#!/usr/bin/env python
"""Text Summarizer — paste text or a URL, get a clean TL;DR + key points.
Streams the summary live. Backend auto-selects: Anthropic Claude API if
ANTHROPIC_API_KEY is set, otherwise a local Ollama model.
pip install flask requests anthropic beautifulsoup4
python app.py # http://127.0.0.1:8500
Project #2 of the "30 Projects in 15 Days" challenge — GritAI.
"""
import os, re, json, html
from flask import Flask, request, Response
PORT = int(os.environ.get("PORT", "8500"))
ANTHROPIC_KEY = os.environ.get("ANTHROPIC_API_KEY")
ANTHROPIC_MODEL = os.environ.get("ANTHROPIC_MODEL", "claude-sonnet-5")
OLLAMA_URL = os.environ.get("OLLAMA_URL", "http://127.0.0.1:11434")
OLLAMA_MODEL = os.environ.get("OLLAMA_MODEL", "qwen2.5:7b")
MAX_CHARS = 40000 # cap input to control tokens/cost
LENGTHS = {
"brief": "a tight 2-4 sentence TL;DR that captures the essence. No bullet points, no headings.",
"standard": "first a 2-3 sentence overview paragraph, then a blank line, then 4-6 key points, each on its own line starting with '- '.",
"detailed": "first a short overview paragraph, then a blank line, then 7-10 key points each on its own line starting with '- ', preserving important specifics (names, numbers, dates, outcomes).",
}
def system_prompt(length):
return (
"You are an expert editor who writes clear, coherent summaries.\n"
"OUTPUT FORMAT: " + LENGTHS.get(length, LENGTHS["standard"]) + "\n"
"RULES:\n"
"- Be faithful to the source; never add facts, opinions, or details not present.\n"
"- Write clean, natural plain text. Do NOT use markdown (**bold**, # headings). "
"Strip any citation markers like [1], footnote symbols, or wiki artifacts.\n"
"- Bullets (when used) start with '- ', are clear parallel phrases or sentences, and are non-redundant.\n"
"- Lead with the single most important point. Be specific, not vague. Do not restate these instructions."
)
_CITE = re.compile(r'\[(?:\d+|[a-z]{1,3}|note \d+|citation needed|clarification needed|edit|update|when\?|who\?|why\?)\]', re.I)
def _clean(text):
text = _CITE.sub("", text or "")
return "\n".join(l.strip() for l in text.splitlines() if l.strip())
def fetch_url_text(url):
"""Fetch a page and extract the main article text (trafilatura, bs4 fallback)."""
import requests
r = requests.get(url, timeout=20, headers={"User-Agent": "Mozilla/5.0 (compatible; GritAI-Summarizer/1.0)"})
r.raise_for_status()
doc = r.text
text = ""
try:
import trafilatura
text = trafilatura.extract(doc, include_comments=False, include_tables=False, favor_precision=True) or ""
except Exception:
text = ""
if len(text) < 200: # fallback: main-content via bs4
from bs4 import BeautifulSoup
soup = BeautifulSoup(doc, "html.parser")
for tag in soup(["script", "style", "nav", "header", "footer", "aside", "noscript", "form"]):
tag.decompose()
main = (soup.find("main") or soup.find("article") or soup.find(id="mw-content-text")
or soup.body or soup)
text = main.get_text("\n")
text = _clean(text)
if len(text) < 200:
raise ValueError("Couldn't extract readable text from that page (it may be JavaScript-only or blocked).")
return text
def summarize_stream(text, length):
sys = system_prompt(length)
user = "Summarize the following:\n\n" + text[:MAX_CHARS]
if ANTHROPIC_KEY:
import anthropic
client = anthropic.Anthropic(api_key=ANTHROPIC_KEY)
with client.messages.stream(model=ANTHROPIC_MODEL, max_tokens=800, system=sys,
messages=[{"role": "user", "content": user}]) as s:
for t in s.text_stream:
yield t
return
import requests
with requests.post(f"{OLLAMA_URL}/api/chat", stream=True, timeout=120, json={
"model": OLLAMA_MODEL, "stream": True,
"messages": [{"role": "system", "content": sys}, {"role": "user", "content": user}],
"options": {"temperature": 0.2}}) as r:
r.raise_for_status()
for line in r.iter_lines():
if not line:
continue
d = json.loads(line)
if d.get("message", {}).get("content"):
yield d["message"]["content"]
if d.get("done"):
break
app = Flask(__name__)
@app.route("/")
def home():
return Response(PAGE, mimetype="text/html")
@app.route("/api/summarize", methods=["POST"])
def api_summarize():
body = request.get_json(force=True)
text = (body.get("text") or "").strip()
url = (body.get("url") or "").strip()
length = body.get("length", "standard")
def gen():
import sys as _sys
try:
src = text
if url:
yield "data: " + json.dumps({"t": "Fetching the page…\n\n"}) + "\n\n"
src = fetch_url_text(url)
if not src:
yield "data: " + json.dumps({"t": "Please paste some text or a URL to summarize."}) + "\n\n"
yield "data: [DONE]\n\n"
return
if url:
yield "data: " + json.dumps({"reset": True}) + "\n\n" # clear the "fetching" note
for chunk in summarize_stream(src, length):
yield "data: " + json.dumps({"t": chunk}) + "\n\n"
except Exception as e:
print("summarize error:", type(e).__name__, file=_sys.stderr, flush=True)
yield "data: " + json.dumps({"t": "Sorry — something went wrong. Please try again."}) + "\n\n"
yield "data: [DONE]\n\n"
return Response(gen(), mimetype="text/event-stream",
headers={"Cache-Control": "no-cache", "X-Accel-Buffering": "no"})
PAGE = """
Text Summarizer — GritAI
GRITAIText Summarizer▸ LiveSDVOSBLawton OK
Get the gist in seconds.
Paste an article, email, or notes — or drop in a link — and get a faithful TL;DR with the key points, streamed live.
01Source
— or summarize a link —
›
02Length
03Summary
"""
if __name__ == "__main__":
backend = "Claude API" if ANTHROPIC_KEY else f"Ollama ({OLLAMA_MODEL} @ {OLLAMA_URL})"
print(f"Text Summarizer on http://127.0.0.1:{PORT} [backend: {backend}]", flush=True)
app.run(host="0.0.0.0", port=PORT, threaded=True)