rubot/README.md
reverse e801b04d6e
All checks were successful
ci / check (push) Successful in 47s
karma: reasons, milestones, no-highlight replies
2026-07-29 22:29:01 +00:00

197 lines
8 KiB
Markdown

# rustbot
A small, modular IRC bot in Rust — **almost dependency-free** (only `native-tls`
for the TLS transport), with a hand-rolled IRC protocol layer, HTTP client, JSON
parser, and hashes, built to the [modern IRC client spec](https://modern.ircdocs.horse/).
## Structure
rustbot is a **library crate** (`src/lib.rs`) with a thin binary on top:
| Path | Layer | Responsibility |
|-----------------|----------------|-------------------------------------------------------------------|
| `src/irc/` | protocol + I/O | `message` (parse IRCv3 wire format), `connection` (TCP/TLS transport), `encoding` (base64, server-time) |
| `src/bot/` | behavior | `mod` (state + loop), `caps` (IRCv3 + SASL), `commands` (routing + help), `module` (plugin trait), `modules/` (feature modules) |
| `src/config.rs` | configuration | Layered `Config`; multi-network `[section]` parsing |
| `src/http.rs` | utility | Minimal HTTPS GET over `native-tls`, with retry |
| `src/json.rs` | utility | Recursive-descent JSON parser |
| `src/lib.rs` | crate root | Ties the modules into the reusable `rustbot` library |
| `src/main.rs` | binary | Thin per-network connect/reconnect supervisor |
| `tests/` | tests | Integration tests against the public API, one file per subsystem |
The layers are deliberately separable: `irc` knows nothing about the bot, `bot`
knows nothing about how the process is launched, and `main` only supervises.
Adding a feature is a new **module** — see [Modules](#modules).
## Modules
Features are **modules**: self-contained units behind the `Module` trait
(`src/bot/module.rs`). A module declares the commands it provides and turns each
invocation into `Action`s — it never touches the socket, so it stays pure and
unit-testable:
```rust
pub trait Module {
fn name(&self) -> &'static str;
fn commands(&self) -> &'static [CommandSpec];
fn on_command(&mut self, cmd: &Command) -> Vec<Action>;
}
```
At startup the bot builds a `command → module` route table and auto-generates
`help` from every module's `CommandSpec`s. Each network gets its own fresh set
of modules, so stateful ones (like the dice roller's PRNG or the weather cache)
keep per-network state with no locking.
**To add a module**, create `src/bot/modules/<name>.rs` implementing `Module`,
then add one line to `all()` in `src/bot/modules/mod.rs`:
```rust
pub fn all(network: &str) -> Vec<Box<dyn Module>> {
vec![
Box::new(builtins::Builtins),
Box::new(dice::Dice::new()),
Box::new(myfeature::MyFeature::new()), // <- your module
]
}
```
Web modules fetch with the shared `http::get` and parse with `json::Json`; each
has a pure formatter unit-tested against a canned API body (no network needed).
## Commands
Triggered by the configured prefix (default `!`, set with `prefix`); `help`
lists them at runtime and `help <command>` describes one.
| Command | Does |
|---|---|
| `ping` / `echo <text>` / `hello` | basics |
| `roll [NdM]` | roll dice (default `1d6`) |
| `karma [thing]` / `karmabottom` (bump with `thing++ [reason]` / `thing--`) | karma + rank + last reason, or top/bottom; milestones announced |
| `md5` / `sha1` / `sha256` / `hash <text>` | hex digests |
| `add <location>` then `w [location]` | weather (wttr.in), per-user location |
| `crypto <sym>` | coin price in USD (Bitstamp) |
| `wiki <term>` | Wikipedia summary |
| `define <word>` / `urban <term>` | dictionary / Urban Dictionary |
| `tinyurl <url>` | shorten a URL |
| `down <host>` | is a site reachable? |
## Build & run
```sh
cargo build # or: cargo build --release
cargo test # integration tests in tests/
cargo run
```
## Configuration
Settings are layered, each overriding the previous:
1. **Built-in defaults** (`Config::default()` in `src/config.rs`)
2. **A config file**`key = value`, see [`rustbot.conf`](rustbot.conf)
3. **Environment variables** — handy for one-off overrides
The config file is found via: the **first CLI argument**, else `RUSTBOT_CONFIG`,
else `./rustbot.conf` if present.
