skeino 0.3.0

Reusable, embeddable replacement for the langgraph dev HTTP server: threads, runs, streaming, and assistants over any LangGraph graph.

skeino

A reusable, embeddable replacement for the langgraph dev HTTP server.

📖 Full documentation: https://ranjitjana027.github.io/skeino/

skeino exposes the LangGraph Studio-compatible REST surface (threads, runs, streaming, assistants) over any user-supplied LangGraph graph. It is built to be:

• Modular — api, ops, persistence, streaming, serialization are separate concerns with explicit dependencies.
• Pluggable — checkpointers register themselves via a small decorator-based registry; Postgres and in-memory ship out of the box, Redis / Mongo / etc. can be added without touching skeino core.
• Two entry points — programmatic create_app(graphs={...}) and a high-level from_langgraph_json("langgraph.json") loader.

Install

pip install skeino

Quickstart

# Programmatic
from skeino import create_app, SkeinoSettings
from my_project.graph import graph

app = create_app(
    graphs={"my_agent": graph},
    settings=SkeinoSettings(postgres_uri="postgresql://..."),
)

# langgraph.json driven
from skeino import from_langgraph_json
app = from_langgraph_json("langgraph.json")

Run with uvicorn:

uvicorn app:app --reload --port 8000

Endpoints (v1)

• GET /info, GET /api/health, GET /api/initial-message
• POST /assistants/search, GET /assistants/{id}, GET /assistants/{id}/{schemas|graph|subgraphs}
• POST /threads, GET /threads/{id}, POST /threads/search, GET /threads/{id}/state, GET|POST /threads/{id}/history
• POST /threads/{id}/runs, POST /threads/{id}/runs/stream, GET /threads/{id}/runs, GET /threads/{id}/runs/{run_id}

Out of scope for v1: /store/*, /runs/crons, webhooks, auth.

Status

Pre-release. API may change.

Contributing

Contributions are welcome — see CONTRIBUTING.md for local setup and the required checks.

License

Licensed under the Apache License 2.0.