Installation

Install LM Deluge using pip (Python 3.10+):

python -m pip install -U lm-deluge

Optional extras:

pip install plyvel if you want LevelDB-backed local caching
pip install pdf2image pillow if you plan to turn PDFs into images via Image.from_pdf

API Keys

LM Deluge reads API keys from environment variables so the client can contact each provider directly. Load them at process startup (for example with python-dotenv) and pass the values down to your workers, CLI scripts, or notebook kernels.

Required Environment Variables

Depending on which providers you plan to use, set the appropriate API keys:

# OpenAI
OPENAI_API_KEY=sk-...

# Anthropic
ANTHROPIC_API_KEY=sk-ant-...

# Google AI Studio (Gemini)
GEMINI_API_KEY=...

# Cohere
COHERE_API_KEY=...

# OpenRouter (any model prefixed with openrouter:)
OPENROUTER_API_KEY=...

# Meta, Groq, DeepSeek, etc. use provider-specific keys defined in src/lm_deluge/models

# AWS Bedrock (for Amazon-hosted Anthropic and Meta models)
AWS_ACCESS_KEY_ID=...
AWS_SECRET_ACCESS_KEY=...

.env File

Create a .env file in your project root:

OPENAI_API_KEY=sk-...
ANTHROPIC_API_KEY=sk-ant-...
GEMINI_API_KEY=...

Load the file inside your application:

import dotenv

dotenv.load_dotenv()

Verification

Verify your installation by running a simple test:

import dotenv
from lm_deluge import LLMClient

dotenv.load_dotenv()

client = LLMClient("gpt-4.1-mini")
responses = client.process_prompts_sync(["Say hello!"])
print(responses[0].completion)

If you see a response from the model, you’re all set!

Skill for AI Coding Assistants

If you use AI coding assistants like Claude Code or Codex, you can install the lm-deluge skill to give them instant familiarity with the library:

# Install to default location (~/.claude/skills)
deluge skill install

# Or specify a custom directory
deluge skill install ~/.codex/skills

This installs a SKILL.md file that teaches your coding assistant how to use lm-deluge correctly—including the right imports, common patterns, and API gotchas—so it can help you write code that works on the first try.

Next Steps

Head over to the Quick Start guide to learn how to use LM Deluge effectively.