The user wanted to add new AI models through their AI gateway service using a specific key. The assistant guided them on how to configure the gateway and successfully added the requested model for use. The user was instructed on setting environment variables and restarting the system to activate the new model.
/model
If you want to change the active model, tell me which one to use (or I can list common options and recommend one by task).
If you meant the /model slash command behavior in pi, say that and I’ll check the exact implementation/docs in this repo.
I want to add models via my ai gateway.
https://ai-gateway.isw.la/
Key: REDACTED
Perfect — you can do this without code changes using ~/.pi/agent/models.json.
Also: since you pasted a live key here, rotate it after setup.
Use any variable name, e.g.:
~/.pi/agent/models.jsonUse the exact model IDs your gateway exposes (from its /v1/models).
/model → provider isw-gatewayIf your gateway is not fully OpenAI-compatible, add compat flags per model (e.g. maxTokensField, supportsUsageInStreaming).
If you want, I can generate a ready-to-paste models.json from the exact model IDs you have.
Go ahead. I want to use gpt-5-codex
Done.
Created:
C:\Users\Daniel.Esuga\.pi\agent\models.jsonwith provider isw-gateway and model gpt-5-codex.
Next, set your key env var (PowerShell):
Persist it for future terminals:
Then restart pi (or pi-ui server) and run /model, select:
isw-gatewayGPT-5 Codex (ISW Gateway)