Merge d0e0eea7b8 into 5745957429

README.md

@@ -404,6 +404,7 @@ PicoClaw supports 30+ LLM providers through the `model_list` configuration. Use
 | [Xiaomi MiMo](https://platform.xiaomimimo.com/) | `mimo/` | Required | MiMo models |
 | [Ollama](https://ollama.com/) | `ollama/` | Not needed | Local models, self-hosted |
 | [vLLM](https://docs.vllm.ai/) | `vllm/` | Not needed | Local deployment, OpenAI-compatible |
+| [OVMS](https://docs.openvino.ai/2026/model-server/ovms_docs_llm_quickstart.html) | `ovms/` | Not needed | Intel OpenVINO Model Server, OpenAI-compatible |
 | [LiteLLM](https://docs.litellm.ai/) | `litellm/` | Varies | Proxy for 100+ providers |
 | [Azure OpenAI](https://portal.azure.com/) | `azure/` | Required | Enterprise Azure deployment |
 | [GitHub Copilot](https://github.com/features/copilot) | `github-copilot/` | OAuth | Device code login |
@@ -413,7 +414,7 @@ PicoClaw supports 30+ LLM providers through the `model_list` configuration. Use
 > \* AWS Bedrock requires build tag: `go build -tags bedrock`. Set `api_base` to a region name (e.g., `us-east-1`) for automatic endpoint resolution across all AWS partitions (aws, aws-cn, aws-us-gov). When using a full endpoint URL instead, you must also configure `AWS_REGION` via environment variable or AWS config/profile.
 
 <details>
-<summary><b>Local deployment (Ollama, vLLM, etc.)</b></summary>
+<summary><b>Local deployment (Ollama, vLLM, OVMS, etc.)</b></summary>
 
 **Ollama:**
 ```json
@@ -441,6 +442,19 @@ PicoClaw supports 30+ LLM providers through the `model_list` configuration. Use
 }
 ```
 
+**OVMS (OpenVINO Model Server):**
+```json
+{
+  "model_list": [
+    {
+      "model_name": "local-ovms",
+      "model": "ovms/your-model",
+      "api_base": "http://localhost:8000/v3"
+    }
+  ]
+}
+```
+
 For full provider configuration details, see [Providers & Models](docs/guides/providers.md).
 
 </details>

ROADMAP.md

@@ -43,7 +43,7 @@
 
 * **Provider**
   * [**Architecture Upgrade**](https://github.com/sipeed/picoclaw/issues/283): Refactor from "Vendor-based" to "Protocol-based" classification (e.g., OpenAI-compatible, Ollama-compatible). *(Status: In progress by @Daming, ETA 5 days)*
-  * **Local Models**: Deep integration with **Ollama**, **vLLM**, **LM Studio**, and **Mistral** (local inference).
+  * **Local Models**: Deep integration with **Ollama**, **vLLM**, **OVMS**, **LM Studio**, and **Mistral** (local inference).
   * **Online Models**: Continued support for frontier closed-source models.
 
 

cmd/picoclaw/internal/model/command.go

@@ -30,11 +30,13 @@ Examples:
   picoclaw model                    # Show current default model
   picoclaw model gpt-5.2           # Set gpt-5.2 as default
   picoclaw model claude-sonnet-4.6 # Set claude-sonnet-4.6 as default
-  picoclaw model local-model       # Set local VLLM server as default
+  picoclaw model local-model       # Set local vLLM server as default
+  picoclaw model local-ovms        # Set local OVMS server as default
   picoclaw model add -b URL -k KEY # Add a model from a custom endpoint
 
-Note: 'local-model' is a special value for using a local VLLM server
+Note: 'local-model' is a built-in entry for a local vLLM server (localhost:8000/v1).
-(running at localhost:8000 by default) which does not require an API key.`,
+      'local-ovms' is a built-in entry for a local OVMS server (localhost:8000/v3).
+      Neither requires an API key.`,
 		Args: cobra.MaximumNArgs(1),
 		RunE: func(cmd *cobra.Command, args []string) error {
 			configPath := internal.GetConfigPath()

docs/guides/configuration.fr.md

@@ -368,7 +368,7 @@ L'ancienne configuration `providers` est **dépréciée** et a été supprimée
 
 PicoClaw route les providers par famille de protocole :
 
-- **Compatible OpenAI** : OpenRouter, Groq, Zhipu, endpoints vLLM et la plupart des autres.
+- **Compatible OpenAI** : OpenRouter, Groq, Zhipu, endpoints vLLM, OVMS et la plupart des autres.
 - **Gemini natif** : Google Gemini via les endpoints natifs `models/*:generateContent` et `models/*:streamGenerateContent`.
 - **Anthropic** : Comportement natif de l'API Claude.
 - **Codex/OAuth** : Route d'authentification OAuth/token OpenAI.

docs/guides/configuration.ja.md

@@ -369,7 +369,7 @@ HEARTBEAT_OK を返信        ユーザーが直接結果を受信
 
 PicoClaw はプロトコルファミリーで Provider をルーティングします：
 
-- **OpenAI 互換**：OpenRouter、Groq、Zhipu、vLLM スタイルのエンドポイントなど。
+- **OpenAI 互換**：OpenRouter、Groq、Zhipu、vLLM スタイルのエンドポイント、OVMSなど。
 - **Gemini ネイティブ**：Google Gemini のネイティブ `models/*:generateContent` / `models/*:streamGenerateContent` エンドポイント。
 - **Anthropic**：Claude ネイティブ API の動作。
 - **Codex/OAuth**：OpenAI OAuth/トークン認証ルート。

docs/guides/configuration.md

@@ -592,6 +592,7 @@ For complete documentation, see [`../security/security_configuration.md`](../sec
 | **OpenRouter**          | `openrouter`      | `https://openrouter.ai/api/v1`                      | OpenAI    | [Get Key](https://openrouter.ai/keys)                            |
 | **LiteLLM Proxy**       | `litellm`         | `http://localhost:4000/v1`                          | OpenAI    | Your LiteLLM proxy key                                           |
 | **VLLM**                | `vllm`            | `http://localhost:8000/v1`                          | OpenAI    | Local                                                            |
+| **OVMS**                | `ovms`           | `http://localhost:8000/v3`                          | OpenAI    | Local (no key needed)                                            |
 | **Cerebras**            | `cerebras`        | `https://api.cerebras.ai/v1`                        | OpenAI    | [Get Key](https://cerebras.ai)                                   |
 | **VolcEngine (Doubao)** | `volcengine`      | `https://ark.cn-beijing.volces.com/api/v3`          | OpenAI    | [Get Key](https://www.volcengine.com/activity/codingplan?utm_campaign=PicoClaw&utm_content=PicoClaw&utm_medium=devrel&utm_source=OWO&utm_term=PicoClaw) |
 | **神算云**              | `shengsuanyun`    | `https://router.shengsuanyun.com/api/v1`            | OpenAI    | —                                                                |
@@ -848,7 +849,7 @@ The old `providers` configuration is **deprecated** and has been removed in V2.
 