```sh
cargo run # uses ./rustbot.conf
cargo run -- /etc/rustbot.conf # explicit path
RUSTBOT_CONFIG=/etc/rustbot.conf cargo run
```
Config file keys: `server`, `port`, `tls`, `nick`, `user`, `realname`,
`channels` (comma/space-separated), `prefix`, `verbose` (wire logging, default
on), `password`, `sasl_user`, `sasl_pass`, and `name` (a network's log label).
Whole-line `#`/`;` comments only — inline comments would clash with channel `#`s.
Web modules write per-network state under `RUSTBOT_DATA_DIR` (default `.`).
### Multiple networks
Group keys under `[name]` section headers to connect to several networks at
once — one supervisor thread each, with independent reconnect/backoff. Keys
*before* the first header are shared defaults every network inherits; each
section overrides them. A file with no headers is a single network (the
historical behaviour), and its logs stay unprefixed.
```ini
# shared defaults, inherited by every network below
nick = rubot
realname = rubot
[tchatou]
server = irc.tchatou.fr
tls = true
channels = #devs
[libera]
server = irc.libera.chat
tls = true
channels = #rust, #rust-beginners
nick = rubot_ # override just for this network
```
With more than one network, each network's log lines are tagged with its
section name (`[libera] << …`) so the interleaved output stays readable.
Environment overrides: `IRC_SERVER`, `IRC_PORT`, `IRC_TLS`, `IRC_NICK`,
`IRC_REALNAME`, `IRC_CHANNELS`, `IRC_PREFIX`, `IRC_VERBOSE`, `IRC_PASSWORD`,
`IRC_SASL_USER`, `IRC_SASL_PASS`.
```sh
IRC_NICK=mybot IRC_CHANNELS='#test' cargo run # override the file for one run
```
> If you put a real `password` in `rustbot.conf`, add it to `.gitignore` so the
> secret isn't committed.
## Deployment (systemd)
Runs as a system service (`User=debian`), auto-restarts on crash, and starts on
boot. The unit is kept in the repo as [`rustbot.service`](rustbot.service).
```sh
cargo build --release
sudo cp rustbot.service /etc/systemd/system/rustbot.service
sudo systemctl daemon-reload
sudo systemctl enable --now rustbot # start now + on boot
systemctl status rustbot # health
journalctl -u rustbot -f # live logs (<< in, >> out; verbose=false to quiet)
sudo systemctl restart rustbot # after a rebuild
```
One-off overrides without touching `rustbot.conf`: put `IRC_*` lines in
`/etc/default/rustbot` (env wins over the config file), then restart.
## IRCv3
The bot negotiates capabilities (`CAP LS 302``CAP REQ``CAP END`) and
enables everything useful the server offers: `message-tags`, `server-time`,
`batch`, `echo-message`, `account-tag`, `extended-join`, `multi-prefix`,
`away-notify`, `chghost`, `setname`, `userhost-in-names`, `cap-notify`,
`labeled-response`.
Two of these do real work:
- **`server-time` + `batch`** let the bot recognise **replayed history**. On
join, InspIRCd replays recent channel lines (wrapped in a `chathistory`
batch). Without this the bot re-answered old commands on every reconnect;
now such messages are ignored (see `Bot::is_historical`).
- **`sasl`** (PLAIN) authenticates to services during negotiation. Set
`sasl_user`/`sasl_pass` in the config (or `IRC_SASL_USER`/`IRC_SASL_PASS`).
base64 is hand-rolled, so this stays dependency-free.
## TLS
Set `tls = true` and the port defaults to 6697. The transport wraps the socket
with `native-tls`, which verifies certificates against the **system OpenSSL
trust store** — so TLS security fixes arrive via OS updates rather than a
vendored crypto crate. Plaintext still works with `tls = false` / `port = 6667`.
Plain TCP and TLS share one code path: `Connection` holds a single
`Box<dyn Read + Write>`, since the single-threaded event loop never needs the
old read/write socket split (which TLS can't do anyway).