Pre-release
This release is a pre-release and may not be stable for production use.
Django AI SDK
Build AI agents in Django. Batteries included.
Project Status: Read This First
This is a beta. The API may still change as we find better patterns. Here's what that means for you:
- API may still evolve: some interfaces will shift as we find better patterns.
- Not for production yet: use this for experimentation, prototypes, and side projects. Keep critical workloads elsewhere until we hit stable.
- Watch the repo: Things change quickly. Star & watch to stay in the loop.
- Your feedback shapes the SDK: Break things, open issues, tell us what hurts.
We'd love to have you along for the ride, just keep your seatbelt on.
Install
pip install django-ai-sdk
Or with uv:
uv add django-ai-sdk
Quick Start
1. Add to INSTALLED_APPS
# settings.py
INSTALLED_APPS = [
...
"django_ai_sdk",
]
Then run python manage.py migrate.
2. Define your agent
# agents.py
from django_ai_sdk import Agent
from django_ai_sdk.adapters.base import Stream
from django_ai_sdk.generators import openai_responses_chat
from haystack import Pipeline
class HelpDeskAgent(Agent):
name = "Help Desk"
model = "gpt-5-mini"
instructions = "You are a helpful support agent."
llm = openai_responses_chat
async def get_pipeline_adapter(self, thread_id=None, user=None):
storage_adapter = await self.get_storage_adapter(thread_id)
generator = self.get_llm()
return Stream(
pipeline=Pipeline(),
generator=generator,
storage_adapter=storage_adapter,
)
For non-streaming tasks (title generation, structured output), use Run instead:
from django_ai_sdk.adapters.base import Run
async def get_run_adapter(self, thread_id=None, user=None):
return Run(generator=self.get_llm())
3. Return a streaming response
# views.py
from .agents import HelpDeskAgent
agent = HelpDeskAgent()
@router.post("/chat")
async def chat(request, payload: ChatRequest):
return await agent.as_view(
payload.messages,
thread_id=payload.thread_id,
)
Features
- Agents: A Django
Agentclass you can run from a view, a task, or the API, streaming or not, with tool calling built in. - Workflows: Multi-step, definition-driven runs with persisted steps and actions, so a process can span multiple agent calls with a record of what happened.
- Subagents: Agents can delegate to other agents as tools, so one agent can orchestrate specialists instead of doing everything itself.
- Tool Calling: MCP, memory, and custom tools, all managed by your Agent.
- Streaming Responses: Built-in SSE streaming. Works with Vercel AI SDK protocol.
- Conversation Storage: Automatic message persistence. Thread-based history out of the box.
- RAG Pipelines: BM25, ChromaDB, and Qdrant hybrid search with query expansion, for when an agent needs to ground itself in your documents.
- Artifacts: 16 structured UI types (tables, plans, approval cards, code blocks, and more) submitted by the LLM via tool calls.
- File Processing: Document upload with pipeline-based processing (text, CSV, JSON, DOCX, PPTX, XLSX. Extraction transforms for metadata embedding.
- Integrations: Third-party tools as self-registering Django apps, with caching, circuit breaking and OAuth built in.
- Tracing: Opt-in Haystack tracing persisted to the ORM, with per-thread and per-message token accounting.
- Reindexing: Hot-reload documents. Cached embeddings with simple refresh API.
Documentation
Full documentation and examples: github.com/django-ai-sdk/django-ai-sdk
1 maintainer
Metadata
1 maintainer
Release files for django-ai-sdk 0.1.2a13
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | |
|---|---|---|---|---|
| django_ai_sdk-0.1.2a13.tar.gz | 233.6 kB | Without provenance | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| django_ai_sdk-0.1.2a13-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 531.1 kB
Release files / django_ai_sdk-0.1.2a13.tar.gz
| Download URL | django_ai_sdk-0.1.2a13.tar.gz |
|---|---|
| Size | 233.6 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
03915ed85241ab4a6f35777888ca03832d8c84699af585e8d2ab1406120f27c1
|
|
BLAKE2b-256 checksum How to use checksums |
c198d2199c90977b4600c3c62948adfaeaab99f8c548b54a4f04a6a16520c3da
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
uv/0.11.2 {"installer":{"name":"uv","version":"0.11.2","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"macOS","version":null,"id":null,"libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":null}
|
Release files / django_ai_sdk-0.1.2a13-py3-none-any.whl
| Download URL | django_ai_sdk-0.1.2a13-py3-none-any.whl |
|---|---|
| Size | 297.5 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
800ab13fc782eb6812fe121cd3c6e43c4a10bb67d7535f3a1ebe291a183e7bf8
|
|
BLAKE2b-256 checksum How to use checksums |
f2baecabd07f2f2d6b59b2238aebff42b85253d16adfff7bbcf043befee6b69a
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
uv/0.11.2 {"installer":{"name":"uv","version":"0.11.2","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"macOS","version":null,"id":null,"libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":null}
|
Security for django-ai-sdk 0.1.2a13
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
