Chatbox API — LobeChat, Chatbox, LibreChat & OpenAI-Compatible UIs

LumeAPI gives chat clients a single OpenAI-compatible Chat Completions endpoint at https://api.lumeapi.site/v1. LobeChat, Chatbox desktop, and LibreChat can send standard chat requests to that gateway with a Bearer API key from the LumeAPI Console. The client chooses the model by its exact catalog id; LumeAPI does not translate approximate names or provider aliases.

The implementation detail that causes most setup failures is the base URL. Configure https://api.lumeapi.site/v1 as the API host or base URL, not a completed Chat Completions URL. OpenAI-compatible SDKs and clients normally append /chat/completions themselves. Supplying https://api.lumeapi.site/v1/chat/completions as the base URL can produce duplicated paths or connection failures.

This architecture is useful when teams want one chat workspace while routing work across GPT, Claude, and Gemini tiers. A support team can use gemini-3-flash for high-volume conversations, move difficult analysis to claude-sonnet-4-6, and reserve gpt-5.6-terra for work where its stronger tier is justified. The conversation client remains the user interface; LumeAPI is the independent gateway behind it.

Do not treat LumeAPI as OpenAI, Anthropic, or Google. Use a LumeAPI API key from the Console, point the client to the LumeAPI gateway, and select the LumeAPI catalog model ids exactly as published. This keeps credentials, endpoint ownership, and billing expectations clear for administrators and end users.

Gateway: https://api.lumeapi.site/v1. Authentication: Bearer API key from the LumeAPI Console.

Use LobeChat's OpenAI-compatible environment configuration when deploying or self-hosting LobeChat. Keep the endpoint at the versioned API base URL.

• Create or copy a LumeAPI API key from the LumeAPI Console.
• Set OPENAI_API_KEY to the LumeAPI API key in the environment used by the LobeChat deployment.
• Set OPENAI_PROXY_URL=https://api.lumeapi.site/v1.
• Set ENABLED_OPENAI=1 so the OpenAI-compatible provider is available in LobeChat.
• Restart or redeploy LobeChat after changing the environment variables; environment changes are not applied to an already running container or process automatically.
• Open the LobeChat model selector and add or select exact LumeAPI catalog ids such as gpt-5.6-terra, claude-sonnet-4-6, gemini-3.5-flash, or gemini-3-flash.
• Start with one short non-streaming-style prompt such as “Reply with the word connected” to separate connectivity problems from model or prompt behavior.
• If LobeChat returns empty responses, test the documented /v1 suffix adjustment for OPENAI_PROXY_URL. Do not append /chat/completions to the configured base URL.
• Keep the API key server-side in deployment environment variables. Do not expose a shared gateway key in browser-delivered client configuration.

Verified LobeChat settings: OPENAI_API_KEY, OPENAI_PROXY_URL=https://api.lumeapi.site/v1, ENABLED_OPENAI=1.

Chatbox desktop is a good fit for individual operators, internal evaluators, and support leads who need a local chat client without changing application code.

• Open Chatbox and go to Settings → Model Provider → Add.
• Choose OpenAI API compatible as the provider type.
• Enter a provider name that makes ownership obvious to users, such as LumeAPI.
• Paste a LumeAPI Console API key into the provider API key field.
• Set API Host to https://api.lumeapi.site/v1.
• Leave API Path at its default value, /chat/completions.
• Add catalog model ids exactly, beginning with gemini-3-flash for a low-cost connectivity test and gpt-5.6-terra or claude-sonnet-4-6 for higher-tier evaluation.
• Use Check connection before starting production conversations.
• Send a short prompt to each configured model and confirm that the response is attributed to the intended model selection.
• If connection validation fails, verify that API Host ends at /v1 and that /chat/completions appears only in the API Path field, not inside API Host.

