Build an earnings call summary bot for X
8-K in, earnings thread out

import json, pathlib, re, time, requests

OT_KEY = "ot_your_key"
SEC = {"User-Agent": "Your Name you@example.com"}   # the SEC requires a contact
WATCH = {"MSFT": 789019, "GOOGL": 1652044, "AMZN": 1018724, "AAPL": 320193}
POSTED = pathlib.Path("posted_8k.json")
seen = set(json.loads(POSTED.read_text())) if POSTED.exists() else set()

def earnings_8ks(cik):
    r = requests.get(f"https://data.sec.gov/submissions/CIK{cik:010d}.json", headers=SEC, timeout=20).json()["filings"]["recent"]
    for form, items, acc, when in zip(r["form"], r["items"], r["accessionNumber"], r["acceptanceDateTime"]):
        if form == "8-K" and "2.02" in items.split(","):
            yield acc, when

def press_release(cik, acc):
    base = f"https://www.sec.gov/Archives/edgar/data/{cik}/{acc.replace('-', '')}"
    files = requests.get(f"{base}/index.json", headers=SEC, timeout=20).json()["directory"]["item"]
    ex99 = next(f["name"] for f in files if re.search(r"ex-?99", f["name"], re.I))
    html = requests.get(f"{base}/{ex99}", headers=SEC, timeout=30).text
    return re.sub(r"<[^>]+>", " ", html), f"{base}/{ex99}"

for ticker, cik in WATCH.items():
    for acc, when in list(earnings_8ks(cik))[:1]:          # newest earnings 8-K only
        if acc in seen:
            continue
        text, link = press_release(cik, acc)
        thread = summarize(ticker, text)                    # your model, see below
        thread[-1] += f"\n\nRelease: {link}"              # one link, last post only
        requests.post(
            "https://opentweet.io/api/v1/posts",
            headers={"Authorization": f"Bearer {OT_KEY}"},
            json={"is_thread": True, "thread_tweets": thread, "publish_now": True},
        )
        seen.add(acc)
    time.sleep(0.2)                                         # stay far below 10 req/s

POSTED.write_text(json.dumps(sorted(seen)))

def summarize(ticker, release_text):
    """Return 3-4 posts. Extract from the text, never from memory."""
    prompt = f"""From this {ticker} earnings press release, write an X thread
of 4 posts, each under 280 characters:
1. ${ticker} headline: revenue, EPS, and year-over-year change.
2. The segment that moved the most, with its number.
3. Guidance for next quarter, quoted, or "No guidance given".
4. One line on what to watch on the call.
Use only numbers that appear in the text. No emoji, no "BREAKING",
no investment advice. Return JSON: {{"posts": [...]}}

{release_text[:30000]}"""
    posts = call_your_model(prompt)["posts"]
    # Reject the draft if any number in it is not in the release.
    for n in re.findall(r"\d[\d,.]*", " ".join(posts)):
        if n.rstrip(".,") not in release_text:
            raise ValueError(f"{n} not in source")
    return posts