feat: add EmpirioLabs LLM provider

[Adam-Dalloul]

What this does

Adds EmpirioLabs (https://empiriolabs.ai) as a built-in LLM provider.

EmpirioLabs is an OpenAI-compatible, multi-model API that hosts open and proprietary models (Qwen, DeepSeek, GLM, Kimi, MiniMax, Gemma, and more) behind a single base URL with pay-as-you-go pricing. It exposes:

• Chat Completions at https://api.empiriolabs.ai/v1 (OpenAI-compatible, streaming supported)
• GET /v1/models for the live model catalog, so the existing dynamic model-listing pattern works as-is
• API keys from https://platform.empiriolabs.ai/dashboard/api-keys

Users could already reach EmpirioLabs through the Generic OpenAI connector, but a named provider with a logo, onboarding entry, and data-handling disclosure makes it discoverable and removes base-URL setup friction, the same way the CometAPI integration (#4379) did.

Resolves #5814.

Approach

This mirrors the CometAPI integration (#4379) file-for-file, following the current in-tree CometAPI shape:

Server

• server/utils/AiProviders/empiriolabs/index.js (LLM class, OpenAI-compatible, cached /v1/models listing)
• server/utils/AiProviders/empiriolabs/constants.js (filters the catalog down to chat-capable models)
• server/utils/agents/aibitat/providers/empiriolabs.js (agent provider with native tool calling + UnTooled fallback)
• Registered in server/utils/helpers/index.js, server/utils/helpers/customModels.js, server/utils/helpers/updateENV.js, server/models/systemSettings.js, server/utils/agents/index.js, server/utils/agents/aibitat/index.js, and server/utils/agents/aibitat/providers/index.js and ai-provider.js

Frontend

• frontend/src/components/LLMSelection/EmpirioLabsOptions/index.jsx (API key + dynamic model selection + stream-timeout advanced control)
• Registered in GeneralSettings/LLMPreference, both OnboardingFlow steps (LLM preference + data-handling privacy disclosure via ProviderPrivacy/constants.js), and WorkspaceSettings/AgentConfig/AgentLLMSelection
• frontend/src/media/llmprovider/empiriolabs.png (logo)

Config / docs

• Env blocks in docker/.env.example and server/.env.example
• Provider rows in README.md, locales/README.ja-JP.md, locales/README.zh-CN.md
• Storage cache dir in server/storage/models/.gitignore

Env vars

LLM_PROVIDER='empiriolabs'
EMPIRIOLABS_API_KEY='your-empiriolabs-api-key-here'
EMPIRIOLABS_MODEL_PREF='qwen3-7-plus'
EMPIRIOLABS_LLM_TIMEOUT_MS=3000 # Optional; stream idle timeout in ms (min 500ms)

Notes

• Prettier-clean against the repo .prettierrc; all new server modules pass node --check.
• The agent provider supports native OpenAI-style tool calling (EmpirioLabs serves tool-capable models such as DeepSeek, GLM, and Kimi), with the standard UnTooled fallback for models without tool support.