Verified Chatbox settings path: Settings → Model Provider → Add → OpenAI API compatible. API Host: https://api.lumeapi.site/v1. API Path default: /chat/completions.

endpoints:
  custom:
    - name: LumeAPI
      baseURL: https://api.lumeapi.site/v1
      apiKey: ${LUMEAPI_KEY}
      models:
        default:
          - gpt-5.6-terra
          - gpt-5.4-mini
          - claude-sonnet-4-6
          - gemini-3.5-flash
          - gemini-3-flash
      titleConvo: true

Save this in librechat.yaml and provide LUMEAPI_KEY through the LibreChat runtime environment. The baseURL is the API base only; do not add /chat/completions.

LibreChat works best when endpoint configuration is treated as deployable infrastructure: version the YAML, inject credentials through the runtime environment, and validate every exposed model after changes.

• Add the LumeAPI endpoint under librechat.yaml → endpoints.custom using the name LumeAPI.
• Set baseURL to https://api.lumeapi.site/v1 exactly.
• Set apiKey to ${LUMEAPI_KEY} instead of placing the raw key in librechat.yaml.
• Add LUMEAPI_KEY to the environment used by the LibreChat service, container, or deployment platform.
• Populate models.default with exact catalog strings that your users are allowed to select, such as gpt-5.6-terra, gpt-5.4-mini, claude-sonnet-4-6, gemini-3.5-flash, and gemini-3-flash.
• Set titleConvo: true if you want LibreChat to generate conversation titles through the configured endpoint behavior.
• Restart the LibreChat service after updating librechat.yaml or its environment variables.
• Open LibreChat, select the LumeAPI endpoint, and send a one-sentence test prompt to every model listed under models.default.
• Remove models that your organization does not intend to offer rather than using the model picker as an unrestricted catalog browser.
• When troubleshooting, first compare the selected model string with librechat.yaml character for character. A near-match is not a valid LumeAPI model id.

Verified LibreChat configuration fields: librechat.yaml → endpoints.custom; name LumeAPI; baseURL https://api.lumeapi.site/v1; apiKey ${LUMEAPI_KEY}; models.default; titleConvo true.

Use the lower-cost tiers as defaults for routine chat, then make premium models deliberate user-selectable options. Catalog pricing was updated 2026-07-22; figures below are listed as input / output catalog rates.

Prices are catalog figures supplied for 2026-07-22. Model availability and exact model ids should be reviewed against the current LumeAPI catalog before changing production defaults.

The practical operating model is simple: each chat client owns its interface, user session behavior, saved prompts, and local configuration; LumeAPI receives OpenAI-compatible Chat Completions requests at one gateway. That separation lets an operations team standardize endpoint configuration across LobeChat, Chatbox, and LibreChat while still giving different groups the client experience they prefer.

Model selection is the main economics control. Do not make every conversation default to a premium tier because it is available. Set gemini-3-flash or gpt-5.4-mini as the first-choice model for routine work, then make gpt-5.6-terra, claude-sonnet-4-6, or higher-priced tiers available for tasks that justify them. This is easier to govern when LibreChat's models.default list and Chatbox's manually added model list contain only approved ids.

Credential handling deserves the same discipline as any production API integration. Store LumeAPI keys in server or deployment environment variables for LobeChat and LibreChat. For a desktop tool such as Chatbox, treat the configured API key as access granted to the desktop user and avoid distributing one shared unrestricted key across a large organization. Rotate keys through the LumeAPI Console when access changes.

There are client limitations to account for. OpenAI-compatible chat configuration applies to Chat Completions workflows, not every feature branded as AI inside a developer tool. For example, Cursor tab completion may not use a custom URL even when its chat features can use a compatible endpoint. Test the exact feature your team plans to deploy, including streaming behavior, model selection, title generation, and connection checks, before presenting it as supported.

LumeAPI is independent from OpenAI, Anthropic, and Google. Use LumeAPI-issued credentials and exact LumeAPI catalog model ids.