 PicoClaw routes providers by protocol family:
 
-- **OpenAI-compatible**: OpenRouter, Groq, Zhipu, vLLM-style endpoints, and most others.
+- **OpenAI-compatible**: OpenRouter, Groq, Zhipu, vLLM-style endpoints, OVMS, and most others.
 - **Gemini native**: Google Gemini via the native `models/*:generateContent` and `models/*:streamGenerateContent` endpoints.
 - **Anthropic**: Claude-native API behavior.
 - **Codex/OAuth**: OpenAI OAuth/token authentication route.

docs/guides/configuration.pt-br.md

@@ -369,7 +369,7 @@ A configuração antiga `providers` está **depreciada** e foi removida no V2. C
 
 PicoClaw roteia providers por família de protocolo:
 
-- **Compatível com OpenAI**: OpenRouter, Groq, Zhipu, endpoints vLLM e a maioria dos outros.
+- **Compatível com OpenAI**: OpenRouter, Groq, Zhipu, endpoints vLLM, OVMS e a maioria dos outros.
 - **Gemini nativo**: Google Gemini via endpoints nativos `models/*:generateContent` e `models/*:streamGenerateContent`.
 - **Anthropic**: Comportamento nativo da API Claude.
 - **Codex/OAuth**: Rota de autenticação OAuth/token OpenAI.

docs/guides/configuration.vi.md

@@ -369,7 +369,7 @@ Cấu hình `providers` cũ đã **bị deprecated** và đã được loại b
 
 PicoClaw định tuyến provider theo họ giao thức:
 
-- **Tương thích OpenAI**: OpenRouter, Groq, Zhipu, endpoint kiểu vLLM và hầu hết các provider khác.
+- **Tương thích OpenAI**: OpenRouter, Groq, Zhipu, endpoint kiểu vLLM, OVMS và hầu hết các provider khác.
 - **Gemini native**: Google Gemini qua các endpoint native `models/*:generateContent` và `models/*:streamGenerateContent`.
 - **Anthropic**: Hành vi API Claude gốc.
 - **Codex/OAuth**: Tuyến xác thực OAuth/token OpenAI.

docs/guides/configuration.zh.md

@@ -452,6 +452,7 @@ Agent 读取 HEARTBEAT.md
 | **OpenRouter**          | `openrouter`      | `https://openrouter.ai/api/v1`                      | OpenAI    | [获取](https://openrouter.ai/keys)                               |
 | **LiteLLM Proxy**       | `litellm`         | `http://localhost:4000/v1`                          | OpenAI    | 你的 LiteLLM 代理 Key                                            |
 | **VLLM**                | `vllm`            | `http://localhost:8000/v1`                          | OpenAI    | 本地                                                             |
+| **OVMS**                | `ovms`           | `http://localhost:8000/v3`                          | OpenAI    | 本地（无需密钥）                                                 |
 | **Cerebras**            | `cerebras`        | `https://api.cerebras.ai/v1`                        | OpenAI    | [获取](https://cerebras.ai)                                      |
 | **火山引擎 (豆包)**     | `volcengine`      | `https://ark.cn-beijing.volces.com/api/v3`          | OpenAI    | [获取](https://www.volcengine.com/activity/codingplan?utm_campaign=PicoClaw&utm_content=PicoClaw&utm_medium=devrel&utm_source=OWO&utm_term=PicoClaw) |
 | **神算云**              | `shengsuanyun`    | `https://router.shengsuanyun.com/api/v1`            | OpenAI    | —                                                                |
@@ -674,7 +675,7 @@ Agent 读取 HEARTBEAT.md
 
 PicoClaw 按协议族路由提供商：
 
-- **OpenAI 兼容**：OpenRouter、Groq、智谱、vLLM 风格端点及大多数其他提供商。
+- **OpenAI 兼容**：OpenRouter、Groq、智谱、vLLM 风格端点、OVMS 及大多数其他提供商。
 - **Gemini 原生**：Google Gemini 通过原生 `models/*:generateContent` 和 `models/*:streamGenerateContent` 端点接入。
 - **Anthropic**：Claude 原生 API 行为。
 - **Codex/OAuth**：OpenAI OAuth/Token 认证路由。

docs/guides/providers.fr.md

@@ -55,6 +55,7 @@ Cette conception permet également le **support multi-agents** avec une sélecti
 | **OpenRouter**      | `openrouter/`     | `https://openrouter.ai/api/v1`                      | OpenAI    | [Get Key](https://openrouter.ai/keys)                            |
 | **LiteLLM Proxy**   | `litellm/`        | `http://localhost:4000/v1`                          | OpenAI    | Your LiteLLM proxy key                                            |
 | **VLLM**            | `vllm/`           | `http://localhost:8000/v1`                          | OpenAI    | Local                                                            |
+| **OVMS**            | `ovms/`           | `http://localhost:8000/v3`                          | OpenAI    | Local (no key needed)                                            |
 | **Cerebras**        | `cerebras/`       | `https://api.cerebras.ai/v1`                        | OpenAI    | [Get Key](https://cerebras.ai)                                   |
 | **VolcEngine (Doubao)** | `volcengine/`     | `https://ark.cn-beijing.volces.com/api/v3`          | OpenAI    | [Get Key](https://www.volcengine.com/activity/codingplan?utm_campaign=PicoClaw&utm_content=PicoClaw&utm_medium=devrel&utm_source=OWO&utm_term=PicoClaw)                        |
 | **神算云**          | `shengsuanyun/`   | `https://router.shengsuanyun.com/api/v1`            | OpenAI    | -                                                                |
@@ -105,7 +106,7 @@ Cette conception permet également le **support multi-agents** avec une sélecti
 |-------|------|--------|-------------|
 | `model_name` | string | Oui | Nom unique pour référencer ce modèle dans la config agent |
 | `model` | string | Oui | Identifiant fournisseur/modèle (ex : `openai/gpt-5.4`, `azure/gpt-5.4`, `anthropic/claude-sonnet-4.6`) |
-| `api_keys` | string[] | Oui* | Clé(s) API pour l'authentification. Plusieurs clés permettent la rotation par requête. Non requis pour les fournisseurs locaux (Ollama, LM Studio, VLLM) |
+| `api_keys` | string[] | Oui* | Clé(s) API pour l'authentification. Plusieurs clés permettent la rotation par requête. Non requis pour les fournisseurs locaux (Ollama, LM Studio, VLLM, OVMS) |
 | `api_base` | string | Non | Remplace l'URL de base API par défaut |
 | `proxy` | string | Non | URL du proxy HTTP pour cette entrée de modèle |
 | `user_agent` | string | Non | En-tête `User-Agent` personnalisé pour les requêtes API (supporté par les providers compatibles OpenAI, Gemini, Anthropic et Azure) |
@@ -298,7 +299,7 @@ Pour un guide de migration détaillé, voir [migration/model-list-migration.md](
 
 PicoClaw route les fournisseurs par famille de protocoles :
 
-- Protocole compatible OpenAI : OpenRouter, passerelles compatibles OpenAI, Groq, Zhipu et endpoints de type vLLM.
+- Protocole compatible OpenAI : OpenRouter, passerelles compatibles OpenAI, Groq, Zhipu et endpoints de type vLLM et OVMS.
 - Protocole Gemini natif : Google Gemini via les endpoints natifs `models/*:generateContent` et `models/*:streamGenerateContent`.
 - Protocole Anthropic : Comportement natif de l'API Claude.
 - Chemin Codex/OAuth : Route d'authentification OAuth/token OpenAI.

