subreddit-announcer

Good project. Before I scaffold anything, let me nail down a couple of forks that change the whole stack — everything else I can pick sensible defaults for.

• AskUserQuestion questions 36.1s 271 B

Locked in: **Python + discord.py (full bot) + asyncpraw, long-running poll loop.**

Let me preflight the environment (templates, tea login, git creds) before scaffolding.

• Bash echo "=== templates ===" && ls -la ~/claude/_templates/project/ 2>&1; echo "=== existing projects ===" && ls ~/claude/pr… 1.8s 1.3 KB

Environment's ready — `uv` is available, tea login works, git creds present. There's already a `quotebot` project; let me peek at it for conventions, and read the templates.

• Bash echo "=== project CLAUDE.md template ===" && cat ~/claude/_templates/project/CLAUDE.md; echo; echo "=== memory.md templa… 1.9s 3.2 KB

• Read /home/megaproxy/claude/projects/quotebot/bot.py 20ms 15.0 KB

There's an established `quotebot` pattern (discord.py + sqlite + `config.env`). Let me read it so the new bot matches conventions.

This is a fresh project, so I'll follow the documented new-project flow (template → git → Forgejo → push). I'll model it on `quotebot`'s conventions (`config.env` + `config.example.env`, SQLite for state, discord.py), adding `asyncpraw` for Reddit.

Here's the design I'm building:

- **`bot.py`** — discord.py bot with an `asyncpraw` read-only Reddit client and a background poll loop (`tasks.loop`) hitting each watched subreddit's `/new`.
- **State in SQLite** — `watches` (subreddit → channel mapping) + `seen_posts` (dedup). When you first add a sub it's *bootstrapped silently* so you don't get spammed with the existing backlog; only genuinely new posts announce.
- **Slash commands** — `/watch`, `/unwatch`, `/watching`, `/setinterval`.
- **Config** — `config.env` (gitignored) for `DISCORD_TOKEN` + Reddit app creds.

Let me scaffold.

• Bash set -e cd ~/claude/projects NAME=subreddit-announcer if [ -e "$NAME" ]; then echo "ALREADY EXISTS — stopping"; exit 1; f… 1.5s 361 B

Now let me write the project files. Starting with the bot itself.

• Write /home/megaproxy/claude/projects/subreddit-announcer/bot.py 89ms 156 B

Let me fix the timestamp construction in the embed — that chained `snowflake_time` call is convoluted. I'll use `datetime` directly.

• Edit /home/megaproxy/claude/projects/subreddit-announcer/bot.py 15ms 166 B

• Edit /home/megaproxy/claude/projects/subreddit-announcer/bot.py 10ms 166 B

Now the supporting files — config example, pyproject, README, and gitignore tweak:

• Write /home/megaproxy/claude/projects/subreddit-announcer/config.example.env 16ms 168 B

• Write /home/megaproxy/claude/projects/subreddit-announcer/pyproject.toml 14ms 164 B

• Write /home/megaproxy/claude/projects/subreddit-announcer/README.md 19ms 159 B

The template `.gitignore` covers `.venv/` but not `config.env`/`announcer.db` (they don't match `.env`). Let me append project-specific entries, then verify deps install and the bot compiles.

• Bash cd ~/claude/projects/subreddit-announcer cat >> .gitignore <<'EOF' # Project-specific: local secrets and runtime state c… 7.8s 373 B

Deps installed. Now compile-check the bot:

• Bash cd ~/claude/projects/subreddit-announcer uv run python -m py_compile bot.py && echo "OK: bot.py compiles" || echo "COMPI… 2.4s 19 B

Let me do a deeper import check (py_compile only catches syntax, not bad imports/attribute references at module load):

• Bash cd ~/claude/projects/subreddit-announcer uv run python -c " import bot # Exercise the pure helpers and class constructio… 455ms 43 B

Everything resolves and the state layer behaves. Now fill in the project docs (CLAUDE.md, memory.md) and commit.

• Write error /home/megaproxy/claude/projects/subreddit-announcer/CLAUDE.md 1ms 96 B

• Read /home/megaproxy/claude/projects/subreddit-announcer/CLAUDE.md 6ms 959 B

• Write /home/megaproxy/claude/projects/subreddit-announcer/CLAUDE.md 38ms 169 B

• Write error /home/megaproxy/claude/projects/subreddit-announcer/memory.md 1ms 96 B

• Read /home/megaproxy/claude/projects/subreddit-announcer/memory.md 19ms 29 B

• Write /home/megaproxy/claude/projects/subreddit-announcer/memory.md 38ms 169 B

Now git init, verify no secrets are staged, commit, create the Forgejo repo, and push.

