From 9d1faca4cc8eb68362a3ce8ceecd082a52c70f57 Mon Sep 17 00:00:00 2001 From: hei Date: Wed, 15 Jul 2026 01:03:07 +0800 Subject: [PATCH] docs(admins): add DaoXE OpenAI-compatible LLM gateway guide Document configuring DaoXE via Onyx's OpenAI-Compatible provider card (API base https://daoxe.com/v1, account-scoped model IDs, multi-protocol context). Wire the page into Language Models navigation and overview. Co-Authored-By: Claude Fable 5 --- .../ai_models/custom_inference_provider.mdx | 6 ++ admins/ai_models/daoxe.mdx | 89 +++++++++++++++++++ admins/ai_models/overview.mdx | 10 ++- docs.json | 1 + 4 files changed, 103 insertions(+), 3 deletions(-) create mode 100644 admins/ai_models/daoxe.mdx diff --git a/admins/ai_models/custom_inference_provider.mdx b/admins/ai_models/custom_inference_provider.mdx index dac135ac..a89c3f56 100644 --- a/admins/ai_models/custom_inference_provider.mdx +++ b/admins/ai_models/custom_inference_provider.mdx @@ -15,6 +15,12 @@ you can configure a custom inference provider. Your custom provider must provide OpenAI-compatible API endpoints. + + For a hosted multi-model gateway example that uses Onyx's built-in + **OpenAI-Compatible** card (API base + key + fetch models), see + [DaoXE](/admins/ai_models/daoxe). + + Determine your provider's API base URL. diff --git a/admins/ai_models/daoxe.mdx b/admins/ai_models/daoxe.mdx new file mode 100644 index 00000000..55f84fdb --- /dev/null +++ b/admins/ai_models/daoxe.mdx @@ -0,0 +1,89 @@ +--- +title: "DaoXE" +description: "Configure DaoXE as an OpenAI-compatible multi-model gateway for Onyx" +icon: "server" +--- + +import NavigateToAiModels from "/snippets/navigate-to-ai-models.mdx"; +import StandardAiModelConfigs from "/snippets/standard-ai-model-configs.mdx"; + +## Guide + +Configure Onyx to use models served through [DaoXE](https://daoxe.com), a multi-model, +multi-protocol API gateway. + +Onyx talks to DaoXE over the **OpenAI-compatible** path (Chat Completions and +`/v1/models`). DaoXE also exposes other client protocols outside Onyx — including +OpenAI Responses and Anthropic Messages — so the same account can serve different +tools; this page only covers the Onyx admin setup. + + + DaoXE model IDs are **account-scoped** and change over time. Prefer **Fetch Available + Models** in the Onyx UI (or copy IDs from your DaoXE account catalog). Do not hardcode a + static public model list. + + + + DaoXE is not a built-in Onyx provider card. Use **OpenAI-Compatible** under + **Self-hosted & Custom** and point the API base at DaoXE. + + + + + Sign in at [daoxe.com](https://daoxe.com) and create an API key for your workspace. + + Confirm which chat models are visible in your account catalog. Those model IDs are + what Onyx will call after import. + + + DaoXE does not serve mainland China. If `https://daoxe.com` is unreachable from your + network, use the regional endpoint documented on the DaoXE site rather than guessing + an alternate host. + + + + + + + Under **Self-hosted & Custom**, select **OpenAI-Compatible**. + + Give the provider a **Display Name** (for example, `DaoXE`). + + Set **API Base URL** to: + + ```text + https://daoxe.com/v1 + ``` + + Paste your DaoXE **API Key**. + + Click **Fetch Available Models** so Onyx can load the language models currently + exposed by your DaoXE account. + + + + Onyx stores the model IDs returned by DaoXE's OpenAI-compatible `/v1/models` endpoint + and makes those models available to users you grant access to. + + Keep the visible set small: pick the chat models you actually want for agents and + RAG rather than every ID in the catalog. + + + For graph extraction, tool calling, or long agent loops, prefer higher-capability + chat models from your DaoXE catalog. Smaller models may work for simple chat but can + be less reliable on structured outputs. + + + + + + +## Notes + +- **Base URL:** use `https://daoxe.com/v1` (include the `/v1` suffix). +- **Auth:** `Authorization: Bearer `. +- **Protocols:** Onyx uses OpenAI-compatible Chat Completions through this integration. + DaoXE separately supports OpenAI Responses and Anthropic Messages for other clients; + configure those outside Onyx when needed. +- **Examples:** OpenAI SDK samples for DaoXE live at + [seven7763/DaoXE-AI](https://github.com/seven7763/DaoXE-AI). diff --git a/admins/ai_models/overview.mdx b/admins/ai_models/overview.mdx index ff0c9343..7ac635e6 100644 --- a/admins/ai_models/overview.mdx +++ b/admins/ai_models/overview.mdx @@ -34,7 +34,8 @@ which models are visible, and which models should be the default or fast option These providers expose language models through a broader cloud or routing layer. Common choices include **Azure OpenAI**, **Amazon Bedrock**, **Google Vertex AI**, **OpenRouter**, - **LiteLLM Proxy**, and **Bifrost**. They are useful when you need enterprise controls, cloud alignment, + **LiteLLM Proxy**, **Bifrost**, and hosted multi-model gateways such as **DaoXE** (via the + OpenAI-Compatible provider). They are useful when you need enterprise controls, cloud alignment, regional hosting options, or access to multiple model families from one integration point. @@ -65,8 +66,9 @@ which models are visible, and which models should be the default or fast option - Start with one primary provider for most users. - Use a recent **GPT**, **Claude**, or **Gemini** family model as your default if you want the strongest out-of-the-box experience. -- Use **Bedrock**, **Vertex AI**, **Azure OpenAI**, **OpenRouter**, **LiteLLM Proxy**, or **Bifrost** when procurement, - routing, or cloud alignment matters more than a direct vendor integration. +- Use **Bedrock**, **Vertex AI**, **Azure OpenAI**, **OpenRouter**, **LiteLLM Proxy**, **Bifrost**, or an + OpenAI-compatible gateway such as [DaoXE](/admins/ai_models/daoxe) when procurement, routing, or cloud + alignment matters more than a direct vendor integration. - Use open-weight families such as **Llama**, **Qwen**, **DeepSeek**, or **gpt-oss** if you are self-hosting. - Keep the visible model list short so users are choosing between a few intentional options instead of every possible version. @@ -95,6 +97,8 @@ which models are visible, and which models should be the default or fast option + + diff --git a/docs.json b/docs.json index 4f6158b2..b6fc9167 100644 --- a/docs.json +++ b/docs.json @@ -176,6 +176,7 @@ "admins/ai_models/openrouter", "admins/ai_models/bifrost", "admins/ai_models/litellm_proxy", + "admins/ai_models/daoxe", "admins/ai_models/custom_inference_provider" ] }