BottleCRM Backend - Django REST API

BottleCRM is a self-hosted CRM you run on your own infrastructure, MIT licensed and free to use. This package is the backend: a Django REST Framework API that serves both the SvelteKit web app and the Flutter mobile client from one set of endpoints.

pip install django-crm

What ships in it: leads, contacts, customer accounts, a sales pipeline with deal tracking, tasks, support tickets with solutions, approvals and escalation, invoices, estimates, recurring invoices, products, orders, business hours and holiday calendars for SLA timing, and saved-reply macros. Every app is listed below.

Multi-tenancy is enforced in the database, not just the ORM. Tenant isolation uses PostgreSQL row-level security keyed on the organization claim in the JWT, so a missing filter in application code cannot leak another tenant's rows. Setup and the non-negotiable rules are in RLS_SETUP.md.

Source, issues and releases: https://github.com/Django-CRM/Django-CRM

Tech Stack

Minimum versions, as declared in pyproject.toml. That file is the source of truth; the list here is a summary and uv.lock pins what actually gets installed.

• Django 6.0.7+ - Web framework
• Django REST Framework 3.17+ - API toolkit
• PostgreSQL - Database, via psycopg 3.2.10+ with the pool extra
• Celery 5.6+ - Async task queue
• Redis 8.0+ - Message broker for Celery
• djangorestframework-simplejwt 5.5+ - JWT authentication
• drf-spectacular 0.30+ - OpenAPI/Swagger documentation
• django-ses 4.7+ - AWS SES email backend
• WeasyPrint 69+ - Invoice and estimate PDF generation
• Sentry SDK 2.66+ - Error tracking

psycopg 3, not psycopg2. The pool extra is required rather than optional: Django raises ImproperlyConfigured for pool options under psycopg2, and DATABASES["default"]["OPTIONS"]["pool"] depends on it.

Django Apps

teams was merged into common. The emails, events, planner and boards apps were removed after 0.9.0; see the release notes if you are upgrading from that version.

Prerequisites

• Python 3.12+ (uv installs a matching Python automatically if needed)
• PostgreSQL
• Redis (for Celery)
• uv: Python package & venv manager (replaces pip + virtualenv)

Installation

1. Install uv

# Linux / macOS
curl -LsSf https://astral.sh/uv/install.sh | sh

# Or via Homebrew
brew install uv

2. Install Python dependencies

uv sync reads pyproject.toml + uv.lock, picks the Python version from .python-version, and creates .venv/ with everything installed.

cd backend
uv sync

Run any backend command with uv run <cmd> (e.g. uv run python manage.py migrate). uv resolves binaries from .venv/bin/ automatically, no manual source .venv/bin/activate needed (though that still works if you prefer it).

3. Install PDF generation system dependencies

Invoice PDF generation uses WeasyPrint (a runtime dep). It requires system libraries that must be installed separately:

Ubuntu/Debian:

sudo apt-get install -y \
    libpango-1.0-0 \
    libpangocairo-1.0-0 \
    libcairo2 \
    libgdk-pixbuf2.0-0 \
    libffi-dev \
    shared-mime-info

macOS:

brew install pango cairo libffi gdk-pixbuf

Fedora/CentOS:

sudo dnf install -y \
    pango \
    cairo \
    gdk-pixbuf2 \
    libffi-devel

Windows: Follow the WeasyPrint Windows installation guide.

Note: If you skip this step, the CRM will work but PDF download for invoices will show "PDF generation unavailable".

4. Configure environment variables

Create a .env file in the backend/ directory:

# Django
# At least 32 bytes: this key also signs the JWTs, and HS256 needs one that long.
# python -c "import secrets; print(secrets.token_urlsafe(48))"
SECRET_KEY=your-secret-key-here
ENV_TYPE=dev
DEBUG=True
ALLOWED_HOSTS=localhost,127.0.0.1

# Database
DBNAME=bottlecrm
DBUSER=postgres
DBPASSWORD=root
DBHOST=localhost
DBPORT=5432

# Email
DEFAULT_FROM_EMAIL=noreply@bottlecrm.com
ADMIN_EMAIL=admin@bottlecrm.com

# Celery
CELERY_BROKER_URL=redis://localhost:6379/0
CELERY_RESULT_BACKEND=redis://localhost:6379/0

# The web app, NOT this API. Every link the backend emails is built from it:
# the magic-link sign-in URL, the customer invoice and estimate portals, the
# CSAT survey. Point it at the API host and every one of those links 404s.
FRONTEND_URL=http://localhost:5173

