🤖 feat: enable/disable providers without removing API keys

[ibetitsmike]

Summary

Add per-provider enable/disable toggles that let users turn off AI providers without deleting their saved API keys or credentials. The enabled state persists server-side in providers.jsonc and is enforced at both the backend (request-time) and frontend (UI gating) layers.

Background

Addresses #2034. Users debugging provider issues or isolating behavior need a way to temporarily deactivate providers. Previously, the only way was to delete the API key, which is destructive and requires re-entry.

Implementation

Backend

• providerService.ts: Reads enabled field from raw JSONC, computes isEnabled and isConfigured. Disabled providers surface as not configured in the UI. Both setConfig() and setConfigValue() normalize the enabled key (stores false as boolean, deletes on re-enable).
• providerModelFactory.ts: Early guard in createModel() returns provider_disabled error when a non-gateway provider is disabled. Gateway bypass is preserved — disabled providers are still accessible via Mux Gateway if allowlisted.
• voiceService.ts: Checks OpenAI enabled state before transcription to prevent unexpected API calls.
• Shared helper: isProviderDisabledInConfig() centralizes the enabled check, handling both boolean false and string "false" (manual JSONC edits).

Frontend

• ProvidersSection: New "Enabled" switch per provider (excludes gateway). Status dot uses design-system CSS variables: green (configured), yellow/amber (disabled), gray (not configured). Warning banner when no providers are configured.
• VoiceInputButton: Three-tier disabled states (HTTPS → provider disabled → no API key). Mid-session exemption prevents lockout during active recording.
• ChatInput: Tracks openAIProviderEnabled state for voice input keybind gating.
• Model selector: Custom models from disabled providers are filtered from suggestions.

Error Plumbing

• New provider_disabled discriminant in SendMessageErrorSchema
• Formatted with proper provider display names (e.g., "OpenAI" not "openai")
• Marked non-retryable in retry eligibility
• Toast with resolution hint and docs link

Schema

• isEnabled: z.boolean().default(true) in ProviderConfigInfoSchema — backward-compatible default for version skew resilience.

Validation

• make static-check passes (ESLint, typecheck, prettier, mintlify links)
• 93 targeted tests pass across 8 test files covering:
  • Provider service (getConfig, setConfig, Codex OAuth + disabled)
  • Model factory (provider_disabled error, gateway bypass, positive cases)
  • Voice service (disabled blocks fetch, enabled proceeds)
  • Schema conformance (isEnabled default, all provider variants)
  • Error formatting, retry eligibility, toast creation
  • Model filtering (disabled providers excluded from custom models)

Risks

Low. The isEnabled field defaults to true when missing, so existing configs are unaffected. The gateway bypass is explicitly tested. The main risk is cosmetic — the yellow status dot color may need design review for theme consistency.

---

Generated with mux • Model: anthropic:claude-opus-4-6 • Thinking: xhigh • Cost: $9.24

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: e5649b37df

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[ibetitsmike]

@codex review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: c1af1109ca

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[ibetitsmike]

@codex review

[chatgpt-codex-connector]

Codex Review: Didn't find any major issues. You're on a roll.

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[Shelnutt2]

Thanks for this!