# Model Radar

[![CI](https://github.com/oiupoyt/model-radar/actions/workflows/ci.yml/badge.svg)](https://github.com/oiupoyt/model-radar/actions)
[![License: MIT](https://img.shields.io/badge/License-MIT-blue.svg)](LICENSE)

A lightweight, crash-resilient Discord bot built in Rust with [poise](https://github.com/serenity-rs/poise) and [serenity](https://github.com/serenity-rs/serenity). Model Radar monitors [OpenRouter](https://openrouter.ai)'s live model catalog for free LLMs (`:free` and exact $0 pricing) and dispatches real-time catalog release alerts and live auto-updating dashboards directly to Discord servers.

A live instance is running 24/7. You can see the bot in action and test commands here:  
**[Join the Live Discord Server](https://discord.gg/dJ5MhVCGh)**

---

## Features

- **Automated OpenRouter Monitoring:** Periodically polls the public OpenRouter models API (`/api/v1/models`) with configurable intervals.
- **Accurate Zero-Price Detection:** String-based decimal parsing identifies genuine free-tier models (including scientific notation like `0e-9999`) while rejecting precision artifacts, non-zero pricing, and tiered overrides.
- **Live Auto-Updating Dashboard:** Running `/watch` posts a persistent public dashboard embed in your channel that automatically refreshes every poll interval (`POLL_SECONDS`), featuring new drops at the top and active stealth/experimental models directly on the first page.
- **Interactive Component Buttons:** Multi-page `/free-models` and `/status` views include interactive `[◀ Prev]` and `[Next ▶]` buttons with automatic timeout disabling.
- **Model ID Autocomplete:** Live slash command autocomplete for `/model <id>` to instantly find any model in the catalog as you type.
- **Direct OpenRouter Links:** Embed titles link directly to the model's page on OpenRouter for quick browser inspection.
- **Pricing per 1M Tokens:** Model inspection computes readable rates per 1 Million tokens (e.g. `$0.15 / 1M tokens`).
- **Flexible Subscription Filters:**
  - `free_only`: Target free models or watch all new catalog entries.
  - `search`: Filter releases by model identifier or name (e.g., `qwen`, `llama`, `coder`).
  - `min_context`: Ignore models with context windows below a given token threshold (e.g. `32,768`).
  - `compact`: Dense notification rows for high-volume channels.
- **Opt-In Role Pings:** Server-level ping configuration supporting custom roles as well as `@everyone`.
- **Crash-Resilient State:** Subscriptions and catalog states are stored atomically (`.tmp` write followed by atomic filesystem rename) to guarantee state integrity across crashes and container reboots.
- **Queue Bounds:** Unreachable alert channels (e.g., deleted channel or revoked permissions) are capped at 50 queued events, preventing unbounded state growth.
- **Discord API Safety:** Strict UTF-16 character limit enforcement and markdown character escaping prevent broken embeds or Discord 400 Bad Request rejections.
- **Dry-Run Diagnostic Mode:** Test public catalog availability and parsing without needing a Discord token (`--check`).

---

## Slash Commands

| Command | Scope | Description |
| :--- | :--- | :--- |
| `/free-models [search] [page]` | Public / Ephemeral | Browse all currently available free models in OpenRouter's catalog with interactive `[◀ Prev]` / `[Next ▶]` buttons and search filter. |
| `/status [free_only] [search] [min_context] [page] [page_size] [compact] [public]` | Configurable | View a catalog freshness snapshot with model counts, interactive pagination, and token constraints. |
| `/model <id> [compact] [show_pricing] [public]` | Configurable | Inspect detailed specifications with live model ID autocomplete, cost per 1M tokens, and direct clickable links to OpenRouter. |
| `/watch [ping_role] [free_only] [search] [min_context] [ping_enabled] [compact] [clear_role]` | Manage Server | Posts a public live auto-updating dashboard in the channel and configures drop alerts. Supports any role or `@everyone`. |
| `/unwatch` | Manage Server | Remove the server's subscription and delete all queued pending alerts. |
| `/stats` | Ephemeral | View catalog freshness, last successful fetch timestamp, polling interval, and complete server subscription settings. |
| `/testping` | Manage Server | Dispatch a labeled test notification to verify channel permissions and role mentionability without modifying catalog state. |
| `/about` | Ephemeral | Display bot information, command guide, poll interval, and optional operator support link. |

---

## Getting Started

### 1. Create a Discord Bot Application

1. Open the [Discord Developer Portal](https://discord.com/developers/applications) and create a **New Application**.
2. Navigate to the **Bot** tab:
   - Click **Reset Token** to generate a bot token for your `DISCORD_TOKEN` environment variable.
   - *Note:* No Privileged Gateway Intents are required (Model Radar does not require Message Content, Server Members, or Presence intents).
3. Navigate to **OAuth2 → URL Generator**:
   - **Scopes:** Select `bot` and `applications.commands`.
   - **Bot Permissions:** Select `View Channels`, `Send Messages`, and `Embed Links`.
4. Open the generated invite URL in your browser to invite the bot to your server.

---

### 2. Configuration & Environment Variables

| Variable | Required | Default | Description |
| :--- | :---: | :---: | :--- |
| `DISCORD_TOKEN` | Yes | — | Discord Bot Token from the Developer Portal. |
| `POLL_SECONDS` | No | `600` | Catalog polling interval in seconds (allowed: `60` to `86400`). |
| `STATE_PATH` | No | `state.json` | Path where subscription settings and catalog state are stored. |
| `SUPPORT_URL` | No | — | Optional HTTPS URL (e.g. Ko-fi, GitHub Sponsors) shown in `/about`. |
| `DISCORD_GUILD_ID` | No | — | Guild ID for instant slash command registration during testing (global registration takes up to an hour). |

---

### 3. Deployment Options

#### Option A: Docker Compose (Recommended)

1. Clone the repository:
   ```bash
   git clone https://github.com/oiupoyt/model-radar.git
   cd model-radar
   ```

2. Create `.env` from example:
   ```bash
   cp .env.example .env
   ```
   Edit `.env` and set `DISCORD_TOKEN`.

3. Launch the container:
   ```bash
   docker compose up -d
   ```
   State persists in `./data`.

#### Option B: Native Cargo Run

Prerequisites: [Rust 1.85+ (Edition 2024)](https://rustup.rs/)

```bash
# Verify catalog connection and parsing (no token required)
cargo run --locked -- --check

# Run the bot
export DISCORD_TOKEN="your_bot_token"
export POLL_SECONDS=600
cargo run --release --locked
```

---

## Setting Up Alerts in Discord

1. Pick a role for notifications (e.g. `@ModelAlerts`, `@everyone`, or any custom role).
2. In the channel where alerts should be posted, run:
   ```text
   /watch ping_role:@ModelAlerts ping_enabled:true
   ```
   *(Or `/watch ping_role:@everyone ping_enabled:true` to notify everyone).*
3. Verify the setup by running:
   ```text
   /testping
   ```
4. Check your active server configuration at any time with `/stats`.

---

## Development & Testing

```bash
# Run unit and integration tests
cargo test --locked

# Run linter
cargo clippy --all-targets -- -D warnings

# Check code formatting
cargo fmt --check
```

---

## License

This project is licensed under the [MIT License](LICENSE).