app.py

import re
import os
import html

import gradio as gr
from opensearchpy import OpenSearch, RequestsHttpConnection
from requests_aws4auth import AWS4Auth


awsauth = AWS4Auth(
    os.environ.get("ACCESS_KEY"),
    os.environ.get("SECRET_KEY"),
    "us-east-1",
    "es",
)

es = OpenSearch(
    hosts=[{"host": os.environ.get("HOST"), "port": 443}],
    http_auth=awsauth,
    use_ssl=True,
    verify_certs=True,
    connection_class=RequestsHttpConnection,
    http_compress=True,
    timeout=200,
)


def mark_tokens_bold(text, tokens):
    for token in tokens:
        if token in ["<", "b", "/", ">"]:
            continue
        pattern = re.escape(token)  # r"\b" + re.escape(token) + r"\b"
        text = re.sub(pattern, "<b>" + token + "</b>", text)
    return text


def process_results(results, query):
    if len(results) == 0:
        return """<br><p>No results retrieved.</p><br><hr>"""

    results_html = ""
    for result in results:
        text_html = result["text"]
        if query.startswith('"') and query.endswith('"'):
            text_html = mark_tokens_bold(text_html, query[1:-1].split(" "))
        else:
            text_html = mark_tokens_bold(text_html, query.split(" "))
        repository = result["repository"]
        commit_id = result["commit_id"]
        path = result["path"]
        license = result["license"]
        language = result["language"]
        code_height = min(
            600, len(text_html.split("\n")) * 20
        )  # limit to maximum height of 600px
        results_html += """\
        <p style='font-size:16px; text-align: left;'><b>Source: </b><a target="_blank" href="https://github.com/{}/blob/{}{}">{}</a>&nbsp;&nbsp;&nbsp;|&nbsp;&nbsp;&nbsp;<b>Language:</b> \
        <span style='color: #00134d;'>{}</span>&nbsp;&nbsp;&nbsp;|&nbsp;&nbsp;&nbsp;<b>Licenses: </b><span style='color: #00134d;'>{}</span></p>
        <pre style='height: {}px; overflow-y: scroll; overflow-x: hidden; color: #d9d9d9;border: 1px solid #e6b800; padding: 10px'><code>{}</code></pre>
        <hr>
        """.format(
            repository,
            commit_id,
            path,
            f"{repository}/blob/{commit_id}{path}",
            language,
            license,
            code_height,
            text_html,
        )
    return results_html


def match_query(query, num_results=10):
    query_body = {"query": {"match": {"content": query}}, "size": num_results}

    response = es.search(index=os.environ.get("INDEX"), body=query_body)
    hits = [hit["_source"] for hit in response["hits"]["hits"]]
    return hits


def phrase_query(query, num_results=10):
    query_body = {"query": {"match_phrase": {"content": query}}, "size": num_results}

    response = es.search(index=os.environ.get("INDEX"), body=query_body)
    hits = [hit["_source"] for hit in response["hits"]["hits"]]
    return hits


def search(query, num_results=10):
    print(es.ping())
    query = query[:200]
    if query.startswith('"') and query.endswith('"'):
        response = phrase_query(query[1:-1], num_results=num_results)
    else:
        response = match_query(query, num_results=num_results)
    results = [
        {
            "text": html.escape(hit["content"]),
            "repository": hit["repository"],
            "commit_id": hit["commit_id"],
            "path": hit["path"],
            "license": ", ".join(hit["scancode_licenses"]) if hit["gh_license"] is None else hit["gh_license"],
            "language": hit["language"],
        }
        for hit in response
    ]
    return process_results(results, query)


description = """# <p style="text-align: center;"><span style='color: #e6b800;'>StarCoder2:</span> Dataset Search 🔍 </p>
<span>When using <a href="https://huggingface.co/bigcode/starcoder2-15b" style="color: #e6b800;">StarCoder2</a> to generate code, it might produce close or exact copies of code in the pretraining dataset. Identifying such cases can provide important context, and help credit the original developer of the code. With this search tool, our aim is to help in identifying if the code belongs to an existing repository. For exact matches, enclose your query in double quotes. <br><br><i>This first iteration of the search tool truncates queries down to 200 characters, so as not to overwhelm the server it is currently running on. Please note that this is not a production-ready app, but rather a research tool that we make available as a proof-of-concept. If you need a reliable search app for your business or research, we would advise you to index the dataset yourself.</i></span>"""

theme = gr.themes.Monochrome(
    primary_hue="indigo",
    secondary_hue="blue",
    neutral_hue="slate",
    radius_size=gr.themes.sizes.radius_sm,
    font=[
        gr.themes.GoogleFont("Open Sans"),
        "ui-sans-serif",
        "system-ui",
        "sans-serif",
    ],
)
css = ".generating {visibility: hidden}"

monospace_css = """
#q-input textarea {
    font-family: monospace, 'Consolas', Courier, monospace;
}
"""

css = monospace_css + ".gradio-container {color: black}"


if __name__ == "__main__":
    demo = gr.Blocks(
        theme=theme,
        css=css,
    )

    with demo:
        with gr.Row():
            gr.Markdown(value=description)
        with gr.Row():
            query = gr.Textbox(
                lines=5,
                placeholder="Type your query here...",
                label="Query",
                elem_id="q-input",
            )
        with gr.Row():
            k = gr.Slider(1, 100, value=10, step=1, label="Max Results")
        with gr.Row():
            submit_btn = gr.Button("Submit")
        with gr.Row():
            results = gr.HTML(label="Results", value="")

        def submit(query, k, lang="en"):
            query = query.strip()
            if query is None or query == "":
                return "", ""
            return {
                results: search(query, k),
            }

        query.submit(fn=submit, inputs=[query, k], outputs=[results])
        submit_btn.click(submit, inputs=[query, k], outputs=[results])

    demo.queue()
    demo.launch(debug=True)