interlatent 0.4.0

Interlatent SDK — hosted API client for RL interpretability dashboards.

Interlatent Python SDK

Official Python client for the Interlatent hosted API.

Browse environments, inspect runs, read latent activations, and view episode data from your RL interpretability dashboards.

Install

pip install interlatent

Quickstart

from interlatent import Interlatent

client = Interlatent(api_key="sk-...")

# Browse
envs = client.environments.list()
runs = client.environments.runs("Ant-v5")

# Inspect
run = client.runs.retrieve("run_abc123")
latents = client.latents.list("run_abc123")
meta = client.episodes.meta("episode_id")
chunk = client.episodes.chunk("episode_id", 0)

# Upload
client.runs.create(run_id="new_run", environment="Ant-v5", layer="policy_net.2")
client.latents.create("new_run", latents=[{"latent_id": 0, "name": "hip torque"}])
client.checkpoints.create("new_run", report={"failure_summary": {}})

Resources

Resource	Methods
client.index	retrieve()
client.environments	list(), runs(env_id)
client.runs	retrieve(), create(), update(), upload_urls(), upload_complete()
client.episodes	meta(), chunk(), frame()
client.latents	list(), retrieve(), create()
client.checkpoints	list(), retrieve(), create()
client.auth	login()