hermes - 💡(How to fix) Fix [Bug]: 'No inference provider configured' in interactive chat despite '-z' working, openai-codex provider (v0.13.0)

Description

Interactive chat mode (hermes chat / hermes --tui chat / profile-aliased equivalents) reports "No inference provider configured" and refuses to send the first message, even though:

• hermes config show confirms provider: openai-codex and model: gpt-5.5
• The same profile's hermes -z "..." one-shot mode runs inference successfully against openai-codex with no error
• hermes auth status shows the openai-codex OAuth token as valid
• The gateway service (which uses its own provider-resolution path) successfully serves inference to a connected Telegram client using the same profile config

Looks like chat mode and -z mode (and gateway mode) use different provider-resolution paths and chat mode misses the openai-codex setting somehow.

Reproduction

hermes profile create personal
hermes auth add openai-codex --type oauth   # complete the OAuth flow
hermes -p personal config set provider openai-codex
hermes -p personal config set model gpt-5.5

# This works — inference returns a real response:
hermes -p personal -z "say OK"
# → OK

# This fails immediately:
hermes -p personal chat
# Send any message → "No inference provider configured. Run 'hermes model' to choose a provider"

Environment

• Hermes Agent v0.13.0 (2026.5.7) — macOS Homebrew install
• Python 3.14.4
• OpenAI SDK 2.35.0
• macOS 15.5 (24F74)
• Provider: openai-codex (OAuth, not API key)
• Model: gpt-5.5

Workaround

Stick to hermes -z "..." non-interactive mode for CLI use, or run inference through the gateway service. The bug doesn't block gateway operation.

What I expected

Chat mode should resolve providers the same way -z mode and the gateway do.