Adding AI Models via User's Gateway ServicePublic

Done.

Created:

• C:\Users\Daniel.Esuga\.pi\agent\models.json

with provider isw-gateway and model gpt-5-codex.

Next, set your key env var (PowerShell):

$env:ISW_AI_GATEWAY_KEY=[REDACTED]

Persist it for future terminals:

setx ISW_AI_GATEWAY_KEY "sk-..."

Then restart pi (or pi-ui server) and run /model, select:

• Provider: isw-gateway
• Model: GPT-5 Codex (ISW Gateway)

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.

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.

1) Set your gateway key as env var

Use any variable name, e.g.:

export ISW_AI_GATEWAY_KEY=[REDACTED]

2) Add provider + models in ~/.pi/agent/models.json

{  "providers": {    "isw-gateway": {      "baseUrl": "https://ai-gateway.isw.la/v1",      "api": "openai-completions",      "apiKey": "ISW_AI_GATEWAY_KEY",      "models": [        {          "id": "openai/gpt-4o-mini",          "name": "GPT-4o Mini (ISW Gateway)",          "reasoning": false,          "input": ["text", "image"],          "contextWindow": 128000,          "maxTokens": 16384,          "cost": { "input": 0, "output": 0, "cacheRead": 0, "cacheWrite": 0 }        }      ]    }  }}

Use the exact model IDs your gateway exposes (from its /v1/models).

3) Select it

• In pi CLI: /model → provider isw-gateway
• In pi-ui app: open model selector (same underlying registry)

If 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.