start_app.py

import json
import os
import re
import subprocess
import time
import yaml

import gradio as gr
import pandas as pd
import requests
from huggingface_hub import HfApi, get_token


CMD = ["python" ,"run_job.py"]
ARG_NAMES = ["<src>", "<dst>", "<query>", "[-c config]", "[-s split]", "[-p private]"]
SPACE_ID = os.environ.get("SPACE_ID") or "lhoestq/run-duckdb-jobs"

CONTENT = """
## Usage: 

```bash
curl -L 'https://huggingface.co/api/jobs/<username>' \
-H 'Content-Type: application/json' \
-H 'Authorization: Bearer <hf_token>' \
-d '{{
    "spaceId": "{SPACE_ID}", 
    "command": {CMD},
    "arguments": {ARG_NAMES},
    "environment": {{"HF_TOKEN": <hf_token>}},
    "flavor": "cpu-basic" 
}}'
```

## Example:
"""

with open("README.md") as f:
    METADATA = yaml.safe_load(f.read().split("---\n")[1])
TITLE = METADATA["title"]
SHORT_DESCRIPTION = METADATA.get("short_description")
EMOJI = METADATA["emoji"]

try:
    process = subprocess.run(CMD + ["--help"], stdout=subprocess.PIPE, stderr=subprocess.PIPE)
    HELP = not process.returncode and (process.stdout or process.stderr).decode()
except Exception:
    HELP = False

DRY_RUN = bool(HELP) and bool(m :=re.search("--dry(-|_)run", HELP)) and m.group(0)

def parse_log(line: str, pbars: dict[str, float] = None):
    if line.startswith("data: {"):
        data = json.loads(line[len("data: "):])
        data, timestamp = data["data"], data["timestamp"]
        if pbars is not None and data.startswith("===== Job started at"):
            pbars.pop("Starting ⚙️", None)
            pbars["Running 🏃"] = 0.0
            return f"[{timestamp}] {data}\n\n"
        elif pbars is not None and (percent_match := re.search("\\d+(?:\\.\\d+)?%", data)) and any(c in data.split("%")[1][:10] for c in "|█▌"):
            pbars.pop("Running 🏃", None)
            [pbars.pop(desc) for desc, percent in pbars.items() if percent == 1.]
            percent = float(percent_match.group(0)[:-1]) / 100
            desc = data[:percent_match.start()].strip() or "Progress"
            pbars[desc] = percent
        else:
            return f"[{timestamp}] {data}\n\n"
    return ""

def dry_run(src, config, split, dst, query):
    if not all([src, dst, query]):
        raise gr.Error("Please fill source, destination and query.")
    args = ["--src", src] + (["--config", config] if config else []) + (["--split", split] if split else []) + [ "--dst", dst, "--query", query, DRY_RUN]
    cmd = CMD + args
    logs = "Job:\n\n```bash\n" + " ".join('"' + arg.replace('"', '\"""') + '"' if " " in arg else arg for arg in cmd) + "\n```\nOutput:\n\n"
    yield {output_markdown: logs, progress_labels: gr.Label(visible=False), details_accordion: gr.Accordion(open=True)}
    process = subprocess.Popen(cmd, stdout=subprocess.PIPE)
    for line in iter(process.stdout.readline, b""):
        logs += line.decode()
        yield {output_markdown: logs}

