([^"]+)"', text):
	target = m.group(1).split("#")[0].strip()
	if not target or target.startswith(("http://", "https://", "mailto:")):
	continue
	if not (html_path.parent / target).resolve().exists():
	rel_src = html_path.relative_to(monorepo)
	errors.append(f"broken link in {rel_src}: {target}")

	# Check STUDY.md ordered list numbering in Laguna section
	study_md = ROOT / "STUDY.md"
	if study_md.exists():
	s_txt = study_md.read_text(encoding="utf-8")
	if "8. [code-study/04-chat-and-tools.md]" in s_txt and "\n7. benches + limits:" in s_txt:
	errors.append("STUDY.md: out-of-order list item '7. benches + limits' after item 8")

	if errors:
	for e in errors:
	print(f"ERROR: {e}")
	print(f"\npackage check FAILED ({len(errors)} errors)")
	return 1

	print(f"package check OK ({len(packages)} models)")
	return 0


	if __name__ == "__main__":
	raise SystemExit(main())