Problem

The Honcho memory plugin's get_config_schema() returns a field with key base_url (snake_case). This is what hermes memory setup writes to ~/.hermes/honcho.json.

However, HonchoClientConfig.from_global_config() in plugins/memory/honcho/client.py reads baseUrl (camelCase) from that same file:

base_url = (
    raw.get("baseUrl")    # ← camelCase
    or os.environ.get("HONCHO_BASE_URL", "").strip()
    or None
)

Impact

Users who run hermes memory setup to configure a self-hosted Honcho instance (e.g. http://localhost:8000) have their base URL silently ignored. The client falls back to the default https://api.honcho.dev cloud endpoint, then fails with Invalid API key because the user's API key was set for their self-hosted instance.

Symptoms:

• hermes honcho status shows Connection... OK (because connectivity check passes differently)
• But Peer data unavailable: Invalid API key appears
• Honcho SDK internally hits https://api.honcho.dev instead of the configured URL

Fix

Align the schema key to baseUrl to match the reader. One-character change.

Verification

Tested locally against self-hosted Honcho 3.0.5 at http://localhost:8000:

Before fix (with base_url):

Peer data unavailable: Invalid API key

After fix (with baseUrl):

Connection... OK
No peer data yet (accumulates after first conversation)

Note

save_config() in the same file has the docstring "Write config to $HERMES_HOME/honcho.json (Honcho SDK native format)". The Honcho Python SDK uses camelCase (baseUrl), so the reader already uses the correct native format. The schema key was the outlier.