Keep a Drive folder searchable in your Supabase vector table

from clickruntime import on_trigger, credentials, http, state, notify, report_fetched
FOLDER, SB = "your-knowledge-folder-id", "https://yourproject.supabase.co/rest/v1/documents"
D = "https://www.googleapis.com/drive/v3/files"
@on_trigger
def run(event):
    h = {"Authorization": f"Bearer {credentials.connection('google', 'work')}"}
    since = state.get("since") or "2026-01-01T00:00:00Z"
    files = http.get(D, headers=h, params={"q": f"'{FOLDER}' in parents and modifiedTime > '{since}' and trashed = false",
                                           "orderBy": "modifiedTime", "fields": "files(id,name,mimeType,modifiedTime)"}).json()["files"]
    report_fetched(len(files))
    if not files:
        return {"changed": 0}
    key = credentials.secret("SUPABASE_SERVICE_KEY")
    sb = {"apikey": key, "Authorization": f"Bearer {key}"}
    emb = {"Authorization": f"Bearer {credentials.secret('OPENAI_API_KEY')}"}
    failed = []
    for f in files:
        try:
            gdoc = f["mimeType"].startswith("application/vnd.google-apps")
            text = (http.get(f"{D}/{f['id']}/export", headers=h, params={"mimeType": "text/plain"}) if gdoc
                    else http.get(f"{D}/{f['id']}", headers=h, params={"alt": "media"})).text
            chunks = [text[i:i + 1500] for i in range(0, len(text), 1300)][:200]
            vecs = http.post("https://api.openai.com/v1/embeddings", headers=emb, json={"model": "text-embedding-3-small", "input": chunks}).json()["data"]
            http.delete(SB, headers=sb, params={"file_id": f"eq.{f['id']}"}).raise_for_status()
            http.post(SB, headers=sb, json=[{"file_id": f["id"], "file_name": f["name"], "content": c, "embedding": v["embedding"]}
                                            for c, v in zip(chunks, vecs)]).raise_for_status()
        except Exception as e: failed.append(f"{f['name']}: {e}")
    if failed: notify.email_me("Drive indexing failed for some files", "\n".join(failed))
    state.set("since", files[-1]["modifiedTime"])
    return {"indexed": len(files) - len(failed), "failed": len(failed)}