Hey Pickaxe community - looking for implementation advice.
Use case: A client has a VR school walkthrough built in Unreal Engine. Users move through the building and can click beacons/nodes inside rooms to learn more (ex: classroom features, equipment, policies, accessibility, etc.).
We have built a Pickaxe “Guide” AI trained on the school + every room’s details. The goal is to make it feel situational:
- If the user enters the gym (or clicks a gym beacon), Unreal should send a small update like: user_location=gym
- Then the Pickaxe AI should immediately shift context, like: “I see you’re in the gym - want to hear about the equipment, seating capacity, or booking rules?”
What we need: A reliable way for Unreal Engine to communicate the user’s current room/beacon context to Pickaxe so the AI can tailor its responses in real time.
Questions for the community:
1. What’s the best pattern to “inject” external context into a Pickaxe chat session? (URL params, hidden fields, API/webhook, session variables, etc.)
-
Can Pickaxe maintain state per user session (current room) and update it repeatedly as the user moves?
-
If there’s an API option: what does the request flow typically look like from an external app to “append” context before the next AI response?
-
Any example setups for game engine → Pickaxe integrations (Unreal/Unity/WebXR) would be amazing.
Goal outcome: When the user changes rooms or clicks a beacon, the next AI message should behave as if it “knows where they are” without the user having to type it.
If anyone has a recommended architecture, docs, or a working example, I’d really appreciate it.