CinvanaAI
← All projects

Developer starter · Games & interaction / Workflows & capabilities

Astra-l-Projection

Connect conversation to a camera and bounded movement in your own Unreal project.

Repository: CinvanaAI/Astra-l-Projection

Recorded native checks and diagnostic provider. Experimental source developer starter. Your own Unreal C++ project supplies its map, character, collision, and assets.

Give the conversation a viewpoint.

The release includes a 32-second walkthrough using labeled diagrams and actual eye-camera stills. The portable client, bridge, adapter, and installer checks were rerun during this catalogue preparation.

Example input
A local Play session with a blocking floor, the agent actor, and the source plugin compiled into a compatible Unreal project.
Captured result
Release record: 11 native checks; two inspected 640 × 360 camera captures; diagnostic human-event → bridge → provider → native reply. Fresh portable checks: 31 passed.

Try the example.

Start with the linked native setup guide. Unreal and native build tools must already be installed. Set QUEUE_PATH to your project’s Saved/AgentEmbodiment directory after starting Play.

python client/player.py --root QUEUE_PATH observe
python client/player.py --root QUEUE_PATH look --yaw-delta-deg 15
python client/player.py --root QUEUE_PATH stop

Inspect the returned eye-camera image and correlated action result. The native guide explains the expected records.

Complete setup and instructions ↗

The interesting part.

A local command queue links a native body to a Python client and a replaceable conversation provider.

Adapted developer starter

Where it came from.

Embodiment runtime controls and conversation routing built with Astra.

Extracts that foundation and adds portable Unreal connection tools and release-specific checks.

Source starter requires the recipient Unreal project; deterministic native evidence does not establish live model inference.

Source ↗