Pantheon v0.3.1 — Independent API connections and optional mixed models

Pantheon now lets you use one provider for every AI role, or optionally mix providers per role without changing their skills.

• Save each supplier's API key and endpoint independently in Settings → Setup. Adding a connection no longer switches Hermes or other role models.
• In Models, use Apply to all AI roles for a shared model, or choose providers/models per role, including Hermes.
• Current chat only saves that conversation's choices. New chat defaults requires an explicit save and preserves existing conversations, including older chats and choices from other tabs.
• Test each selected role's actual model and see its connection state. Connection tests do not establish access to every model at a supplier.
• Preserve existing dotenv/keyring credentials and older YAML configurations. OpenAI and DeepSeek connections stay independent across Web, CLI, SDK and media.

One provider is sufficient. Additional providers are optional and connecting them does not itself consume tokens; actual models, tokens and calls determine cost. Use the existing usage/budget controls to manage spending.

Validation: 243 backend tests and 18 Chromium checks passed locally; 3 intentional live-account/native-platform tests skipped. OpenAI and Anthropic calls were simulated, so verify your own account access with Test API. Real credentials are not required for the default test suite.

Install/update commands and platform-specific instructions: English, 简体中文.