docs: auto-sync gateway with merged PRs (through 2026-09-19)

packages/kilo-docs/pages/ai-providers/index.md

@@ -89,6 +89,15 @@ To allow only specific providers and disable everything else, use `enabled_provi
 
 Both fields accept provider IDs — the lowercase identifier used in the `provider/model` format (e.g. `kilo`, `anthropic`, `openai`, `google`, `groq`).
 
+## Managing Provider Connections
+
+In the VS Code extension, open the **Providers** tab in Settings to manage a connected provider:
+
+- **Change API key** — replaces the stored API key for providers that use a standard API-key form. Use it to rotate a key without disconnecting the provider first.
+- **Disconnect** — removes the provider's stored credentials and its saved provider configuration from both global and project config. The provider returns to the available catalog so you can reconnect it with a different key, with no restart or config edit.
+
+Disconnect takes effect immediately and refreshes the connected and disabled provider lists. Explicit `disabled_providers` entries and unrelated configuration are preserved, and deleting a custom provider is unchanged.
+
 ## Next Steps
 
 - **New to Kilo Code?** Start with the [Kilo Code provider](/docs/ai-providers/kilocode) - no setup required

packages/kilo-docs/pages/ai-providers/openai-chatgpt-plus-pro.md

@@ -50,6 +50,15 @@ Then set your default model to one of the OpenAI Codex models available in Kilo
 {% /tab %}
 {% /tabs %}
 
+## Checking your Codex usage
+
+When you are signed in with ChatGPT, Kilo reports your Codex quota alongside your other connected providers:
+
+- **VS Code:** open the **Profile** view and find the **Plans & usage** section.
+- **CLI:** run `/usage` (aliases `/plans` and `/quota`) and press `ctrl+r` to refresh.
+
+Codex shows its plan and each usage window — for example, a five-hour and a weekly limit — with the percentage used and the reset time. A **Manage** link opens your Codex usage settings at chatgpt.com. If your ChatGPT sign-in has expired, the view says so and you can reconnect from the provider settings.
+
 ## Tips and Notes
 
 - **Subscription Required:** You need an active ChatGPT Plus or Pro subscription. This provider won't work with free ChatGPT accounts. [Codex is included](https://developers.openai.com/codex/pricing/) in ChatGPT Plus, Pro, Business, Edu, and Enterprise plans. See [OpenAI's ChatGPT plans](https://chatgpt.com/pricing/) for more information.

packages/kilo-docs/pages/gateway/authentication.md

@@ -92,7 +92,6 @@ BYOK lets you use your own provider API keys with the Kilo AI Gateway. When a BY
 | BytePlus Coding Plan | `byteplus-coding` |
 | Chutes BYOK | `chutes-byok` |
 | Codestral (FIM) | `codestral` |
-| CrofAI | `crofai` |
 | Inceptron BYOK | `inceptron-byok` |
 | Kimi Code | `kimi-coding` |
 | Martian | `martian` |

packages/kilo-docs/pages/gateway/models-and-providers.md

@@ -56,7 +56,7 @@ This returns model information including pricing, context window, and supported
 
 ### Free models
 
-Several models are available at no cost, subject to rate limits. Availability changes; check the [live model catalog](https://kilo.ai/models) for current free options and model IDs.
+Several models are available at no cost, subject to rate limits. Availability changes; check the [live model catalog](https://kilo.ai/models) for current free options and model IDs. This includes OpenRouter stealth alpha models, whose IDs are in the `stealth/` namespace and end in `-alpha` (for example, `stealth/ox-alpha`).
 
 Free models are available to both authenticated and anonymous users. Anonymous users are rate-limited to 200 requests per hour per IP address.
 
@@ -66,7 +66,7 @@ For NVIDIA free endpoints (Super/Ultra/etc): Trial use only - do not submit pers
 
 ## Auto models
 
-Auto virtual models select an underlying model using tier-specific routing. Frontier uses the `x-kilocode-mode` request header. Efficient classifies task difficulty in session context and falls back to the API interface for its baseline model, Free uses deterministic affinity across available candidates, and Small uses account balance.
+Auto virtual models select an underlying model using tier-specific routing. Frontier uses the `x-kilocode-mode` request header. Efficient classifies task difficulty in session context and falls back to a fixed baseline model, Free uses deterministic affinity across available candidates, and Small uses account balance.
 
 {% callout type="info" title="Underlying models can change" %}
 The mappings below reflect the current routing. The underlying models behind each `kilo-auto/*` tier are updated server-side as better options become available or as providers change pricing and availability — the tier IDs themselves remain stable.
@@ -84,13 +84,7 @@ Highest performance and capability for any task. Frontier requests are sent with
 
 ### `kilo-auto/efficient`
 
-Session-aware routing that classifies each request by difficulty and routes to the cheapest model proven accurate enough for the task. When no confident routing decision can be made, requests fall back to a baseline model resolved by the API interface used by the client.
-
-| API interface | Resolved Model | Reasoning effort |
-|---|---|---|
-| Completions (default) | `qwen/qwen3.6-plus` | enabled |
-| Responses API | `openai/gpt-5.5` | low |
-| Messages API | `anthropic/claude-sonnet-4.6` | low |
+Session-aware routing that classifies each request by difficulty and routes to the cheapest model proven accurate enough for the task. When no confident routing decision can be made, requests fall back to `z-ai/glm-5.3-flash`.
 
 ### `kilo-auto/free`
 
@@ -106,7 +100,7 @@ Automatically routes to a small, fast model for lightweight background tasks (se
 
 | Condition | Resolved Model |
 |---|---|
-| Account has paid balance | `google/gemma-4-31b-it` |
+| Account has paid balance | `google/gemma-4-26b-a4b-it` |
 | No balance / free account | `google/gemma-4-26b-a4b-it:free` |
 
 ### Example usage