Skip to main content

​
LiteLLM

LiteLLM is an open-source LLM gateway that provides a unified OpenAI-compatible API to many model providers. Route Fased through LiteLLM to centralize usage tracking, logging, virtual keys, and backend switching behind one proxy.

​
Why use LiteLLM with Fased?

  • Usage tracking — Review Fased traffic across the models your proxy exposes
  • Model routing — Switch between Claude, GPT, Gemini, Bedrock, local models, and other upstreams behind one proxy
  • Virtual keys — Create keys with spend limits and model access limits for Fased
  • Logging — Full request/response logs for debugging
  • Fallbacks — Automatic failover if your primary provider is down

​
Quick start

For normal browser setup, open Agents, select the Agent, then use Agent > Models > LiteLLM to add the proxy URL/key and choose the Agent’s model roles. Use the CLI examples below when provisioning repeatable local or hosted proxies.

​
Via onboarding

fased onboard --auth-choice litellm-api-key

​
Manual setup

  1. Start LiteLLM Proxy:
pip install 'litellm[proxy]'
litellm --model gpt-5.5
  1. Point Fased to LiteLLM:
export LITELLM_API_KEY="your-litellm-key"

fased
That’s it. Fased now routes through LiteLLM. LiteLLM is a proxy, so the available model IDs are whatever your proxy exposes. Fased registers litellm/default as the default initial ref, then refreshes or accepts the model IDs exposed by your proxy.

​
Configuration

​
Environment variables

export LITELLM_API_KEY="sk-litellm-key"

​
Config file

{
  models: {
    providers: {
      litellm: {
        baseUrl: "http://localhost:4000",
        apiKey: "${LITELLM_API_KEY}",
        api: "openai-completions",
        request: { allowPrivateNetwork: true },
        models: [
          {
            id: "default",
            name: "LiteLLM Default",
            reasoning: true,
            input: ["text", "image"],
            contextWindow: 200000,
            maxTokens: 32000,
          },
          {
            id: "gpt-5.5",
            name: "GPT-5.5",
            reasoning: true,
            input: ["text", "image"],
            contextWindow: 1000000,
            maxTokens: 128000,
          },
        ],
      },
    },
  },
  agents: {
    defaults: {
      model: { primary: "litellm/default" },
    },
  },
}

​
Virtual keys

Create a dedicated key for Fased with spend limits: 
curl -X POST "http://localhost:4000/key/generate" \
  -H "Authorization: Bearer $LITELLM_MASTER_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "key_alias": "fased",
    "max_budget": 50.00,
    "budget_duration": "monthly"
  }'
Use the generated key as LITELLM_API_KEY.

​
Model routing

LiteLLM can route model requests to different backends. Configure in your LiteLLM config.yaml: 
model_list:
  - model_name: default
    litellm_params:
      model: openai/gpt-5.5
      api_key: os.environ/OPENAI_API_KEY

  - model_name: gpt-5.5
    litellm_params:
      model: openai/gpt-5.5
      api_key: os.environ/OPENAI_API_KEY
Fased requests the model ID you choose, such as default or gpt-5.5. LiteLLM handles the upstream routing.

​
Model discovery

LiteLLM exposes the models available on your proxy through /models when configured for model discovery. That model list is local to your LiteLLM instance, not a global LiteLLM catalog. 
curl "http://localhost:4000/models" \
  -H "Authorization: Bearer $LITELLM_API_KEY"
Because this catalog is instance-local, fased providers refresh does not fetch a central LiteLLM model list. In the normal UI, LiteLLM is treated as a dynamic provider: litellm/<model-id> is allowed when your proxy exposes that model.

​
Viewing usage

Fased Usage shows the local model usage history for calls made by Fased, grouped by provider, model, Agent, task, channel, and session. LiteLLM’s dashboard or API shows proxy-level spend and quota from LiteLLM’s perspective: 
# Key info
curl "http://localhost:4000/key/info" \
  -H "Authorization: Bearer sk-litellm-key"

# Spend logs
curl "http://localhost:4000/spend/logs" \
  -H "Authorization: Bearer $LITELLM_MASTER_KEY"

​
Notes

  • LiteLLM runs on http://localhost:4000 by default
  • Fased connects via the OpenAI-compatible chat completions endpoint
  • Capabilities depend on the upstream model and how your LiteLLM proxy exposes it

​
See also