Lumoswitch quickstartCopy Anchor Link

Lumoswitch is an LLM access platform and console for AI applications, agents, and developer tools. A Lumoswitch access key can reach configured model providers while Lumoswitch synchronizes fixed, failover, automatic model selection, and load-balancing policies to Lumoswitch Runtime.

What you will haveCopy Anchor Link

• Optional provider keys for connected upstreams
• Platform models and/or connected model resources
• Model selection and failover rules
• A base URL for your selected downstream protocol
• A Lumoswitch access key whose full value is shown only once

PrerequisitesCopy Anchor Link

1. A Lumoswitch account with console access.
2. No provider key is needed when using only platform models.
3. To connect an upstream, prepare a provider key that can list and invoke models.

Never paste real keys into documentation, screenshots, source control, or chat. Lumoswitch does not reveal stored upstream secrets in console lists.

Connect in five stepsCopy Anchor Link

1. Choose model sourcesCopy Anchor Link

Open Quickstart. Platform models appear directly. Add a provider and key only when you also need connected models. Platform models and connected models have the same role in later routing and candidate selection. You can continue without a connected upstream. See Upstreams.

2. Discover modelsCopy Anchor Link

This step shows platform models and fetches connected models from enabled upstreams. You may inspect later steps and save a draft when no model is available, but a callable API configuration still needs at least one healthy routable candidate. See Model resources.

3. Choose a routing modeCopy Anchor Link

Auto-routing order maps candidates onto the SIMPLE, MEDIUM, COMPLEX, and REASONING tiers; adjacent tiers reuse candidates when fewer than four are selected. Least-busy and lowest-latency modes do not treat list order as priority.

4. Configure downstream outputCopy Anchor Link

Set the default client-facing model name and select the client formats to expose. OpenAI uses a /v1 Base URL; Anthropic and Gemini use the inference origin. After creation, the configuration editor can add more client-facing model mappings that target either the whole routing policy or one candidate. Saving creates an API configuration and its initial access key.

Platform models never bypass routing or receive a special direct-access key. Add them to candidates and configure routing exactly like connected models.

5. Save the credentialsCopy Anchor Link

Copy the base URL and access key before closing the dialog.

The complete access key is shown once. Store it immediately in a secrets manager or environment variable.

Make the first requestCopy Anchor Link

export LUMOSWITCH_BASE_URL="https://api.lumoswitch.com/v1"
export LUMOSWITCH_API_KEY="your-lumoswitch-access-key"

curl "$LUMOSWITCH_BASE_URL/models" \
  -H "Authorization: Bearer $LUMOSWITCH_API_KEY"

curl "$LUMOSWITCH_BASE_URL/chat/completions" \
  -H "Authorization: Bearer $LUMOSWITCH_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "lumos-chat",
    "messages": [{"role": "user", "content": "Say hello in one sentence."}]
  }'

Replace lumos-chat with the client-facing model name from your API configuration.

Supported endpointsCopy Anchor Link

GET  /v1/models
POST /v1/chat/completions
POST /v1/responses
POST /v1/embeddings
POST /v1/messages
POST /v1/messages/count_tokens
POST /v1beta/models/{model}:generateContent
POST /v1beta/models/{model}:streamGenerateContent
POST /v1beta/models/{model}:countTokens

Gemini also supports the /models/{model}:... aliases. See Protocols and compatibility for the exact public route allowlist and authentication boundary.

Next stepsCopy Anchor Link

• Confirm the full chain on the Dashboard.
• Tune routing and limits in API configurations.
• Isolate applications or environments with Projects.