def run(src, config, split, dst, query, oauth_token: gr.OAuthToken | None, profile: gr.OAuthProfile | None):
    if not all([src, dst, query]):
        raise gr.Error("Please fill source, destination and query.")
    if oauth_token and profile:
        token = oauth_token.token
        username = profile.username
    elif (token := get_token()):
        username = HfApi().whoami(token=token)["name"]
    else:
        raise gr.Error("Please log in to run the job.")
    args = ["--src", src] + (["--config", config] if config else []) + (["--split", split] if split else []) + [ "--dst", dst, "--query", query]
    cmd = CMD + args
    logs = "Job:\n\n```bash\n" + " ".join('"' + arg.replace('"', '\"""') + '"' if " " in arg else arg for arg in cmd) + "\n```\nOutput:\n\n"
    pbars = {}
    yield {output_markdown: logs, progress_labels: gr.Label(pbars, visible=bool(pbars))}
    resp = requests.post(
        f"https://huggingface.co/api/jobs/{username}",
        json={
            "spaceId": SPACE_ID,
            "arguments": args,
            "command":  CMD,
            "environment": {"HF_TOKEN": token},
            "flavor": "cpu-basic" 
        },
        headers={"Authorization": f"Bearer {token}"}
    )
    if resp.status_code != 200:
        logs += resp.text
        pbars = {"Finished with an error ❌": 1.0}
    else:
        job_id = resp.json()["metadata"]["job_id"]
        pbars = {"Starting ⚙️": 0.0}
        yield {output_markdown: logs, progress_labels: gr.Label(pbars, visible=bool(pbars))}
        resp = requests.get(
            f"https://huggingface.co/api/jobs/{username}/{job_id}/logs-stream",
            headers={"Authorization": f"Bearer {token}"},
            stream=True
        )
        for line in resp.iter_lines():
            logs += parse_log(line.decode("utf-8"), pbars=pbars)
            yield {output_markdown: logs, progress_labels: gr.Label(pbars, visible=bool(pbars))}
        job_status = {"status": {"stage": "RUNNING"}}
        while True:
            job_status = requests.get(
                f"https://huggingface.co/api/jobs/{username}/{job_id}",
                headers={"Authorization": f"Bearer {token}"}
            ).json()
            if job_status["status"]["stage"] == "RUNNING":
                time.sleep(1)
            else:
                break
        if job_status["status"]["stage"] == "COMPLETED":
            pbars = {"Finished ✅": 1.0}
        else:
            logs += f'{job_status["status"]["message"]} ({job_status["status"]["error"]})'
            pbars = {"Finished with an error ❌": 1.0}
    yield {output_markdown: logs, progress_labels: gr.Label(pbars, visible=bool(pbars))}


READ_FUNCTIONS = ("pl.read_parquet", "pl.read_csv", "pl.read_json")
NUM_TRENDING_DATASETS = 10

with gr.Blocks() as demo:
    with gr.Row():
        with gr.Column(scale=10):
            gr.Markdown(f"# {TITLE} {EMOJI}")
            if SHORT_DESCRIPTION:
                gr.Markdown(SHORT_DESCRIPTION)
        with gr.Column():
            gr.LoginButton()
    gr.Markdown(CONTENT.format(SPACE_ID=SPACE_ID, CMD=json.dumps(CMD), ARG_NAMES=json.dumps(ARG_NAMES)))
    with gr.Row():
        with gr.Column(scale=10):
            with gr.Row():
                loading_codes_json = gr.JSON([], visible=False)
                dataset_dropdown = gr.Dropdown(label="Source Dataset", allow_custom_value=True, scale=10)
                subset_dropdown = gr.Dropdown(info="Subset", allow_custom_value=True, show_label=False, visible=False)
                split_dropdown = gr.Dropdown(info="Split", allow_custom_value=True, show_label=False, visible=False)
        with gr.Column(min_width=60):
            gr.HTML("<div style='font-size: 4em;'>→</div>")
        with gr.Column(scale=10):
            dst_dropdown = gr.Dropdown(label="Destination Dataset", allow_custom_value=True)
    query_textarea = gr.Textbox(label="SQL Query", lines=2, max_lines=300, placeholder="SELECT * FROM src;", value="SELECT * FROM src;")
    with gr.Row():
        run_button = gr.Button("Run", scale=10, variant="primary")
        if DRY_RUN:
            dry_run_button = gr.Button("Dry-Run")
    progress_labels= gr.Label(visible=False, label="Progress")
    with gr.Accordion("Details", open=False) as details_accordion:
        output_markdown = gr.Markdown(label="Output logs")
    run_button.click(run, inputs=[dataset_dropdown, subset_dropdown, split_dropdown, dst_dropdown, query_textarea], outputs=[details_accordion, progress_labels, output_markdown])
    if DRY_RUN:
        dry_run_button.click(dry_run, inputs=[dataset_dropdown, subset_dropdown, split_dropdown, dst_dropdown, query_textarea], outputs=[details_accordion, progress_labels, output_markdown])