docs/guides/providers.ja.md

@@ -56,6 +56,7 @@
 | **OpenRouter**      | `openrouter/`         | `https://openrouter.ai/api/v1`                      | OpenAI     | [キーを取得](https://openrouter.ai/keys)                          |
 | **LiteLLM Proxy**   | `litellm/`            | `http://localhost:4000/v1`                          | OpenAI     | LiteLLM プロキシキー                                              |
 | **VLLM**            | `vllm/`               | `http://localhost:8000/v1`                          | OpenAI     | ローカル                                                          |
+| **OVMS**            | `ovms/`               | `http://localhost:8000/v3`                          | OpenAI     | ローカル（キー不要）                                              |
 | **Cerebras**        | `cerebras/`           | `https://api.cerebras.ai/v1`                        | OpenAI     | [キーを取得](https://cerebras.ai)                                 |
 | **VolcEngine (Doubao)** | `volcengine/`     | `https://ark.cn-beijing.volces.com/api/v3`          | OpenAI     | [キーを取得](https://www.volcengine.com/activity/codingplan?utm_campaign=PicoClaw&utm_content=PicoClaw&utm_medium=devrel&utm_source=OWO&utm_term=PicoClaw) |
 | **神算云**          | `shengsuanyun/`       | `https://router.shengsuanyun.com/api/v1`            | OpenAI     | -                                                                 |
@@ -106,7 +107,7 @@
 |-----------|------|------|------|
 | `model_name` | string | はい | agent 設定でこのモデルを参照するための一意の名前 |
 | `model` | string | はい | ベンダー/モデル識別子（例：`openai/gpt-5.4`、`azure/gpt-5.4`、`anthropic/claude-sonnet-4.6`） |
-| `api_keys` | string[] | はい* | 認証キー。複数キーでリクエストごとのローテーションが可能。ローカル provider（Ollama、LM Studio、VLLM）には不要 |
+| `api_keys` | string[] | はい* | 認証キー。複数キーでリクエストごとのローテーションが可能。ローカル provider（Ollama、LM Studio、VLLM、OVMS）には不要 |
 | `api_base` | string | いいえ | デフォルトの API エンドポイント URL を上書き |
 | `proxy` | string | いいえ | このモデルエントリの HTTP プロキシ URL |
 | `user_agent` | string | いいえ | カスタム `User-Agent` リクエストヘッダー（OpenAI 互換、Gemini、Anthropic、Azure provider で対応） |
@@ -310,7 +311,7 @@ PicoClaw はリクエスト送信前に外側の `litellm/` プレフィック
 
 PicoClaw はプロトコルファミリーごとに Provider をルーティングします：
 
-- OpenAI 互換プロトコル：OpenRouter、OpenAI 互換ゲートウェイ、Groq、Zhipu、vLLM スタイルのエンドポイント。
+- OpenAI 互換プロトコル：OpenRouter、OpenAI 互換ゲートウェイ、Groq、Zhipu、vLLM スタイルのエンドポイント、OVMS。
 - Gemini ネイティブプロトコル：Google Gemini のネイティブ `models/*:generateContent` / `models/*:streamGenerateContent` エンドポイント。
 - Anthropic プロトコル：Claude ネイティブ API 動作。
 - Codex/OAuth パス：OpenAI OAuth/Token 認証ルート。

docs/guides/providers.md

@@ -64,6 +64,7 @@ This design also enables **multi-agent support** with flexible provider selectio
 | **OpenRouter**      | `openrouter`      | `https://openrouter.ai/api/v1`                      | OpenAI    | [Get Key](https://openrouter.ai/keys)                            |
 | **LiteLLM Proxy**   | `litellm`         | `http://localhost:4000/v1`                          | OpenAI    | Your LiteLLM proxy key                                           |
 | **VLLM**            | `vllm`            | `http://localhost:8000/v1`                          | OpenAI    | Local                                                            |
+| **OVMS**            | `ovms`           | `http://localhost:8000/v3`                          | OpenAI    | Local (no key needed)                                            |
 | **Cerebras**        | `cerebras`        | `https://api.cerebras.ai/v1`                        | OpenAI    | [Get Key](https://cerebras.ai)                                   |
 | **VolcEngine (Doubao)** | `volcengine`  | `https://ark.cn-beijing.volces.com/api/v3`          | OpenAI    | [Get Key](https://www.volcengine.com/activity/codingplan?utm_campaign=PicoClaw&utm_content=PicoClaw&utm_medium=devrel&utm_source=OWO&utm_term=PicoClaw) |
 | **神算云**          | `shengsuanyun`    | `https://router.shengsuanyun.com/api/v1`            | OpenAI    | -                                                                |
@@ -121,7 +122,7 @@ This design also enables **multi-agent support** with flexible provider selectio
 | `model_name` | string | Yes | Unique name used to reference this model in agent config                                                                                                                                                                                    |
 | `provider` | string | No | Preferred provider identifier. When present, PicoClaw sends `model` unchanged to that provider                                                                                                                                              |
 | `model` | string | Yes | Native model ID when `provider` is set. If `provider` is omitted, the legacy `provider/model` form is still supported                                                                                                                       |
-| `api_keys` | string[] | Yes* | API key(s) for authentication. Multiple keys enable per-request rotation. Not required for local providers (Ollama, LM Studio, VLLM)                                                                                                        |
+| `api_keys` | string[] | Yes* | API key(s) for authentication. Multiple keys enable per-request rotation. Not required for local providers (Ollama, LM Studio, VLLM, OVMS)                                                                                                        |
 | `api_base` | string | No | Override the default API endpoint URL                                                                                                                                                                                                       |
 | `proxy` | string | No | HTTP proxy URL for this model entry                                                                                                                                                                                                         |
 | `user_agent` | string | No | Custom `User-Agent` header sent with API requests (supported by OpenAI-compatible, Gemini, Anthropic, and Azure providers)                                                                                                                  |
@@ -476,7 +477,7 @@ For detailed migration guide, see [migration/model-list-migration.md](../migrati
 
 PicoClaw routes providers by protocol family:
 
-- OpenAI-compatible protocol: OpenRouter, OpenAI-compatible gateways, Groq, Zhipu, and vLLM-style endpoints.
+- OpenAI-compatible protocol: OpenRouter, OpenAI-compatible gateways, Groq, Zhipu, OVMS, and vLLM-style endpoints.
 - Gemini native protocol: Google Gemini via the native `models/*:generateContent` and `models/*:streamGenerateContent` endpoints.
 - Anthropic protocol: Claude-native API behavior.
 - Codex/OAuth path: OpenAI OAuth/token authentication route.

