Sakrylle API

English

简体中文

English

简体中文

Appearance

NextChat ​

Introduction ​

NextChat is a lightweight, cross-platform open-source AI chat frontend covering Web, desktop (Tauri: Windows / macOS / Linux), iOS, and Android. It supports the OpenAI, Anthropic, Google, and other protocols, and lets you override the default API endpoint in Settings — a very convenient way to plug in the Sakrylle API.

Prerequisites ​

  1. A NextChat entry point: the official NextChatAI, a Vercel/Docker self-hosted build, or a downloaded desktop/mobile app.
  2. An account at https://ai1.sakrylle.com with sk-xxxxxxxxxxxxxxxx generated per the API Keys guide.
  3. Model IDs ready: e.g. gpt-5.6-sol, claude-haiku-4-5-20251001.

Override in the UI ​

If you just want to try out the Sakrylle API on a single device, override the endpoint and key directly in NextChat — no redeploy needed:

  1. Open NextChat and tap Settings (gear icon) at the bottom-left.

  2. Scroll to the Custom Endpoint toggle and turn it on.

  3. Set Model Provider to OpenAI.

  4. Fill in per this table:

    FieldValue
    OpenAI Endpoint / API Hosthttps://api.sakrylle.com
    OpenAI API Keysk-xxxxxxxxxxxxxxxx
    Custom Models+gpt-5.6-sol,+claude-haiku-4-5-20251001

    Notes:

    • For "Endpoint", enter only the root domain https://api.sakrylle.com. NextChat appends /v1/chat/completions automatically.
    • "Custom Models" uses +modelID to add models to the picker; separate multiple models with commas. The -all prefix clears the existing list (e.g. -all,+gpt-5.6-sol).

  5. Close the settings panel, return to the chat view, tap the model picker at the top right, and switch the current model to gpt-5.6-sol or claude-haiku-4-5-20251001.

  6. Send a message to verify connectivity.

To use the Anthropic native protocol, switch Model Provider to Anthropic, set Anthropic API Endpoint to https://api.sakrylle.com, and use the same Sakrylle key for the API Key; add +claude-haiku-4-5-20251001 to Custom Models.

Self-host env ​

If you self-host NextChat for a team via Vercel / Docker, the recommended way is to fix the Sakrylle connection through environment variables at deploy time:

bash
docker run -d -p 3000:3000 \
  -e OPENAI_API_KEY=sk-xxxxxxxxxxxxxxxx \
  -e BASE_URL=https://api.sakrylle.com \
  -e CUSTOM_MODELS="-all,+gpt-5.6-sol,+claude-haiku-4-5-20251001,+gpt-5.4,+claude-sonnet-4-6" \
  -e DEFAULT_MODEL=gpt-5.6-sol \
  -e CODE=set-a-strong-access-code \
  yidadaa/chatgpt-next-web

Key variables:

  • OPENAI_API_KEY: your Sakrylle API Key.
  • BASE_URL: Sakrylle gateway root https://api.sakrylle.com.
  • CUSTOM_MODELS: clear defaults with -all, then use + to add the model IDs you want exposed to the frontend.
  • DEFAULT_MODEL: the model selected by default when entering a session.
  • CODE: access password — strongly recommend setting a 7+ character password, otherwise anyone with access to the domain can spend your quota.

Redeploy after changes for them to take effect.

FAQ ​

  • Only one or two models in the picker: CUSTOM_MODELS used -all but forgot to + the models back. Re-check the variable.
  • All requests return 401: confirm BASE_URL is not https://api.sakrylle.com/v1 (NextChat appends /v1 automatically; doubling produces /v1/v1/... errors), and check API Key validity.
  • Streaming hangs: in self-hosted setups, check whether the reverse proxy buffers SSE. Nginx needs proxy_buffering off; proxy_read_timeout 600s;.

Official Documentation ​