/.github/scripts/issue_triage.py

#!/usr/bin/env python
from __future__ import annotations
import os, sys, json, datetime, pathlib, textwrap, requests
from openai import OpenAI

REPO = "voideditor/void"
CACHE_FILE = pathlib.Path(".github/triage_cache.json")
STAMP_FILE = pathlib.Path(".github/last_triage.txt")

THEMES_MD = textwrap.dedent("""\
1. 🔗 LLM Integration & Provider Support
2. 🖥 App Build & Platform Compatibility
3. 🎯 Prompt, Token, and Cost Management
4. 🧩 Editor UX & Interaction Design
5. 🤖 Agent & Automation Features
6. ⚙️ System Config & Environment Setup
7. 🗃 Meta: Feature Comparison, Structure, and Naming
""").strip()

client  = OpenAI(api_key=os.environ["OPENAI_API_KEY"])
headers = {"Authorization": f"Bearer {os.environ['GITHUB_TOKEN']}"}


# ───────── helpers ────────────────────────────────────────────────────────
def utc_iso_now() -> str:
    return datetime.datetime.utcnow().replace(microsecond=0, tzinfo=datetime.timezone.utc).isoformat()

def read_stamp() -> str:
    return STAMP_FILE.read_text().strip() if STAMP_FILE.exists() else "1970-01-01T00:00:00Z"

def save_stamp():
    STAMP_FILE.parent.mkdir(parents=True, exist_ok=True)
    STAMP_FILE.write_text(utc_iso_now())

def load_cache() -> dict[int, str]:
    return json.loads(CACHE_FILE.read_text()) if CACHE_FILE.exists() else {}

def save_cache(d: dict[int, str]):
    CACHE_FILE.parent.mkdir(parents=True, exist_ok=True)
    CACHE_FILE.write_text(json.dumps(d, indent=2))

def fetch_open_issues(since_iso: str | None = None) -> list[dict]:
    issues, page = [], 1
    while True:
        url = (
            f"https://api.github.com/repos/{REPO}/issues"
            f"?state=open&per_page=100&page={page}"
            + (f"&since={since_iso}" if since_iso else "")
        )
        chunk = requests.get(url, headers=headers).json()
        if not chunk or (isinstance(chunk, dict) and chunk.get("message")):
            break
        issues.extend(i for i in chunk if "pull_request" not in i)
        page += 1
    return issues


# ───────── main ───────────────────────────────────────────────────────────
last_stamp = read_stamp()
changed    = fetch_open_issues(since_iso=last_stamp)

# Fallback if **nothing** changed AND we have *no* existing output
if not changed:
    cache_exists = CACHE_FILE.exists()
    wiki_exists  = pathlib.Path("wiki/Issue-Categories.md").exists()
    if not cache_exists or not wiki_exists:
        # first run or someone wiped the wiki → build from scratch
        print("⏩ First run or empty wiki — fetching ALL open issues.", file=sys.stderr)
        changed = fetch_open_issues()         # full list
    else:
        print(f"✅ No issues updated since {last_stamp}. Nothing to classify.", file=sys.stderr)
        save_stamp()
        sys.exit(0)

# ---------------------------------------------------------------- prompt
issue_lines = "\n".join(f"- {i['title']} ({i['html_url']})" for i in changed)
prompt = textwrap.dedent(f"""\
You are an AI assistant helping triage GitHub issues into exactly 7 predefined themes.

Each issue must go into exactly one of the themes below:

{THEMES_MD}

Format your output in Markdown like:
## 🎯 Prompt, Token, and Cost Management
- [#123](https://github.com/org/repo/issues/123) – Title here

Classify these issues:
{issue_lines}
""")

resp = client.chat.completions.create(
    model="gpt-4.1",
    messages=[{"role": "user", "content": prompt}],
    temperature=0.2,
)

md = resp.choices[0].message.content

# ---------------------------------------------------------------- parse GPT
new_map: dict[int, str] = {}
current = None
for ln in md.splitlines():
    if ln.startswith("##"):
        current = ln.lstrip("# ").strip()
    elif ln.lstrip().startswith("- [#"):
        try:
            num = int(ln.split("[#")[1].split("]")[0])
            new_map[num] = current
        except Exception:
            pass  # ignore malformed lines

cache = load_cache()
cache.update(new_map)
save_cache(cache)
save_stamp()

# ---------------------------------------------------------------- rebuild wiki
order = [
    "🔗 LLM Integration & Provider Support",
    "🖥 App Build & Platform Compatibility",
    "🎯 Prompt, Token, and Cost Management",
    "🧩 Editor UX & Interaction Design",
    "🤖 Agent & Automation Features",
    "⚙️ System Config & Environment Setup",
    "🗃 Meta: Feature Comparison, Structure, and Naming",
]

sections: dict[str, list[int]] = {t: [] for t in order}

# ── fetch ALL current open issues once  (PRs filtered out) ────────────────
title_map: dict[int, tuple[str, str]] = {}
open_now: set[int] = set()

page = 1
while True:
    batch = fetch_open_issues(since_iso=None) if page == 1 else []
    if not batch:
        break
    for it in batch:
        num = it["number"]
        title_map[num] = (it["title"], it["html_url"])
        open_now.add(num)
    page += 1

# 🧹 drop any cached IDs that are no longer open issues (e.g., became a PR or were closed)
for stale in set(cache) - open_now:
    del cache[stale]
save_cache(cache)            # persist cleaned cache

# build sections from cleaned cache
for num, theme in cache.items():
    if theme in sections:          # extra safety
        sections[theme].append(num)

# ---------------------------------------------------------------- print roadmap
for theme in order:
    issues = sections[theme]
    if issues:
        print(f"## {theme}")
        for n in sorted(issues):
            title, url = title_map.get(n, ("(missing)", f"https://github.com/{REPO}/issues/{n}"))
            print(f"- [#{n}]({url}) – {title}")
        print()