docs/guides/providers.pt-br.md

@@ -55,6 +55,7 @@ Este design também permite **suporte multi-agente** com seleção flexível de
 | **OpenRouter**      | `openrouter/`     | `https://openrouter.ai/api/v1`                      | OpenAI    | [Get Key](https://openrouter.ai/keys)                            |
 | **LiteLLM Proxy**   | `litellm/`        | `http://localhost:4000/v1`                          | OpenAI    | Your LiteLLM proxy key                                            |
 | **VLLM**            | `vllm/`           | `http://localhost:8000/v1`                          | OpenAI    | Local                                                            |
+| **OVMS**            | `ovms/`           | `http://localhost:8000/v3`                          | OpenAI    | Local                                                            |
 | **Cerebras**        | `cerebras/`       | `https://api.cerebras.ai/v1`                        | OpenAI    | [Get Key](https://cerebras.ai)                                   |
 | **VolcEngine (Doubao)** | `volcengine/`     | `https://ark.cn-beijing.volces.com/api/v3`          | OpenAI    | [Get Key](https://www.volcengine.com/activity/codingplan?utm_campaign=PicoClaw&utm_content=PicoClaw&utm_medium=devrel&utm_source=OWO&utm_term=PicoClaw)                        |
 | **神算云**          | `shengsuanyun/`   | `https://router.shengsuanyun.com/api/v1`            | OpenAI    | -                                                                |
@@ -105,7 +106,7 @@ Este design também permite **suporte multi-agente** com seleção flexível de
 |-------|------|-------------|-----------|
 | `model_name` | string | Sim | Nome único para referenciar este modelo na config do agent |
 | `model` | string | Sim | Identificador fornecedor/modelo (ex: `openai/gpt-5.4`, `azure/gpt-5.4`, `anthropic/claude-sonnet-4.6`) |
-| `api_keys` | string[] | Sim* | Chave(s) API para autenticação. Múltiplas chaves permitem rotação por requisição. Não necessário para providers locais (Ollama, LM Studio, VLLM) |
+| `api_keys` | string[] | Sim* | Chave(s) API para autenticação. Múltiplas chaves permitem rotação por requisição. Não necessário para providers locais (Ollama, LM Studio, VLLM, OVMS) |
 | `api_base` | string | Não | Substitui a URL base da API padrão |
 | `proxy` | string | Não | URL do proxy HTTP para esta entrada de modelo |
 | `user_agent` | string | Não | Cabeçalho `User-Agent` personalizado enviado com requisições API (suportado por providers OpenAI-compatible, Gemini, Anthropic e Azure) |
@@ -298,7 +299,7 @@ Para guia de migração detalhado, veja [migration/model-list-migration.md](../m
 
 O PicoClaw roteia provedores por família de protocolo:
 
-- Protocolo compatível com OpenAI: OpenRouter, gateways compatíveis com OpenAI, Groq, Zhipu e endpoints estilo vLLM.
+- Protocolo compatível com OpenAI: OpenRouter, gateways compatíveis com OpenAI, Groq, Zhipu, endpoints estilo vLLM e OVMS.
 - Protocolo Gemini nativo: Google Gemini via endpoints nativos `models/*:generateContent` e `models/*:streamGenerateContent`.
 - Protocolo Anthropic: Comportamento nativo da API Claude.
 - Caminho Codex/OAuth: Rota de autenticação OAuth/token da OpenAI.

docs/guides/providers.vi.md

@@ -55,6 +55,7 @@ Thiết kế này cũng cho phép **hỗ trợ đa agent** với lựa chọn pr
 | **OpenRouter**      | `openrouter/`     | `https://openrouter.ai/api/v1`                      | OpenAI    | [Get Key](https://openrouter.ai/keys)                            |
 | **LiteLLM Proxy**   | `litellm/`        | `http://localhost:4000/v1`                          | OpenAI    | Your LiteLLM proxy key                                            |
 | **VLLM**            | `vllm/`           | `http://localhost:8000/v1`                          | OpenAI    | Local                                                            |
+| **OVMS**            | `ovms/`           | `http://localhost:8000/v3`                          | OpenAI    | Local                                                            |
 | **Cerebras**        | `cerebras/`       | `https://api.cerebras.ai/v1`                        | OpenAI    | [Get Key](https://cerebras.ai)                                   |
 | **VolcEngine (Doubao)** | `volcengine/`     | `https://ark.cn-beijing.volces.com/api/v3`          | OpenAI    | [Get Key](https://www.volcengine.com/activity/codingplan?utm_campaign=PicoClaw&utm_content=PicoClaw&utm_medium=devrel&utm_source=OWO&utm_term=PicoClaw)                        |
 | **神算云**          | `shengsuanyun/`   | `https://router.shengsuanyun.com/api/v1`            | OpenAI    | -                                                                |
@@ -105,7 +106,7 @@ Thiết kế này cũng cho phép **hỗ trợ đa agent** với lựa chọn pr
 |--------|------|----------|------|
 | `model_name` | string | Có | Tên duy nhất để tham chiếu model này trong cấu hình agent |
 | `model` | string | Có | Định danh nhà cung cấp/model (ví dụ: `openai/gpt-5.4`, `azure/gpt-5.4`, `anthropic/claude-sonnet-4.6`) |
-| `api_keys` | string[] | Có* | Khóa API xác thực. Nhiều khóa cho phép xoay vòng theo yêu cầu. Không cần thiết cho provider nội bộ (Ollama, LM Studio, VLLM) |
+| `api_keys` | string[] | Có* | Khóa API xác thực. Nhiều khóa cho phép xoay vòng theo yêu cầu. Không cần thiết cho provider nội bộ (Ollama, LM Studio, VLLM, OVMS) |
 | `api_base` | string | Không | Ghi đè URL endpoint API mặc định |
 | `proxy` | string | Không | URL proxy HTTP cho entry model này |
 | `user_agent` | string | Không | Header `User-Agent` tùy chỉnh gửi với yêu cầu API (được hỗ trợ bởi provider OpenAI-compatible, Gemini, Anthropic và Azure) |
@@ -298,7 +299,7 @@ Cấu hình `providers` cũ đã **bị deprecated** và đã được loại b
 
 PicoClaw định tuyến provider theo họ giao thức:
 
-- Giao thức tương thích OpenAI: OpenRouter, gateway tương thích OpenAI, Groq, Zhipu, và endpoint kiểu vLLM.
+- Giao thức tương thích OpenAI: OpenRouter, gateway tương thích OpenAI, Groq, Zhipu, endpoint kiểu vLLM và OVMS.
 - Giao thức Gemini native: Google Gemini qua các endpoint native `models/*:generateContent` và `models/*:streamGenerateContent`.
 - Giao thức Anthropic: Hành vi API native của Claude.
 - Đường dẫn Codex/OAuth: Tuyến xác thực OAuth/token của OpenAI.

docs/guides/providers.zh.md

@@ -62,6 +62,7 @@
 | **OpenRouter**      | `openrouter`      | `https://openrouter.ai/api/v1`                      | OpenAI    | [获取密钥](https://openrouter.ai/keys)                            |
 | **LiteLLM Proxy**   | `litellm`         | `http://localhost:4000/v1`                          | OpenAI    | 你的 LiteLLM 代理密钥                                             |
 | **VLLM**            | `vllm`            | `http://localhost:8000/v1`                          | OpenAI    | 本地                                                              |
+| **OVMS**            | `ovms`           | `http://localhost:8000/v3`                          | OpenAI    | 本地（无需密钥）                                                  |
 | **Cerebras**        | `cerebras`        | `https://api.cerebras.ai/v1`                        | OpenAI    | [获取密钥](https://cerebras.ai)                                   |
 | **火山引擎（Doubao）** | `volcengine`   | `https://ark.cn-beijing.volces.com/api/v3`          | OpenAI    | [获取密钥](https://www.volcengine.com/activity/codingplan?utm_campaign=PicoClaw&utm_content=PicoClaw&utm_medium=devrel&utm_source=OWO&utm_term=PicoClaw) |
 | **神算云**          | `shengsuanyun`    | `https://router.shengsuanyun.com/api/v1`            | OpenAI    | -                                                                 |
@@ -118,7 +119,7 @@
 | `model_name` | string | 是 | 在 agent 配置中引用此模型的唯一名称 |
 | `provider` | string | 否 | 推荐的 provider 标识。设置后，PicoClaw 会将 `model` 原样发送给该 provider |
 | `model` | string | 是 | 当设置 `provider` 时，这里填写 provider 原生模型 ID。若未设置 `provider`，仍兼容旧的 `provider/model` 写法 |
-| `api_keys` | string[] | 是* | 认证密钥。多个密钥可按请求轮换。本地 provider（Ollama、LM Studio、VLLM）不需要 |
+| `api_keys` | string[] | 是* | 认证密钥。多个密钥可按请求轮换。本地 provider（Ollama、LM Studio、VLLM、OVMS）不需要 |
 | `api_base` | string | 否 | 覆盖默认的 API 端点 URL |
 | `proxy` | string | 否 | 此模型条目的 HTTP 代理 URL |
 | `user_agent` | string | 否 | 自定义 `User-Agent` 请求头（支持 OpenAI 兼容、Gemini、Anthropic 和 Azure provider） |
@@ -421,7 +422,7 @@ PicoClaw 按下面的规则解析 `provider` 和最终发给上游的模型 ID
 
