🦁 Lion Framework

Language InterOperable Network

A powerful Python framework for structured AI conversations and operations

🌟 Features

• 🎯 Dynamic structured output at runtime
• 🔄 Easy composition of multi-step processes
• 🤖 Support for any model via litellm
• 🏗️ Built-in conversation management
• 🧩 Extensible architecture
• 🔍 Type-safe with Pydantic models

🚀 Quick Install

pip install lion-os

Note

the operation API is experimental and may change in future versions. Use with caution.

💡 Usage Examples

1️⃣ Simple Communication

using litellm integration

from lion import LiteiModel, Branch

# Initialize model and branch
imodel = LiteiModel(
    model="openai/gpt-4o",
    api_key="OPENAI_API_KEY",
    temperature=0.2,
)
branch = Branch(imodel=imodel)

# Basic communication
result = await branch.communicate(
    instruction="Give me ideas for FastAPI interview questions",
    context="We're hiring senior engineers"
)

using lion's own service system (only supports openai / anthropic / perplexity / groq)

from lion import iModel

# Initialize model and branch
imodel = iModel(
    provider="openai",
    model="gpt-4o",
    api_key="OPENAI_API_KEY",
    temperature=0.2,
    task="chat",
)

# if use anthropic
# imodel = iModel(
#     provider="anthropic",
#     model="claude-3-5-sonnet-20241022",
#     task="messages",
#     api_key="ANTHROPIC_API_KEY",
#     max_tokens=500,
# )

# use perplexity
# imodel = iModel(
#     provider="perplexity",
#     model="llama-3.1-sonar-small-128k-online",
#     task="chat/completions",
#     api_key="PERPLEXITY_API_KEY",
#     max_tokens=500,
# )

# use groq
# imodel = iModel(
#     provider="groq",
#     model="llama3-8b-8192",
#     task="chat/completions",
#     api_key="GROQ_API_KEY",
#     max_tokens=500,
# )
branch = Branch(imodel=imodel)

# Basic communication
result = await branch.communicate(
    instruction="Give me ideas for FastAPI interview questions",
    context="We're hiring senior engineers"
)

2️⃣ Structured Output with Pydantic

from pydantic import BaseModel

class CodingQuestion(BaseModel):
    question: str
    evaluation_criteria: str

# Get structured responses
questions = await branch.operate(
    instruction="Generate FastAPI coding questions",
    context="Technical interview context",
    operative_model=CodingQuestion
)

3️⃣ Advanced Operations (Brainstorming)

from lion.operations import brainstorm

result = await brainstorm(
    instruct={
        "instruction": "Design API endpoints for a todo app",
        "context": "Building a modern task management system"
    },
    imodel=imodel,
    num_instruct=3,
    operative_model=CodingQuestion,
    auto_run=True
)

🎯 Key Components

Requirements

python 3.11+ required

⭐ Star History