Integration Guide

LiteLLM Proxy in front of Meridian

Route all your LLM traffic through LiteLLM with Meridian as the upstream provider. One proxy, every model, zero vendor lock-in.

Model alias configuration

Map friendly model names to Meridian endpoints. LiteLLM handles load balancing, fallbacks, and rate limiting transparently.

model_list:
  - model_name: gpt-4o
    litellm_params:
      model: openai/gpt-4o
      api_base: https://api.getnimbus.net/v1
      api_key: os.environ/MERIDIAN_API_KEY
      rpm: 500
      timeout: 60

  - model_name: claude-sonnet
    litellm_params:
      model: openai/claude-3-5-sonnet-20241022
      api_base: https://api.getnimbus.net/v1
      api_key: os.environ/MERIDIAN_API_KEY

  - model_name: meridian-llama-70b
    litellm_params:
      model: openai/meta-llama-3.1-70b-instruct
      api_base: https://api.getnimbus.net/v1
      api_key: os.environ/MERIDIAN_API_KEY
      rpm: 300

litellm_settings:
  drop_params: true
  set_verbose: false
  num_retries: 3
  request_timeout: 120

general_settings:
  master_key: os.environ/LITELLM_MASTER_KEY
  database_url: os.environ/LITELLM_DB_URL

Key points

▸api_base always points to https://api.getnimbus.net/v1 regardless of the upstream model.
▸Model names use the provider/model-id convention. Meridian accepts OpenAI-compatible requests.
▸drop_params: true strips unsupported parameters before forwarding — prevents 400 errors from mismatched model capabilities.

← Back to docs Rate limits →