Scriber setup guide

Everything to take Scriber from zero to recording: create the Discord app, get its token and IDs, invite it with the right permissions, admit it to a private voice channel, then configure and run the container.

Roughly 10–15 minutes. You'll need a Discord account and a server where you have the Manage Server permission.

Before you start

You will need:

• Docker or Podman (recommended), or Python 3.12+ with the libopus system library — libopus is required for Discord voice, config alone won't make voice work.
• A Discord server you manage (you need Manage Server to add a bot).
• An API key for Anthropic or OpenAI — or a self-hosted OpenAI-compatible server (Ollama, vLLM, LM Studio, LocalAI) if you want summaries to never leave your machine.

1 · Create the Discord application

2 · Configure the bot & get the token

3 · Find the IDs you need

Application ID (a.k.a. Client ID)

On the General Information page, copy Application ID. (The identical value appears on the OAuth2 tab as Client ID.) You'll use it to build the invite link.

Server ID and Channel IDs (optional but recommended)

To make slash commands appear instantly in one server (instead of waiting up to an hour for a global sync), set DISCORD_GUILD_ID. To get it, enable Developer Mode in the Discord app:

4 · Invite the bot to your server

The bot must be added with two scopes — bot (to join and record) and applications.commands (so /scriber registers). The four permissions Scriber advertises add up to the integer 1084416:

Easiest: use the ready-made invite URL

Replace YOUR_APPLICATION_ID with the ID from step 3 and open the URL in a browser:

https://discord.com/oauth2/authorize?client_id=YOUR_APPLICATION_ID&scope=bot%20applications.commands&permissions=1084416

Pick a server where you have Manage Server, click Authorize, and solve the captcha. The bot joins.

Or build it in the portal

Two equivalent paths — use one consistently:

• OAuth2 → URL Generator: tick the bot and applications.commands scopes, then in the permissions grid tick View Channel, Send Messages, Attach Files, Connect. Copy the generated URL.
• Installation tab (newer UI): enable Guild Install, set the Install Link to Discord Provided Link, and under Default Install Settings for Guild Install add the same scopes and permissions.

5 · Give the bot access to a private voice channel

A private channel denies View Channel to @everyone; only roles or members with an explicit allow can see or join it. Server-wide permissions are not enough — you must add a channel-level override for the bot.

6 · Configure Scriber

Copy the example environment file and edit it. The only things you must set are DISCORD_TOKEN and one summary-provider block.

cp .env.example .env
$EDITOR .env

Discord

DISCORD_TOKEN=your-bot-token-from-step-2
# Optional: instant slash-command sync in one server (see step 3)
DISCORD_GUILD_ID=your-server-id

Summary providers (ordered failover list)

Each provider is a numbered block (_1, _2, …), tried in order until one succeeds. A single block is fine. Secrets can later be edited from the dashboard Settings page.

# Provider 1: Anthropic (Claude) — tried first
SUMMARY_PROVIDER_1=anthropic
SUMMARY_API_KEY_1=sk-ant-xxxxxxxx
SUMMARY_MODEL_1=claude-opus-5
SUMMARY_BASE_URL_1=https://api.anthropic.com

# Provider 2 (optional): OpenAI — tried if provider 1 fails
SUMMARY_PROVIDER_2=openai
SUMMARY_API_KEY_2=sk-xxxxxxxx
SUMMARY_MODEL_2=gpt-4o
SUMMARY_BASE_URL_2=https://api.openai.com/v1

Admin dashboard & the rest

Set the dashboard login and change the default password. Whisper, web bind, and storage keys all have sensible defaults — see the full reference in the README.

ADMIN_USERNAME=admin
ADMIN_PASSWORD=change-me   # change before exposing the port!
WHISPER_MODEL=base       # tiny / base / small / medium / large-v3
WHISPER_LANGUAGE=en       # default; override per meeting with lang:

7 · Run it

With Docker Compose (a compose.yaml is included):

docker compose up -d          # or: podman-compose up -d

Or with a plain container run:

docker build -t scriber .
docker run -d --name scriber \
  --env-file .env \
  -v ./data:/data \
  -v ./.env:/app/.env \
  -p 8080:8080 \
  scriber

Open http://localhost:8080 and sign in with your ADMIN_USERNAME / ADMIN_PASSWORD. The dashboard comes up even if the bot has a problem, and shows exactly what to fix.

8 · Your first recording

9 · Troubleshooting

/scriber doesn't appear

The bot was likely invited without applications.commands, so command registration returns 403 Missing Access. Re-invite with both scopes (step 4) — no need to kick it first — then re-run the sync from the dashboard (Resync commands) or restart. Guild commands appear instantly; a global sync can take up to an hour, so set DISCORD_GUILD_ID for testing.

The bot won't join the voice channel

Check it has View Channel and Connect on that specific channel (step 5) — a channel override can deny Connect even when the server role allows it. Confirm you're in a voice channel of the same server. Scriber's start command reports the missing permission by name.

Empty transcripts

Make sure libopus is installed (it's built into the Docker image). On servers that require voice end-to-end encryption (DAVE), Scriber decrypts it automatically; if it can't, recordings may come out empty — check the container logs.

Summary failed

The transcript is never lost — Scriber attaches the raw transcript to the error message and keeps it in the dashboard. Verify your SUMMARY_API_KEY, SUMMARY_MODEL, and SUMMARY_BASE_URL on the Settings page. Add a second provider for automatic failover.

nginx returns 502

Keep WEB_HOST=0.0.0.0 in .env (a loopback bind inside the container makes a published port unreachable) and keep the container side of -p at 8080. Full reverse-proxy notes are in the README.