Max context always stuck at 128k tokens in pi

[sh1ftred]

Bug Description

The max context size for models in the pi command is always stuck at 128k tokens, regardless of the actual model being used or its documented context window.

Expected Behavior

The max context should reflect the actual supported context window of the selected model (e.g., 200k, 1M tokens for larger models, etc.).

Actual Behavior

Max context remains fixed at 128k tokens regardless of model selection.

Steps to Reproduce

1. Run routstr pi with a model that supports larger context
2. Query the max context size
3. Observe that it always reports 128k tokens

Environment

• routstrd version: [include version]
• OS: [your OS]

Additional Context

This appears to be a hardcoded value somewhere in the codebase. Need to investigate where the context size logic is implemented and ensure it properly reads from model metadata.