    def show_subset_dropdown(dataset: str):
        if dataset and "/" not in dataset.strip().strip("/"):
            return []
        resp = requests.get(f"https://datasets-server.huggingface.co/compatible-libraries?dataset={dataset}", timeout=3).json()
        loading_codes = ([lib["loading_codes"] for lib in resp.get("libraries", []) if lib["function"] in READ_FUNCTIONS] or [[]])[0] or []
        subsets = [loading_code["config_name"] for loading_code in loading_codes]
        subset = (subsets or [""])[0]
        return dict(choices=subsets, value=subset, visible=len(subsets) > 1, key=hash(str(loading_codes))), loading_codes

    def show_split_dropdown(subset: str, loading_codes: list[dict]):
        splits = ([list(loading_code["arguments"]["splits"]) for loading_code in loading_codes if loading_code["config_name"] == subset] or [[]])[0]
        split = (splits or [""])[0]
        return dict(choices=splits, value=split, visible=len(splits) > 1, key=hash(str(loading_codes) + subset))

    @demo.load(outputs=[dataset_dropdown, loading_codes_json, subset_dropdown, split_dropdown])
    def _fetch_datasets(request: gr.Request):
        dataset = "CohereForAI/Global-MMLU"
        datasets = [dataset] + [ds.id for ds in HfApi().list_datasets(limit=NUM_TRENDING_DATASETS, sort="trendingScore", direction=-1) if ds.id != dataset]
        subsets, loading_codes = show_subset_dropdown(dataset)
        splits = show_split_dropdown(subsets["value"], loading_codes)
        return {
            dataset_dropdown: gr.Dropdown(choices=datasets, value=dataset),
            loading_codes_json: loading_codes,
            subset_dropdown: gr.Dropdown(**subsets),
            split_dropdown: gr.Dropdown(**splits),
        }
    
    @dataset_dropdown.select(inputs=[dataset_dropdown], outputs=[subset_dropdown, split_dropdown])
    def _show_subset_dropdown(dataset: str):
        subsets, loading_codes = show_subset_dropdown(dataset)
        splits = show_split_dropdown(subsets["value"], loading_codes)
        return {
            subset_dropdown: gr.Dropdown(**subsets),
            split_dropdown: gr.Dropdown(**splits),
        }
    
    @subset_dropdown.select(inputs=[dataset_dropdown, subset_dropdown, loading_codes_json], outputs=[split_dropdown])
    def _show_split_dropdown(dataset: str, subset: str, loading_codes: list[dict]):
        splits = show_split_dropdown(subset, loading_codes)
        return {
            split_dropdown: gr.Dropdown(**splits),
        }

if HELP:
    with demo.route("Help", "/help"):
        gr.Markdown(f"# Help\n\n```\n{HELP}\n```")

with demo.route("Jobs", "/jobs") as page:
    gr.Markdown("# Jobs")
    jobs_dataframe = gr.DataFrame(datatype="markdown")

    @page.load(outputs=[jobs_dataframe])
    def list_jobs(oauth_token: gr.OAuthToken | None, profile: gr.OAuthProfile | None):
        if oauth_token and profile:
            token = oauth_token.token
            username = profile.username
        elif (token := get_token()):
            username = HfApi().whoami(token=token)["name"]
        else:
            return pd.DataFrame({"Log in to see jobs": []})
        resp = requests.get(
            f"https://huggingface.co/api/jobs/{username}",
            headers={"Authorization": f"Bearer {token}"}
        )
        return pd.DataFrame([
            {
                "id": job["metadata"]["id"],
                "created_at": job["metadata"]["created_at"],
                "stage": job["compute"]["status"]["stage"],
                "output": f'[logs](https://huggingface.co/api/jobs/{username}/{job["metadata"]["id"]}/logs-stream)',
                "command": str(job["compute"]["spec"]["extra"]["command"]),
                "args": str(job["compute"]["spec"]["extra"]["args"]),
            }
            for job in resp.json()
            if job["compute"]["spec"]["extra"]["input"]["spaceId"] == SPACE_ID
        ])

if __name__ == "__main__":
    demo.launch(server_name="0.0.0.0")