# Google sign-in. Without these the OAuth login flow cannot complete, which is
# the only interactive way into the app.
GOOGLE_CLIENT_ID=your-google-client-id
GOOGLE_CLIENT_SECRET=your-google-client-secret

# Where the browser calls this API from. The SvelteKit dev server by default.
CORS_ALLOWED_ORIGINS=http://localhost:5173

# This API's own public origin
DOMAIN_NAME=http://localhost:8000

5. Set up database

# Create PostgreSQL database
sudo -u postgres psql
CREATE DATABASE bottlecrm WITH OWNER = postgres;
ALTER USER postgres WITH PASSWORD 'root';
\q

# Run migrations
uv run python manage.py migrate

# Create superuser (optional)
uv run python manage.py createsuperuser

6. Run the development server

uv run python manage.py runserver

The API will be available at http://localhost:8000

Running Celery

For background tasks (emails, notifications), run the Celery worker:

uv run celery -A crm worker --loglevel=INFO

API Documentation

• Swagger UI: http://localhost:8000/swagger-ui/
• ReDoc: http://localhost:8000/api/schema/redoc/
• Django Admin: http://localhost:8000/admin/

Generating Schema

To generate the OpenAPI schema file:

uv run python manage.py spectacular --file openapi.yml

Architecture

Multi-Tenancy

Every request operates within an organization context:

• Organization (Org): Top-level tenant container
• Users: Regular members with USER role
• Admins: Organization administrators with ADMIN role
• Super Admin: Users with is_superuser set on the user record have platform-wide access. Grant it deliberately (manage.py createsuperuser or the Django admin). It is never inferred from the email address

Authentication

JWT-based authentication:

Authorization: Bearer <token>

• Organization ID is embedded in the JWT token (not sent as header)
• Access token lifetime: 1 hour
• Refresh token lifetime: 14 days
• Refresh tokens are single-use: /api/auth/refresh-token/ blacklists the token you send and returns a replacement, so clients must persist the new refresh value from every response

Middleware

The middleware chain provides security:

1. GetProfileAndOrg (common.middleware.get_company):

   • Extracts org_id from JWT token claims (not headers - prevents spoofing)
   • Validates user has active membership in the organization
   • Sets request.profile and request.org

2. RequireOrgContext (common.middleware.rls_context):

   • Sets PostgreSQL session variable app.current_org for RLS
   • Resets context after each request

Row-Level Security (RLS)

PostgreSQL RLS provides database-level tenant isolation as defense-in-depth.

How It Works

1. Middleware sets context: SET app.current_org = '<org_id>'
2. RLS policies filter queries: Only rows matching org_id are visible
3. Fail-safe design: Empty context returns zero rows (NULLIF pattern)

Protected Tables

ORG_SCOPED_TABLES in common/rls/__init__.py is the list, and the only one worth trusting. It is deliberately not reproduced here: this section used to carry a copy and a count, and both went stale, naming tables that no longer exist and a total that was wrong by more than double.

uv run python manage.py manage_rls --status   # what is actually protected

Configuration

RLS is configured in common/rls/__init__.py:

from common.rls import RLS_CONFIG, get_enable_policy_sql

# List of protected tables
tables = RLS_CONFIG["tables"]

# Enable RLS on a table
cursor.execute(get_enable_policy_sql("my_table"))

Management Commands

# Check RLS status on all tables
uv run python manage.py manage_rls --status

# Verify database user is non-superuser (required for RLS)
uv run python manage.py manage_rls --verify-user

# Test RLS isolation between organizations
uv run python manage.py manage_rls --test

Policies are enabled by migration, using get_enable_policy_sql(), so there is no --enable flag to run by hand and no --disable to reach for when something is in the way. A table becomes protected when its migration says so.

Critical: Database User Setup

PostgreSQL superusers bypass ALL RLS policies. You must use a non-superuser:

-- Create application user
CREATE USER crm_app WITH PASSWORD 'your_secure_password';

-- Grant permissions
GRANT CONNECT ON DATABASE bottlecrm TO crm_app;
GRANT USAGE ON SCHEMA public TO crm_app;
GRANT SELECT, INSERT, UPDATE, DELETE ON ALL TABLES IN SCHEMA public TO crm_app;
GRANT USAGE, SELECT ON ALL SEQUENCES IN SCHEMA public TO crm_app;

-- Future tables inherit permissions
ALTER DEFAULT PRIVILEGES IN SCHEMA public
GRANT SELECT, INSERT, UPDATE, DELETE ON TABLES TO crm_app;