 PicoClaw 按协议族路由 Provider：
 
-- OpenAI 兼容协议：OpenRouter、OpenAI 兼容网关、Groq、智谱、vLLM 风格端点。
+- OpenAI 兼容协议：OpenRouter、OpenAI 兼容网关、Groq、智谱、vLLM 风格端点和 OVMS。
 - Gemini 原生协议：Google Gemini 通过原生 `models/*:generateContent` 和 `models/*:streamGenerateContent` 端点接入。
 - Anthropic 协议：Claude 原生 API 行为。
 - Codex/OAuth 路径：OpenAI OAuth/Token 认证路由。

docs/project/README.fr.md

@@ -403,13 +403,14 @@ PicoClaw supporte plus de 30 providers LLM via la configuration `model_list`. Ut
 | [Xiaomi MiMo](https://platform.xiaomimimo.com/) | `mimo/` | Requise | Modèles MiMo |
 | [Ollama](https://ollama.com/) | `ollama/` | Non requise | Modèles locaux, auto-hébergé |
 | [vLLM](https://docs.vllm.ai/) | `vllm/` | Non requise | Déploiement local, compatible OpenAI |
+| [OVMS](https://docs.openvino.ai/2026/model-server/ovms_docs_llm_quickstart.html) | `ovms/` | Non requise | Intel OpenVINO Model Server, compatible OpenAI |
 | [LiteLLM](https://docs.litellm.ai/) | `litellm/` | Variable | Proxy pour 100+ providers |
 | [Azure OpenAI](https://portal.azure.com/) | `azure/` | Requise | Déploiement Azure entreprise |
 | [GitHub Copilot](https://github.com/features/copilot) | `github-copilot/` | OAuth | Connexion par code appareil |
 | [Antigravity](https://console.cloud.google.com/) | `antigravity/` | OAuth | Google Cloud AI |
 
 <details>
-<summary><b>Déploiement local (Ollama, vLLM, etc.)</b></summary>
+<summary><b>Déploiement local (Ollama, vLLM, OVMS, etc.)</b></summary>
 
 **Ollama :**
 ```json
@@ -437,6 +438,19 @@ PicoClaw supporte plus de 30 providers LLM via la configuration `model_list`. Ut
 }
 ```
 
+**OVMS (OpenVINO Model Server) :**
+```json
+{
+  "model_list": [
+    {
+      "model_name": "local-ovms",
+      "model": "ovms/your-model",
+      "api_base": "http://localhost:8000/v3"
+    }
+  ]
+}
+```
+
 Pour les détails complets de configuration des providers, voir [Providers & Models](../guides/providers.fr.md).
 
 </details>

docs/project/README.id.md

@@ -398,13 +398,14 @@ PicoClaw mendukung 30+ provider LLM melalui konfigurasi `model_list`. Gunakan fo
 | [Xiaomi MiMo](https://platform.xiaomimimo.com/) | `mimo/` | Diperlukan | Model MiMo |
 | [Ollama](https://ollama.com/) | `ollama/` | Tidak perlu | Model lokal, self-hosted |
 | [vLLM](https://docs.vllm.ai/) | `vllm/` | Tidak perlu | Deploy lokal, kompatibel OpenAI |
+| [OVMS](https://docs.openvino.ai/2026/model-server/ovms_docs_llm_quickstart.html) | `ovms/` | Tidak perlu | Intel OpenVINO Model Server, kompatibel OpenAI |
 | [LiteLLM](https://docs.litellm.ai/) | `litellm/` | Bervariasi | Proxy untuk 100+ provider |
 | [Azure OpenAI](https://portal.azure.com/) | `azure/` | Diperlukan | Deploy Azure enterprise |
 | [GitHub Copilot](https://github.com/features/copilot) | `github-copilot/` | OAuth | Login dengan device code |
 | [Antigravity](https://console.cloud.google.com/) | `antigravity/` | OAuth | Google Cloud AI |
 
 <details>
-<summary><b>Deploy lokal (Ollama, vLLM, dll.)</b></summary>
+<summary><b>Deploy lokal (Ollama, vLLM, OVMS, dll.)</b></summary>
 
 **Ollama:**
 ```json
@@ -432,6 +433,19 @@ PicoClaw mendukung 30+ provider LLM melalui konfigurasi `model_list`. Gunakan fo
 }
 ```
 
+**OVMS (OpenVINO Model Server):**
+```json
+{
+  "model_list": [
+    {
+      "model_name": "local-ovms",
+      "model": "ovms/your-model",
+      "api_base": "http://localhost:8000/v3"
+    }
+  ]
+}
+```
+
 Untuk detail konfigurasi provider lengkap, lihat [Providers & Models](../guides/providers.md).
 
 </details>

docs/project/README.it.md

@@ -398,13 +398,14 @@ PicoClaw supporta 30+ provider LLM tramite la configurazione `model_list`. Usa i
 | [Xiaomi MiMo](https://platform.xiaomimimo.com/) | `mimo/` | Richiesta | Modelli MiMo |
 | [Ollama](https://ollama.com/) | `ollama/` | Non necessaria | Modelli locali, self-hosted |
 | [vLLM](https://docs.vllm.ai/) | `vllm/` | Non necessaria | Deploy locale, compatibile OpenAI |
+| [OVMS](https://docs.openvino.ai/2026/model-server/ovms_docs_llm_quickstart.html) | `ovms/` | Non necessaria | Intel OpenVINO Model Server, compatibile OpenAI |
 | [LiteLLM](https://docs.litellm.ai/) | `litellm/` | Variabile | Proxy per 100+ provider |
 | [Azure OpenAI](https://portal.azure.com/) | `azure/` | Richiesta | Deploy Azure enterprise |
 | [GitHub Copilot](https://github.com/features/copilot) | `github-copilot/` | OAuth | Login con device code |
 | [Antigravity](https://console.cloud.google.com/) | `antigravity/` | OAuth | Google Cloud AI |
 
 <details>
-<summary><b>Deploy locale (Ollama, vLLM, ecc.)</b></summary>
+<summary><b>Deploy locale (Ollama, vLLM, OVMS, ecc.)</b></summary>
 
 **Ollama:**
 ```json
@@ -432,6 +433,19 @@ PicoClaw supporta 30+ provider LLM tramite la configurazione `model_list`. Usa i
 }
 ```
 
+**OVMS (OpenVINO Model Server):**
+```json
+{
+  "model_list": [
+    {
+      "model_name": "local-ovms",
+      "model": "ovms/your-model",
+      "api_base": "http://localhost:8000/v3"
+    }
+  ]
+}
+```
+
 Per i dettagli completi sulla configurazione dei provider, vedi [Provider & Modelli](../guides/providers.md).
 
