Multi-model provider implementations for OpenAI Agents, supporting both OpenAI and Ollama models. Works with both local Ollama instances and Ollama Cloud.

This repository contains equivalent implementations in both Python and TypeScript.

• Python Package (timestep) - PyPI
• TypeScript Package (@timestep-ai/timestep) - npm

pip install timestep

from timestep import MultiModelProvider, MultiModelProviderMap, OllamaModelProvider
from agents import Agent, Runner, RunConfig
import os

# Setup provider
model_provider_map = MultiModelProviderMap()
if os.environ.get("OLLAMA_API_KEY"):
    model_provider_map.add_provider(
        "ollama",
        OllamaModelProvider(api_key=os.environ.get("OLLAMA_API_KEY"))
    )

model_provider = MultiModelProvider(
    provider_map=model_provider_map,
    openai_api_key=os.environ.get("OPENAI_API_KEY", ""),
)

# Create agent and run
agent = Agent(model="ollama/llama3")
run_config = RunConfig(model_provider=model_provider)
result = Runner.run_streamed(agent, agent_input, run_config=run_config)

npm install @timestep-ai/timestep
# or
pnpm add @timestep-ai/timestep

import { MultiModelProvider, MultiModelProviderMap, OllamaModelProvider } from '@timestep-ai/timestep';
import { Agent, Runner } from '@openai/agents';

// Setup provider
const modelProviderMap = new MultiModelProviderMap();
if (Deno.env.get('OLLAMA_API_KEY')) {
  modelProviderMap.addProvider(
    'ollama',
    new OllamaModelProvider({ apiKey: Deno.env.get('OLLAMA_API_KEY') })
  );
}

const modelProvider = new MultiModelProvider({
  provider_map: modelProviderMap,
  openai_api_key: Deno.env.get('OPENAI_API_KEY') || '',
});

// Create agent and run
const agent = new Agent({ model: 'ollama/llama3' });
const runner = new Runner({ modelProvider });
const result = await runner.run(agent, agentInput, { stream: true });

Both implementations provide:

• MultiModelProvider: Automatically routes requests to OpenAI or Ollama based on model name prefixes
• OllamaModelProvider: Provides access to Ollama models (local or cloud)
• OllamaModel: Direct model implementation that converts Ollama responses to OpenAI-compatible format
• MultiModelProviderMap: Manages custom mappings of model name prefixes to providers

• Multi-Model Support: Seamlessly switch between OpenAI and Ollama models
• Automatic Routing: Model names with prefixes (e.g., ollama/llama3) automatically route to the correct provider
• Customizable: Add your own providers using MultiModelProviderMap
• OpenAI Compatible: Works with the OpenAI Agents SDK
• Ollama Integration: Full support for both local Ollama instances and Ollama Cloud
• Cross-Language: Equivalent implementations in Python and TypeScript

• Python Documentation
• TypeScript Documentation

We're actively developing additional features for the timestep libraries:

• Additional Abstractions: Gradually abstracting out other logic from Timestep AI into reusable library components
• CLI Tool: A proper command-line interface with tracing support for debugging and monitoring agent interactions

MIT License - see LICENSE file for details.