hermes - 💡(How to fix) Fix [Bug]: Auxiliary compression model does not inherit context_length from custom_providers

Description

When the compression model is set to auto (model: ''), it inherits the main model name from model.default but does NOT look up the context_length from custom_providers for that model.

Steps to reproduce

1. Configure a named custom provider with explicit context_length:

model:
  default: qwen/deepseek-v4-pro
  provider: custom:my-gateway

custom_providers:
- name: my-gateway
  base_url: https://ais.example.com/v1
  models:
    qwen/deepseek-v4-pro:
      context_length: 1000000

auxiliary:
  compression:
    provider: auto
    model: ''

1. Start a session, observe:

Compression model qwen/deepseek-v4-pro (ais.example.com) context is 256,000 tokens,
but the main model compression threshold was 500,000 tokens.
Auto-lowered this session threshold to 256,000 tokens.

Expected

Compression model should inherit context_length: 1000000 from custom_providers.

Actual

Uses hardcoded default 256,000, ignoring custom_providers.

Workaround

Add context_length directly to model block:

model:
  context_length: 1000000

Impact

Every subagent prints this warning. Users with custom OpenAI-compatible providers whose /models endpoint does not return context_length are affected.