• Bash cd ~/claude/projects/subreddit-announcer git init -b main >/dev/null 2>&1 git add -A echo "=== staged files (verify no c… 3.7s 264 B

Clean staging — `uv.lock` is in (good for reproducibility), no secrets/state. Commit, create the private Forgejo repo, wire the remote, and push:

• Bash cd ~/claude/projects/subreddit-announcer git commit -q -m "Initial scaffold: Discord subreddit-announcer bot" && echo "c… 14.1s 582 B

• Bash cd ~/claude/projects/subreddit-announcer git remote add origin https://git.rdx4.com/megaproxy/subreddit-announcer.git gi… 5.8s 136 B

Done. The bot is built, verified, and pushed to Forgejo.

## What's there

**`projects/subreddit-announcer/`** → [git.rdx4.com/megaproxy/subreddit-announcer](https://git.rdx4.com/megaproxy/subreddit-announcer) (private)

- **`bot.py`** — discord.py bot with an async poll loop (`tasks.loop`) that sweeps each watched subreddit's `/new` every 60s (configurable) and posts a reddit-orange embed per new submission.
- **Slash commands:** `/watch <subreddit> [channel]`, `/unwatch`, `/watching`, `/setinterval`.
- **SQLite state** (`announcer.db`): `watches` (subreddit→channel per server) + `seen_posts` (dedup, pruned to 500/sub).
- **Silent bootstrap** — when you first `/watch` a sub, its existing backlog is marked seen *without* announcing; only genuinely new posts fire. No spam when you add a busy subreddit.
- No privileged intents (post-only), so the invite just needs `bot` + `applications.commands` + Send Messages/Embed Links.

**Verified:** `uv sync` installs clean, `py_compile` passes, module imports + the `Store`/parsing helpers smoke-tested in-memory.

## What I couldn't do (needs your tokens)

I can't run it end-to-end since that requires live Discord + Reddit credentials. Your steps:

1. **Discord app** → [developers portal](https://discord.com/developers/applications): create app, copy the bot token, invite it to your server.
2. **Reddit app** → [reddit.com/prefs/apps](https://www.reddit.com/prefs/apps): "create another app" → type **script**, grab client id + secret.
3. Fill creds and run:
   ```bash
   cd ~/claude/projects/subreddit-announcer
   cp config.example.env config.env   # then edit in your tokens
   set -a; source config.env; set +a
   uv run bot.py
   ```
   Then in Discord: `/watch python #some-channel`.

If you want to do the login/run step right here so I can watch the output, type it with the `!` prefix, e.g. `! cd ~/claude/projects/subreddit-announcer && set -a && source config.env && set +a && uv run bot.py`.

## Worth adding next (noted in `memory.md` TODOs)

- **Per-watch filtering** by flair or title keyword — right now *every* new post announces, which gets noisy on big subs.
- A **systemd unit** to keep it alive (sketch is in the README).

Want me to add keyword/flair filtering to `/watch` now, or leave it until you've got it running?

The user is explicitly asking, which satisfies the "explicit confirmation" requirement for going public. Let me flip visibility via the Forgejo API using the stored token.

• Bash TOKEN=$(grep 'git.rdx4.com' ~/.git-credentials | sed -E 's#https://[^:]+:([^@]+)@.*#\1#') echo "=== current visibility =… 22.1s 69 B

Let me confirm with a clean read-back:

• Bash TOKEN=$(grep 'git.rdx4.com' ~/.git-credentials | sed -E 's#https://[^:]+:([^@]+)@.*#\1#') curl -s -H "Authorization: tok… 2.6s 70 B

Done — **subreddit-announcer is now public**: https://git.rdx4.com/megaproxy/subreddit-announcer

`private: False` confirmed. Anyone can now view/clone it. One thing worth a sanity check given it's public: the repo contains **no secrets** — `config.env` is gitignored and was never committed (I verified that before the initial push), and `config.example.env` only has empty placeholders. So nothing sensitive is exposed.