 </details>

docs/project/README.ja.md

@@ -399,13 +399,14 @@ PicoClaw は `model_list` 設定を通じて 30 以上の LLM Provider をサポ
 | [Xiaomi MiMo](https://platform.xiaomimimo.com/) | `mimo/` | 必須 | MiMo モデル |
 | [Ollama](https://ollama.com/) | `ollama/` | 不要 | ローカルモデル、セルフホスト |
 | [vLLM](https://docs.vllm.ai/) | `vllm/` | 不要 | ローカルデプロイ、OpenAI 互換 |
+| [OVMS](https://docs.openvino.ai/2026/model-server/ovms_docs_llm_quickstart.html) | `ovms/` | 不要 | Intel OpenVINO Model Server、OpenAI 互換 |
 | [LiteLLM](https://docs.litellm.ai/) | `litellm/` | 場合による | 100 以上の Provider のプロキシ |
 | [Azure OpenAI](https://portal.azure.com/) | `azure/` | 必須 | エンタープライズ Azure デプロイ |
 | [GitHub Copilot](https://github.com/features/copilot) | `github-copilot/` | OAuth | デバイスコードログイン |
 | [Antigravity](https://console.cloud.google.com/) | `antigravity/` | OAuth | Google Cloud AI |
 
 <details>
-<summary><b>ローカルデプロイ（Ollama、vLLM など）</b></summary>
+<summary><b>ローカルデプロイ（Ollama、vLLM、OVMS など）</b></summary>
 
 **Ollama:**
 ```json
@@ -433,6 +434,19 @@ PicoClaw は `model_list` 設定を通じて 30 以上の LLM Provider をサポ
 }
 ```
 
+**OVMS (OpenVINO Model Server):**
+```json
+{
+  "model_list": [
+    {
+      "model_name": "local-ovms",
+      "model": "ovms/your-model",
+      "api_base": "http://localhost:8000/v3"
+    }
+  ]
+}
+```
+
 Provider の完全な設定詳細は [Provider とモデル](../guides/providers.ja.md) を参照してください。
 
 </details>

docs/project/README.ms.md

@@ -394,6 +394,7 @@ PicoClaw menyokong 30+ penyedia LLM melalui konfigurasi `model_list`. Gunakan fo
 | [Xiaomi MiMo](https://platform.xiaomimimo.com/) | `mimo/` | Diperlukan | Model MiMo |
 | [Ollama](https://ollama.com/) | `ollama/` | Tidak perlu | Model tempatan, self-hosted |
 | [vLLM](https://docs.vllm.ai/) | `vllm/` | Tidak perlu | Deployment tempatan, serasi OpenAI |
+| [OVMS](https://docs.openvino.ai/2026/model-server/ovms_docs_llm_quickstart.html) | `ovms/` | Tidak perlu | Intel OpenVINO Model Server, serasi OpenAI |
 | [LiteLLM](https://docs.litellm.ai/) | `litellm/` | Berbeza | Proksi untuk 100+ penyedia |
 | [Azure OpenAI](https://portal.azure.com/) | `azure/` | Diperlukan | Deployment Azure perusahaan |
 | [GitHub Copilot](https://github.com/features/copilot) | `github-copilot/` | OAuth | Log masuk kod peranti |
@@ -403,7 +404,7 @@ PicoClaw menyokong 30+ penyedia LLM melalui konfigurasi `model_list`. Gunakan fo
 > \* AWS Bedrock memerlukan tag binaan: `go build -tags bedrock`. Tetapkan `api_base` kepada nama rantau (cth. `us-east-1`) untuk resolusi endpoint automatik merentasi semua partition AWS. Apabila menggunakan URL endpoint penuh, anda juga perlu mengkonfigurasi `AWS_REGION` melalui pemboleh ubah persekitaran.
 
 <details>
-<summary><b>Deployment tempatan (Ollama, vLLM, dll.)</b></summary>
+<summary><b>Deployment tempatan (Ollama, vLLM, OVMS, dll.)</b></summary>
 
 **Ollama:**
 ```json
@@ -431,6 +432,19 @@ PicoClaw menyokong 30+ penyedia LLM melalui konfigurasi `model_list`. Gunakan fo
 }
 ```
 
+**OVMS (OpenVINO Model Server):**
+```json
+{
+  "model_list": [
+    {
+      "model_name": "local-ovms",
+      "model": "ovms/your-model",
+      "api_base": "http://localhost:8000/v3"
+    }
+  ]
+}
+```
+
 Untuk butiran konfigurasi penyedia penuh, lihat [Penyedia & Model](../guides/providers.md).
 
 </details>

docs/project/README.pt-br.md

@@ -399,13 +399,14 @@ O PicoClaw suporta mais de 30 providers de LLM através da configuração `model
 | [Xiaomi MiMo](https://platform.xiaomimimo.com/) | `mimo/` | Obrigatória | Modelos MiMo |
 | [Ollama](https://ollama.com/) | `ollama/` | Não necessária | Modelos locais, self-hosted |
 | [vLLM](https://docs.vllm.ai/) | `vllm/` | Não necessária | Implantação local, compatível com OpenAI |
+| [OVMS](https://docs.openvino.ai/2026/model-server/ovms_docs_llm_quickstart.html) | `ovms/` | Não necessária | Intel OpenVINO Model Server, compatível com OpenAI |
 | [LiteLLM](https://docs.litellm.ai/) | `litellm/` | Varia | Proxy para 100+ providers |
 | [Azure OpenAI](https://portal.azure.com/) | `azure/` | Obrigatória | Implantação Azure Enterprise |
 | [GitHub Copilot](https://github.com/features/copilot) | `github-copilot/` | OAuth | Login por código de dispositivo |
 | [Antigravity](https://console.cloud.google.com/) | `antigravity/` | OAuth | Google Cloud AI |
 
 <details>
-<summary><b>Implantação local (Ollama, vLLM, etc.)</b></summary>
+<summary><b>Implantação local (Ollama, vLLM, OVMS, etc.)</b></summary>
 
 **Ollama:**
 ```json
@@ -433,6 +434,19 @@ O PicoClaw suporta mais de 30 providers de LLM através da configuração `model
 }
 ```
 
+**OVMS (OpenVINO Model Server):**
+```json
+{
+  "model_list": [
+    {
+      "model_name": "local-ovms",
+      "model": "ovms/your-model",
+      "api_base": "http://localhost:8000/v3"
+    }
+  ]
+}
+```
+
 Para detalhes completos de configuração de providers, veja [Providers & Models](../guides/providers.pt-br.md).
 
