LegalBot is a Telegram bot for assisting users with legal claims in Russia. It collects the user's problem, builds a "golden" prompt for OpenRouter, generates advice along with PDF and DOCX documents, and sends them back via Telegram.
- Commands:
/start,/help,/claim,/status,/delete,/lang - Input text up to 8000 characters
- Rate limit: 10 requests per minute per user
- Generates PDF and DOCX versions of claim letters and lawsuits
- Data policy and
/deletecommand for removing history
.
├─ cmd/
│ ├─ bot/ # Telegram webhook service
│ ├─ prompt/ # Prompt builder gRPC service
│ └─ worker/ # Task consumer
├─ internal/
│ ├─ telegram/ # Telegram SDK wrapper
│ ├─ openrouter/ # REST client for OpenRouter
│ ├─ prompt/ # Prompt templates
│ └─ db/ # Postgres repositories
├─ deploy/
│ ├─ docker-compose.yml
│ ├─ Dockerfile.bot
│ ├─ Dockerfile.worker
│ └─ Dockerfile.prompt
├─ Makefile
├─ SPEC.md # Technical specification
cp .env.example .env
# set TELEGRAM_SECRET_TOKEN to the value configured in your bot settings
docker compose up --build
DOCS_BASE_URL can be used to customize the base URL for document links.
make lintUses golangci-lint with settings in .golangci.yml.
Automated deployment is handled by GitHub Actions. Secrets such as the server
IP address and SSH key are stored in HashiCorp Vault. The workflow reads these
values using hashicorp/vault-action@v2 and then performs an SSH deployment via
appleboy/ssh-action.
Configure Vault with SERVER_IP and SERVER_SSH_KEY keys under the
secret/data/legalbot path. Provide VAULT_ADDR, VAULT_ROLE_ID and
VAULT_SECRET_ID as GitHub repository secrets so the workflow can fetch the
credentials during the deploy job.