litellm
par ishaan-jaff
Coding Agents & IDEsRésumé
Call 100+ LLM providers through LiteLLM's unified API.
Description
Ce que fait ce skill
LiteLLM provides a unified Python API to call 100+ LLM providers using the same interface. It handles routing, so you can switch between OpenAI, Anthropic, Google, and others without rewriting call logic. A proxy mode adds caching, rate limiting, and observability on top.
Instead of maintaining separate SDK integrations for each provider, one call signature works across all of them.
Quand l'utiliser
- Route code review tasks to GPT-4o while running on Claude
- Compare outputs from three models on the same prompt
- Switch to gpt-4o-mini for simple classification to cut costs
- Access Gemini or Mistral from a runtime that only supports OpenAI format
- Point all LLM calls through a shared proxy for usage tracking
Exemple de workflow
User asks: compare how GPT-4o and Claude answer 'What is the best sorting algorithm?'
- 1
Import litellm and define the prompt as a messages list
- 2
Loop over ["gpt-4o", "claude-sonnet-4-20250514"]
- 3
Call litellm.completion(model=model, messages=prompt) for each
- 4
Extract response.choices[0].message.content from each result
- 5
Print each model name alongside its response
Side-by-side text outputs from GPT-4o and Claude for the same question
Prérequis
- OPENAI_API_KEY environment variable (for OpenAI models)
- ANTHROPIC_API_KEY environment variable (for Anthropic models)
Installation
clawhub install ishaan-jaff/litellmExécute la commande CLI, ou colle l'instruction dans Claude Code, Cursor, Codex ou l'agent de code que tu utilises.