Write context window and max tokens into Pi's models.json on configure

[devin-ai-integration]

Summary

Before this change, requesty configure for Pi wrote only {"id": "<model>"} into ~/.pi/agent/models.json. Pi then used its custom-model defaults (contextWindow ?? 128000, maxTokens ?? 16384 in model-registry.ts), so a 1M-context model run through plain pi stopped at 128k.

The TUI model picker already loads client.Model from /v1/models. It now copies the picked model's ContextWindow and MaxOutputTokens into ConfigureOptions. In both configure paths, Pi builds its entry from those with piCatalogModelFrom, the same function the requesty pi launch catalog uses:

{"id": "openai/gpt-5.5", "contextWindow": 1050000, "maxTokens": 128000}

If the list has no value for one of these, the launch defaults are written (piCatalogContextWindow 200k, piCatalogMaxTokens 8192). Other harnesses ignore the new options.

Link to Devin session: https://app.devin.ai/sessions/feb93c6a56e847888cf65ee01f6e6559
Open in Devin Desktop: https://app.devin.ai/desktop/session/feb93c6a56e847888cf65ee01f6e6559?variant=devin
Requested by: @requesty-JohnCosta27

[devin-ai-integration]

I'll fix CI failures and address comments from users with write access. I'll skip comments containing "(aside)".

• Disable automatic comment, CI, and merge conflict monitoring

[github-actions]

AI Cost

No spend detected for branch devin/1791370894-pi-configure-context-window in the last 30 days.

_{Updated 2026-10-07T11:13Z · last 30 days · data from Requesty}

[requesty-JohnCosta27]

I'm not sure what you're doing here. If we have access to the Model, don't we also have the context window information?

[devin-ai-integration]

Agreed, refetching wasn't needed. The TUI model picker already has the full client.Model. In dfe9c5f the picker copies ContextWindow/MaxOutputTokens into ConfigureOptions with the ID, and Pi builds its entry from those with piCatalogModelFrom. That drops the network call and the httptest server. If the list has no values, the launch defaults (200k / 8192) apply.

Written by Devin