 </details>

docs/project/README.vi.md

@@ -399,13 +399,14 @@ PicoClaw hỗ trợ 30+ Provider LLM thông qua cấu hình `model_list`. Sử d
 | [Xiaomi MiMo](https://platform.xiaomimimo.com/) | `mimo/` | Bắt buộc | Mô hình MiMo |
 | [Ollama](https://ollama.com/) | `ollama/` | Không cần | Mô hình cục bộ, tự lưu trữ |
 | [vLLM](https://docs.vllm.ai/) | `vllm/` | Không cần | Triển khai cục bộ, tương thích OpenAI |
+| [OVMS](https://docs.openvino.ai/2026/model-server/ovms_docs_llm_quickstart.html) | `ovms/` | Không cần | Intel OpenVINO Model Server, tương thích OpenAI |
 | [LiteLLM](https://docs.litellm.ai/) | `litellm/` | Tùy | Proxy cho 100+ provider |
 | [Azure OpenAI](https://portal.azure.com/) | `azure/` | Bắt buộc | Triển khai Azure doanh nghiệp |
 | [GitHub Copilot](https://github.com/features/copilot) | `github-copilot/` | OAuth | Đăng nhập bằng device code |
 | [Antigravity](https://console.cloud.google.com/) | `antigravity/` | OAuth | Google Cloud AI |
 
 <details>
-<summary><b>Triển khai cục bộ (Ollama, vLLM, v.v.)</b></summary>
+<summary><b>Triển khai cục bộ (Ollama, vLLM, OVMS, v.v.)</b></summary>
 
 **Ollama:**
 ```json
@@ -433,6 +434,19 @@ PicoClaw hỗ trợ 30+ Provider LLM thông qua cấu hình `model_list`. Sử d
 }
 ```
 
+**OVMS (OpenVINO Model Server):**
+```json
+{
+  "model_list": [
+    {
+      "model_name": "local-ovms",
+      "model": "ovms/your-model",
+      "api_base": "http://localhost:8000/v3"
+    }
+  ]
+}
+```
+
 Để biết chi tiết cấu hình provider đầy đủ, xem [Providers & Models](../guides/providers.vi.md).
 
 </details>

docs/project/README.zh.md

@@ -399,13 +399,14 @@ PicoClaw 通过 `model_list` 配置支持 30+ LLM Provider，使用 `协议/模
 | [小米 MiMo](https://platform.xiaomimimo.com/) | `mimo/` | 必填 | MiMo 系列模型 |
 | [Ollama](https://ollama.com/) | `ollama/` | 无需 | 本地模型，自托管 |
 | [vLLM](https://docs.vllm.ai/) | `vllm/` | 无需 | 本地部署，兼容 OpenAI |
+| [OVMS](https://docs.openvino.ai/2026/model-server/ovms_docs_llm_quickstart.html) | `ovms/` | 无需 | Intel OpenVINO Model Server，兼容 OpenAI |
 | [LiteLLM](https://docs.litellm.ai/) | `litellm/` | 视情况 | 100+ Provider 代理 |
 | [Azure OpenAI](https://portal.azure.com/) | `azure/` | 必填 | 企业级 Azure 部署 |
 | [GitHub Copilot](https://github.com/features/copilot) | `github-copilot/` | OAuth | 设备码登录 |
 | [Antigravity](https://console.cloud.google.com/) | `antigravity/` | OAuth | Google Cloud AI |
 
 <details>
-<summary><b>本地部署（Ollama、vLLM 等）</b></summary>
+<summary><b>本地部署（Ollama、vLLM、OVMS 等）</b></summary>
 
 **Ollama:**
 ```json
@@ -433,6 +434,19 @@ PicoClaw 通过 `model_list` 配置支持 30+ LLM Provider，使用 `协议/模
 }
 ```
 
+**OVMS (OpenVINO Model Server):**
+```json
+{
+  "model_list": [
+    {
+      "model_name": "local-ovms",
+      "model": "ovms/your-model",
+      "api_base": "http://localhost:8000/v3"
+    }
+  ]
+}
+```
+
 完整 Provider 配置详情请参阅 [Providers & Models](../guides/providers.zh.md)。
 
 </details>

pkg/audio/asr/asr.go

@@ -26,7 +26,7 @@ func supportsAudioTranscription(modelCfg *config.ModelConfig) bool {
 	case "openai", "azure", "azure-openai",
 		"litellm", "openrouter", "groq", "zhipu", "gemini", "nvidia",
 		"ollama", "moonshot", "shengsuanyun", "deepseek", "cerebras",
-		"vivgrid", "volcengine", "vllm", "qwen", "qwen-portal", "qwen-intl", "qwen-international", "dashscope-intl",
+		"vivgrid", "volcengine", "vllm", "ovms", "qwen", "qwen-portal", "qwen-intl", "qwen-international", "dashscope-intl",
 		"qwen-us", "dashscope-us", "mistral", "avian", "minimax", "longcat", "modelscope", "novita",
 		"coding-plan", "alibaba-coding", "qwen-coding", "zai":
 		// These protocols all go through the OpenAI-compatible or Azure provider path in
@@ -47,7 +47,7 @@ func supportsWhisperTranscription(modelCfg *config.ModelConfig) bool {
 	switch protocol {
 	case "openai", "litellm", "openrouter", "groq", "zhipu", "gemini", "nvidia",
 		"ollama", "moonshot", "shengsuanyun", "deepseek", "cerebras",
-		"vivgrid", "volcengine", "vllm", "qwen", "qwen-portal", "qwen-intl", "qwen-international", "dashscope-intl",
+		"vivgrid", "volcengine", "vllm", "ovms", "qwen", "qwen-portal", "qwen-intl", "qwen-international", "dashscope-intl",
 		"qwen-us", "dashscope-us", "mistral", "avian", "minimax", "longcat", "modelscope", "novita",
 		"coding-plan", "alibaba-coding", "qwen-coding", "zai", "mimo":
 		return true

pkg/audio/asr/asr_test.go

@@ -206,6 +206,35 @@ func TestDetectTranscriber(t *testing.T) {
 			},
 			wantName: "audio-model",
 		},
+		{
+			name: "ovms voice model name selects audio model transcriber",
+			cfg: &config.Config{
+				Voice: config.VoiceConfig{ModelName: "local-ovms"},
+				ModelList: []*config.ModelConfig{
+					{
+						ModelName: "local-ovms",
+						Model:     "ovms/custom-model",
+						APIBase:   "http://localhost:8000/v3",
+					},
+				},
+			},
+			wantName: "audio-model",
+		},
+		{
+			name: "ovms voice model name with whisper model selects whisper transcriber",
+			cfg: &config.Config{
+				Voice: config.VoiceConfig{ModelName: "local-ovms"},
+				ModelList: []*config.ModelConfig{
+					{
+						ModelName: "local-ovms",
+						Model:     "ovms/whisper-large-v3",
+						APIBase:   "http://localhost:8000/v3",
+						APIKeys:   config.SimpleSecureStrings("sk-ovms-test"),
+					},
+				},
+			},
+			wantName: "whisper",
+		},
 	}
 
 	for _, tc := range tests {

pkg/config/defaults.go

@@ -273,6 +273,14 @@ func DefaultConfig() *Config {
 				APIBase:   "http://localhost:8000/v1",
 			},
 
