tools/web_tools.py

"""Internet access for the local model (stdlib only).

CLI:
  python web_tools.py search "query" [-n 8]
  python web_tools.py fetch URL [--max 20000]
  python web_tools.py hf-note REPO PATH_IN_REPO "text" [--type model|dataset|space] [--append]

Import:  from web_tools import search, fetch, hf_note

search  -> needs SearXNG running (searxng\\start_searxng.bat) on 127.0.0.1:8888
fetch   -> direct HTTP GET, returns readable text
hf_note -> writes/appends a note file in a Hugging Face repo; token from env
           HF_TOKEN (or HUGGINGFACE_HUB_TOKEN, or file tools\\hf_token.txt)
"""
import argparse, base64, gzip, json, os, re, sys, urllib.error, urllib.parse, urllib.request
from html.parser import HTMLParser

HERE = os.path.dirname(os.path.abspath(__file__))
SEARX = os.environ.get("SEARXNG_URL", "http://127.0.0.1:8888")
UA = "Mozilla/5.0 (Windows NT 10.0; Win64; x64) LocalAgent/1.0"


def _get(url, headers=None, timeout=30, data=None, method=None):
    h = {"User-Agent": UA, "Accept-Encoding": "gzip"}
    h.update(headers or {})
    req = urllib.request.Request(url, data=data, headers=h, method=method)
    with urllib.request.urlopen(req, timeout=timeout) as r:
        body = r.read()
        if r.headers.get("Content-Encoding") == "gzip":
            body = gzip.decompress(body)
        return r.headers, body


def search(query, n=8, engines=None):
    """Return [{title,url,content,engines}] from the local SearXNG."""
    p = {"q": query, "format": "json"}
    if engines:
        p["engines"] = engines
    try:
        _, b = _get(SEARX + "/search?" + urllib.parse.urlencode(p), timeout=40)
    except Exception as e:
        raise RuntimeError("SearXNG not reachable at %s (start searxng\\start_searxng.bat): %s" % (SEARX, e))
    res = json.loads(b).get("results", [])[:n]
    return [{"title": r.get("title"), "url": r.get("url"),
             "content": (r.get("content") or "")[:300], "engines": r.get("engines")} for r in res]


class _Text(HTMLParser):
    SKIP = {"script", "style", "noscript", "svg", "head", "nav", "footer", "form"}

    def __init__(self):
        super().__init__(convert_charrefs=True)
        self.out, self.skip = [], 0

    def handle_starttag(self, t, a):
        if t in self.SKIP:
            self.skip += 1
        elif t in ("p", "div", "br", "li", "tr", "h1", "h2", "h3", "h4", "pre", "section", "article"):
            self.out.append("\n")

    def handle_endtag(self, t):
        if t in self.SKIP and self.skip:
            self.skip -= 1

    def handle_data(self, d):
        if not self.skip:
            self.out.append(d)


def fetch(url, max_chars=20000):
    """GET a URL; HTML is reduced to text, JSON/text returned as is."""
    h, b = _get(url, timeout=30)
    ct = h.get("Content-Type", "")
    m = re.search(r"charset=([\w-]+)", ct)
    txt = b.decode(m.group(1) if m else "utf-8", "replace")
    if "html" in ct.lower() or txt.lstrip()[:15].lower().startswith(("<!doctype", "<html")):
        p = _Text()
        p.feed(txt)
        txt = re.sub(r"[ \t]+", " ", "".join(p.out))
        txt = re.sub(r"\n\s*\n+", "\n\n", txt).strip()
    return txt[:max_chars]


def _token():
    t = os.environ.get("HF_TOKEN") or os.environ.get("HUGGINGFACE_HUB_TOKEN")
    if not t:
        f = os.path.join(HERE, "hf_token.txt")
        if os.path.isfile(f):
            t = open(f).read().strip()
    if not t:
        raise RuntimeError("No Hugging Face token: set HF_TOKEN or put it in tools\\hf_token.txt")
    return t


def hf_note(repo, path, text, repo_type="model", append=False, message="agent note"):
    """Create/overwrite (or append to) a text file in an HF repo via the commit API."""
    tok = _token()
    base = "https://huggingface.co"
    seg = {"model": "models", "dataset": "datasets", "space": "spaces"}[repo_type]
    if append:
        rp = "" if repo_type == "model" else seg + "/"
        try:
            _, old = _get("%s/%s%s/resolve/main/%s" % (base, rp, repo, urllib.parse.quote(path)),
                          headers={"Authorization": "Bearer " + tok})
            text = old.decode("utf-8", "replace").rstrip("\n") + "\n" + text
        except urllib.error.HTTPError as e:
            if e.code != 404:
                raise
    lines = [json.dumps({"key": "header", "value": {"summary": message}}),
             json.dumps({"key": "file", "value": {"content": base64.b64encode(text.encode()).decode(),
                                                  "path": path, "encoding": "base64"}})]
    _, b = _get("%s/api/%s/%s/commit/main" % (base, seg, repo),
                headers={"Authorization": "Bearer " + tok, "Content-Type": "application/x-ndjson"},
                data="\n".join(lines).encode(), method="POST")
    return json.loads(b)


def main():
    ap = argparse.ArgumentParser()
    sub = ap.add_subparsers(dest="cmd", required=True)
    s = sub.add_parser("search"); s.add_argument("query"); s.add_argument("-n", type=int, default=8)
    f = sub.add_parser("fetch"); f.add_argument("url"); f.add_argument("--max", type=int, default=20000)
    n = sub.add_parser("hf-note"); n.add_argument("repo"); n.add_argument("path"); n.add_argument("text")
    n.add_argument("--type", default="model", choices=["model", "dataset", "space"])
    n.add_argument("--append", action="store_true")
    a = ap.parse_args()
    if a.cmd == "search":
        print(json.dumps(search(a.query, a.n), indent=1, ensure_ascii=False))
    elif a.cmd == "fetch":
        print(fetch(a.url, a.max))
    else:
        print(json.dumps(hf_note(a.repo, a.path, a.text, a.type, a.append)))


if __name__ == "__main__":
    sys.stdout.reconfigure(encoding="utf-8")
    main()