Update .env:

DBUSER=crm_app
DBPASSWORD=your_secure_password

Celery Tasks & RLS

Background tasks don't go through middleware, so set RLS context manually:

from common.tasks import set_rls_context


@app.task
def my_background_task(data_id, org_id):
    set_rls_context(org_id)  # Required!
    obj = MyModel.objects.get(id=data_id)
    # ... process

Adding RLS to New Tables

1. Add table name to ORG_SCOPED_TABLES in common/rls/__init__.py
2. Create migration using get_enable_policy_sql()
3. Ensure model has org = models.ForeignKey(Org, ...)

BaseModel Pattern

All models inherit from BaseModel (common.base.BaseModel):

• UUID primary keys (not integer IDs)
• Automatic timestamps: created_at, updated_at
• Audit trail: created_by, updated_by
• Organization isolation: org = models.ForeignKey(Org)

API Endpoint Pattern

GET/POST       /api/<module>/              # List/Create
GET/PUT/DELETE /api/<module>/<pk>/         # Detail/Update/Delete
GET/POST       /api/<module>/comment/<pk>/ # Comments
GET/POST       /api/<module>/attachment/<pk>/ # Attachments

Project Structure

backend/
├── manage.py
├── pyproject.toml          # Python deps + project metadata (uv-managed)
├── uv.lock                 # Pinned, reproducible dependency tree
├── .python-version         # Python version pin (uv reads this on `uv sync`)
├── crm/                    # Django project settings
│   ├── settings.py
│   ├── urls.py
│   ├── celery.py
│   └── wsgi.py
├── common/                 # Core models and utilities
│   ├── models.py           # User, Org, Profile, etc.
│   ├── base.py             # BaseModel
│   ├── middleware/
│   └── tasks.py            # Celery tasks
│   └── templates/          # Email templates, shipped as package data
├── accounts/
├── leads/
├── contacts/
├── opportunity/
├── cases/
├── tasks/
├── invoices/
├── orders/
├── business_hours/
├── macros/
└── static/

Templates live in their owning app's templates/ directory rather than a project-level one. TEMPLATES[0]["DIRS"] is empty on purpose: a BASE_DIR entry resolves to site-packages once the package is installed, where nothing is written, and the login emails would not render.

Development

Code Quality

# Lint (E, F and I as backend/ruff.toml selects them)
uv run ruff check .

# Format the tree. `ruff format` is black's output from one tool, and ruff's
# I rules do isort's job, so neither black nor isort is installed here.
uv run ruff format .

# Run tests
uv run pytest

CI runs ruff check . and ruff format --check . as hard steps, so both must pass.

Managing Dependencies

# Add a runtime dependency (updates pyproject.toml + uv.lock)
uv add <package>

# Add a dev-only dependency (e.g. test or lint tool)
uv add --group dev <package>

# Remove a dependency
uv remove <package>

# Refresh the lockfile
uv lock --upgrade

Creating a New App

1. Create the app:

   uv run python manage.py startapp myapp
   

2. Add to INSTALLED_APPS in crm/settings.py

3. Create models inheriting from BaseModel:

   from common.base import BaseModel
   from common.models import Org
   
   
   class MyModel(BaseModel):
       org = models.ForeignKey(Org, on_delete=models.CASCADE)
       # ... other fields
   

4. Always filter queries by organization:

   queryset = MyModel.objects.filter(org=request.profile.org)
   

5. Run migrations:

   uv run python manage.py makemigrations
   uv run python manage.py migrate
   

Environment Variables Reference

DB_POOL_ENABLED is not just a performance knob. RLS context lives in a session-scoped variable, so a pooled connection carries the previous tenant's org id unless something clears it. The reset callback in common/rls/pool.py is what clears it, and pooling must never be enabled without it.

Troubleshooting

Database Connection Issues

# Check PostgreSQL is running
sudo systemctl status postgresql

# Verify database exists
sudo -u postgres psql -l

Migration Issues

# Show migration status
uv run python manage.py showmigrations

# Reset migrations (development only)
uv run python manage.py migrate --fake <app> zero

Celery Not Processing Tasks

# Check Redis is running
redis-cli ping

# Check Celery worker logs
uv run celery -A crm worker --loglevel=DEBUG

License

MIT License. See LICENSE.

The link is absolute on purpose. This file is the package's long description on PyPI, where a relative ../LICENSE points outside the distribution and 404s.