+			// OVMS (local) - http://localhost:8000
+			{
+				ModelName: "local-ovms",
+				Provider:  "ovms",
+				Model:     "custom-model",
+				APIBase:   "http://localhost:8000/v3",
+			},
+
 			// LM Studio (local) - http://localhost:1234
 			{
 				ModelName: "lmstudio-local",

pkg/providers/factory_provider.go

@@ -54,6 +54,7 @@ var protocolMetaByName = map[string]protocolMeta{
 	"alibaba-coding-anthropic": {defaultAPIBase: "https://coding-intl.dashscope.aliyuncs.com/apps/anthropic"},
 	"zai":                      {defaultAPIBase: "https://api.z.ai/api/coding/paas/v4"},
 	"vllm":                     {defaultAPIBase: "http://localhost:8000/v1", emptyAPIKeyAllowed: true},
+	"ovms":                     {defaultAPIBase: "http://localhost:8000/v3", emptyAPIKeyAllowed: true},
 	"mistral":                  {defaultAPIBase: "https://api.mistral.ai/v1"},
 	"avian":                    {defaultAPIBase: "https://api.avian.io/v1"},
 	"minimax":                  {defaultAPIBase: "https://api.minimaxi.com/v1"},
@@ -248,7 +249,7 @@ func CreateProviderFromConfig(cfg *config.ModelConfig) (LLMProvider, string, err
 
 	case "litellm", "lmstudio", "openrouter", "groq", "zhipu", "nvidia", "venice",
 		"ollama", "moonshot", "shengsuanyun", "deepseek", "cerebras",
-		"vivgrid", "volcengine", "vllm", "qwen", "qwen-portal", "qwen-intl", "qwen-international", "dashscope-intl",
+		"vivgrid", "volcengine", "vllm", "ovms", "qwen", "qwen-portal", "qwen-intl", "qwen-international", "dashscope-intl",
 		"qwen-us", "dashscope-us", "mistral", "avian", "longcat", "modelscope", "novita",
 		"coding-plan", "alibaba-coding", "qwen-coding", "zai", "mimo":
 		// All other OpenAI-compatible HTTP providers

pkg/providers/factory_provider_test.go

@@ -199,6 +199,7 @@ func TestCreateProviderFromConfig_DefaultAPIBase(t *testing.T) {
 		{"vivgrid", "vivgrid"},
 		{"qwen", "qwen"},
 		{"vllm", "vllm"},
+		{"ovms", "ovms"},
 		{"deepseek", "deepseek"},
 		{"ollama", "ollama"},
 		{"lmstudio", "lmstudio"},
@@ -316,6 +317,20 @@ func TestCreateProviderFromConfig_LocalProviders(t *testing.T) {
 			apiKey:      "",
 			wantModelID: "Qwen/Qwen3-8B",
 		},
+		{
+			name:        "OVMS with API key",
+			modelName:   "test-ovms",
+			model:       "ovms/llama3",
+			apiKey:      "test-key",
+			wantModelID: "llama3",
+		},
+		{
+			name:        "OVMS without API key",
+			modelName:   "test-ovms",
+			model:       "ovms/llama3",
+			apiKey:      "",
+			wantModelID: "llama3",
+		},
 	}
 
 	for _, tt := range tests {

web/README.md

@@ -341,7 +341,7 @@ Check these in the dashboard:
 
 - a default model is selected
 - the model has credentials or OAuth state
-- local models such as Ollama or vLLM are reachable
+- local models such as Ollama, vLLM, or OVMS are reachable
 
 ### The launcher cannot find `picoclaw`
 

web/backend/api/model_status.go

@@ -176,7 +176,7 @@ func runLocalModelProbe(m *config.ModelConfig) bool {
 	switch protocol {
 	case "ollama":
 		return probeOllamaModelFunc(apiBase, modelID)
-	case "vllm", "lmstudio":
+	case "vllm", "ovms", "lmstudio":
 		return probeOpenAICompatibleModelFunc(apiBase, modelID, m.APIKey())
 	case "github-copilot", "copilot":
 		return probeTCPServiceFunc(apiBase)

web/backend/api/model_status_test.go

@@ -89,6 +89,56 @@ func TestProbeLocalModelAvailability_LMStudioUsesOpenAICompatibleProbe(t *testin
 	}
 }
 
+func TestRequiresRuntimeProbe_OVMS(t *testing.T) {
+	if !requiresRuntimeProbe(&config.ModelConfig{
+		Model: "ovms/llama3",
+	}) {
+		t.Fatal("requiresRuntimeProbe(ovms with default base) = false, want true")
+	}
+
+	if requiresRuntimeProbe(&config.ModelConfig{
+		Model:   "ovms/llama3",
+		APIBase: "https://api.example.com/v3",
+	}) {
+		t.Fatal("requiresRuntimeProbe(ovms with remote base) = true, want false")
+	}
+}
+
+func TestModelProbeAPIBase_OVMSDefault(t *testing.T) {
+	got := modelProbeAPIBase(&config.ModelConfig{Model: "ovms/llama3"})
+	if got != "http://localhost:8000/v3" {
+		t.Fatalf("modelProbeAPIBase(ovms) = %q, want %q", got, "http://localhost:8000/v3")
+	}
+}
+
+func TestProbeLocalModelAvailability_OVMSUsesOpenAICompatibleProbe(t *testing.T) {
+	originalProbe := probeOpenAICompatibleModelFunc
+	defer func() { probeOpenAICompatibleModelFunc = originalProbe }()
+
+	called := false
+	probeOpenAICompatibleModelFunc = func(apiBase, modelID, apiKey string) bool {
+		called = true
+		if apiBase != "http://localhost:8000/v3" {
+			t.Fatalf("apiBase = %q, want %q", apiBase, "http://localhost:8000/v3")
+		}
+		if modelID != "llama3" {
+			t.Fatalf("modelID = %q, want %q", modelID, "llama3")
+		}
+		if apiKey != "" {
+			t.Fatalf("apiKey = %q, want empty", apiKey)
+		}
+		return true
+	}
+
+	model := &config.ModelConfig{Model: "ovms/llama3"}
+	if !probeLocalModelAvailability(model) {
+		t.Fatal("probeLocalModelAvailability(ovms) = false, want true")
+	}
+	if !called {
+		t.Fatal("probeOpenAICompatibleModelFunc was not called for ovms")
+	}
+}
+
 func TestModelProbeCacheKey_DifferentAPIKeysProduceDifferentKeys(t *testing.T) {
 	base := &config.ModelConfig{
 		ModelName:   "local-vllm",

web/frontend/src/components/models/models-page.tsx

@@ -42,9 +42,10 @@ const PROVIDER_PRIORITY: Record<string, number> = {
   azure: 23,
   ollama: 24,
   vllm: 25,
-  lmstudio: 26,
+  ovms: 26,
-  zai: 27,
+  lmstudio: 27,
-  mimo: 28,
+  zai: 28,
+  mimo: 29,
 }
 
 interface ProviderGroup {

web/frontend/src/components/models/provider-icon.tsx

@@ -41,6 +41,7 @@ const PROVIDER_DOMAINS: Record<string, string> = {
   mistral: "mistral.ai",
   avian: "avian.io",
   vllm: "vllm.ai",
+  ovms: "docs.openvino.ai",
   zhipu: "zhipuai.cn",
   zai: "z.ai",
   mimo: "xiaomi.com",

web/frontend/src/components/models/provider-label.ts

@@ -20,6 +20,7 @@ const PROVIDER_LABELS: Record<string, string> = {
   mistral: "Mistral AI",
   avian: "Avian",
   vllm: "VLLM (local)",
+  ovms: "OVMS (local)",
   zhipu: "Zhipu AI (智谱)",
   zai: "Z.ai",
   mimo: "Xiaomi MiMo",