Getting Started

Quickstart

This is the fast integration loop after setup: install, initialize, ingest, retrieve, feedback.

Before you start

First choose your installation route on Installation & Setup: Orbit Cloud (API key) or Self-Hosted Orbit (local JWT + local API URL).

Using Orbit Cloud? Create your API key in Dashboard, then set ORBIT_API_KEY.

1

Install the SDK

terminal
pip install orbit-memory
2

Create a client

Use API key for Cloud or JWT token for Self-Hosted. Same SDK call, different credential source.

app.py
import os
from orbit import MemoryEngine

engine = MemoryEngine(
    api_key=os.getenv("ORBIT_API_KEY") or os.getenv("ORBIT_JWT_TOKEN"),
    base_url=os.getenv("ORBIT_BASE_URL") or os.getenv("ORBIT_API_BASE_URL", "http://localhost:8000"),
)
3

Ingest user and assistant signals

Ingest both sides of the interaction so retrieval can model progress, style, and outcomes.

app.py
engine.ingest(
    content="User completed lesson 10",
    event_type="learning_progress",
    entity_id="alice",
)
4

Retrieve focused context

Orbit handles ranking, decay, and personalization inference under the hood.

app.py
results = engine.retrieve(
    query="What should I know before I answer?",
    entity_id="alice",
    limit=5,
)
5

Send feedback

Feedback is the learning signal. If you skip this, Orbit cannot tune ranking as effectively.

app.py
engine.feedback(
    memory_id=results.memories[0].memory_id,
    helpful=True,
    outcome_value=1.0,
)

Automatic personalization, no sidecar service

Orbit can create inferred memories from repeated behavior and feedback trends, for example inferred_learning_pattern and inferred_preference. You keep shipping features; Orbit keeps learning users.

Next

Integration Guide ->