django-auto-agent

Django reusable app that turns opted-in DRF endpoints into a LangGraph supervisor, with optional long-term memory, admin views, and evals.

Import name: ai_agent.

Install

pip install django-auto-agent

Redis-backed memory store (optional):

pip install "django-auto-agent[redis]"

From a clone of this repository:

pip install -e .
pip install -e ".[redis]"

Host wiring

INSTALLED_APPS = [
    # ...
    "ai_agent",
    "catalog",  # your domain app
]

AI_AGENT = {
    "SUPERVISOR_MODEL": "openai:gpt-4.1-mini",
    "SUBAGENT_MODEL": "openai:gpt-4.1-nano",
    "AUTHENTICATE_TOKEN": "myproject.auth.authenticate_access_token",
    "PLATFORM_NAME": "Acme",
    "PLATFORM_DOMAINS": ("catalog", "billing"),
    "EVAL_MODULE": "catalog.evals",
}

AUTHENTICATE_TOKEN is required. It must be a callable (or dotted path to one) that accepts a bearer token and returns a user object with .pk.

Optional evals — point EVAL_MODULE at a host module that defines:

• seed_eval_world() -> EvalWorld (users plus prompt_vars for case prompts)
• snapshot_world(world, kind) -> dict (optional; used by mutation cases)
• CASES: list[EvalCase] (or partitioned ROUTING_CASES / TOOL_CASES / MUTATION_CASES / E2E_CASES)

The library always includes safety routing cases (off-policy greeting, jailbreaks). Domain tool and mutation cases come from the host module. See dummy/evals.py in this repo.

HTTP_TRUSTED_PROXY_COUNT (default 0) controls whether the CopilotKit HTTP throttle reads X-Forwarded-For. Leave it at 0 unless the app sits behind that many trusted reverse proxies; otherwise clients can rotate a spoofed header and bypass the limit.

Include the chat/memory HTTP API:

urlpatterns = [
    path("api/agent/", include("ai_agent.urls")),
]

Opt a host app into the agent by setting flags on its AppConfig:

class CatalogConfig(AppConfig):
    name = "catalog"
    agent_expose = True
    agent_description = "List and create catalog items for the authenticated user."
    agent_exclude = ("internal_webhook",)
    agent_memory_profile = "catalog.memory_schemas.CatalogProfile"
    agent_memory_collections = (
        "ai_agent.memory.schemas.SemanticFact",
        "catalog.memory_schemas.CatalogAccount",
    )

Mark individual views even when the app is off:

from ai_agent.expose import agent_expose, agent_exclude

@agent_expose
def public_note(request):
    ...

LangGraph Agent Server

This package is not a host Docker image. Install django-auto-agent and your Django project into the Agent Server image, then point langgraph.json at the library entrypoints (see the example langgraph.json in this repo):

• graphs: ai_agent.studio:graph, ai_agent.studio:memory
• auth: ai_agent.auth:auth
• HTTP app: ai_agent.studio:app

Set DJANGO_SETTINGS_MODULE (or AI_AGENT_STUDIO_DJANGO_SETTINGS) to your settings module before starting Studio.

The example langgraph.json uses cors.allow_origins: ["*"]. Auth is Bearer-token based (not cookies), so CSRF exposure is limited; restrict origins in production.

Tests

From this repo:

python manage.py test ai_agent

The dummy and notes apps plus tests/settings.py are a sample host. They are not installed with the wheel.