spark-control/image/app/download.py

"""Drive `./hf-download.sh <repo>` on Spark 1 via SSH and stream progress.

Parses `huggingface-hub` tqdm-style progress lines like:

    Downloading (incomplete total...):   8%|▏  | 2.06G/25.1G [03:20<18:35, 20.6MB/s]

into a structured percent + bytes done / total + ETA payload that the
front-end can render as a clean progress bar.
"""
from __future__ import annotations
import asyncio
import re
import uuid
from dataclasses import dataclass, field
from datetime import datetime, timezone
from typing import Literal, Optional

from .config import Settings
from .shellsafe import quote_arg, validate_repo
from .ssh import ssh_stream, StreamHandle


Mode = Literal["spark1", "spark2", "cluster"]


def build_download_command(repo: str, flags: str = "") -> str:
    """Remote shell command that drives hf-download.sh on a Spark.

    Prepends ~/.local/bin to PATH. hf-download.sh shells out to `uvx` (Astral's
    uv), and the official uv installer drops its binaries in ~/.local/bin — but
    our SSH session is non-interactive, so it never sources the user's profile
    and ~/.local/bin is off PATH, leaving `uvx` as "command not found". $HOME
    expands server-side, so this stays correct for any adopter/user. `repo` is
    shlex-quoted at the sink (validate_repo gates the charset upstream).
    """
    serve = f"./hf-download.sh {quote_arg(repo)} {flags}".strip()
    return f'export PATH="$HOME/.local/bin:$PATH" && cd ~/spark-vllm-docker && {serve}'


_TQDM_RE = re.compile(
    r"(\d+(?:\.\d+)?)\s*%\s*\|.*?\|\s*"
    r"([\d.]+[KMG]?B?)\s*/\s*([\d.]+[KMG]?B?)\s*"
    r"\[(\d+:\d+:?\d*)\s*<\s*(\d+:\d+:?\d*),?\s*"
    r"([\d.]+\s*\w+/s)?",
    re.IGNORECASE,
)

# huggingface_hub also emits "Fetching N files: pct%|..."
_FETCHING_RE = re.compile(
    r"Fetching\s+(\d+)\s+files:\s+(\d+(?:\.\d+)?)\s*%",
    re.IGNORECASE,
)


@dataclass
class DownloadProgress:
    percent: float = 0.0
    downloaded: str = ""
    total: str = ""
    elapsed: str = ""
    eta: str = ""
    rate: str = ""
    phase: str = "Starting…"


@dataclass
class DownloadJob:
    id: str
    repo: str
    mode: Mode
    started_at: str
    state: str = "starting"   # starting | downloading | copying | done | failed
    lines: list[str] = field(default_factory=list)
    progress: DownloadProgress = field(default_factory=DownloadProgress)
    returncode: Optional[int] = None
    finished_at: Optional[str] = None

    def append(self, line: str) -> None:
        self.lines.append(line)
        if len(self.lines) > 800:
            del self.lines[: len(self.lines) - 800]


class DownloadManager:
    def __init__(self, settings: Settings) -> None:
        self.settings = settings
        self.lock = asyncio.Lock()
        self.jobs: dict[str, DownloadJob] = {}
        self.current_job_id: Optional[str] = None

    def get(self, job_id: str) -> DownloadJob | None:
        return self.jobs.get(job_id)

    async def trigger(self, repo: str, mode: Mode) -> DownloadJob:
        validate_repo(repo)  # raises ValueError on anything but a clean 'org/name'
        if self.lock.locked():
            raise RuntimeError("A download is already in progress")
        job = DownloadJob(
            id=uuid.uuid4().hex[:8],
            repo=repo,
            mode=mode,
            started_at=datetime.now(timezone.utc).isoformat(),
        )
        self.jobs[job.id] = job
        self.current_job_id = job.id
        asyncio.create_task(self._run(job))
        return job

    async def _run(self, job: DownloadJob) -> None:
        async with self.lock:
            try:
                await self._do(job)
                if job.state != "failed":
                    job.state = "done"
                    job.returncode = 0
                    job.progress.percent = 100.0
                    job.progress.phase = "Done"
            except Exception as e:
                job.append(f"[error] {type(e).__name__}: {e}")
                job.state = "failed"
                if job.returncode is None:
                    job.returncode = 1
            finally:
                job.finished_at = datetime.now(timezone.utc).isoformat()
                if self.current_job_id == job.id:
                    self.current_job_id = None

    async def _do(self, job: DownloadJob) -> None:
        s = self.settings
        # Pick the SSH target and hf-download flags from the mode.
        if job.mode == "spark2":
            target_host, target_user = s.spark2_host, s.spark2_user
            flags = ""
        elif job.mode == "cluster":
            target_host, target_user = s.spark1_host, s.spark1_user
            flags = "-c --copy-parallel"
        else:  # spark1
            target_host, target_user = s.spark1_host, s.spark1_user
            flags = ""
        if not target_host or not target_user:
            raise RuntimeError(f"{job.mode} host not configured")

        cmd = build_download_command(job.repo, flags)
        job.append(f"$ {cmd}")
        job.state = "downloading"
        job.progress.phase = "Connecting to Hugging Face…"

        handle = StreamHandle()
        async for line in ssh_stream(target_host, target_user, cmd, s, handle=handle):
            job.append(line)
            self._update_progress(job, line)

        rc = handle.returncode or 0
        if rc != 0:
            job.state = "failed"
            job.returncode = rc

    def _update_progress(self, job: DownloadJob, line: str) -> None:
        p = job.progress
        # Phase transitions from log content
        if "Copying model" in line or "Parallel copy enabled" in line:
            job.state = "copying"
            p.phase = "Copying to peer Sparks…"
        elif "Download completed" in line:
            p.phase = "Download complete, finalizing…"
        elif "Copy complete" in line:
            p.phase = "Copy complete"
        elif "Still waiting to acquire lock" in line:
            p.phase = "Waiting for lock (another download in progress)…"

        # Check the "Fetching N files" pattern first (it could match TQDM_RE otherwise).
        m2 = _FETCHING_RE.search(line)
        if m2:
            p.percent = float(m2.group(2))
            p.phase = f"Fetching {m2.group(1)} files"
            return

        m = _TQDM_RE.search(line)
        if m:
            p.percent = float(m.group(1))
            p.downloaded = m.group(2)
            p.total = m.group(3)
            p.elapsed = m.group(4)
            p.eta = m.group(5)
            p.rate = m.group(6) or p.rate
            if job.state != "copying":
                p.phase = "Downloading"
            return