Build with Pi - SambaNova Integration Guide

Pi is a lightweight, terminal-based coding harness. It keeps its core intentionally minimal while allowing users to extend it with TypeScript extensions, skills, prompt templates, themes, and Pi packages. Follow the steps below to connect Pi with a model hosted on SambaCloud.

Prerequisites

Before you begin, ensure you have:

A SambaCloud account and API key
Node.js and npm
A POSIX shell on macOS, Linux, or WSL

Setup

Install Pi

Install Pi globally with npm:

npm install -g --ignore-scripts @earendil-works/pi-coding-agent

Verify that Pi is installed:

pi --version

Configure SambaNova as a provider

Create the Pi agent configuration directory if it doesn’t exist yet:

mkdir -p ~/.pi/agent

Create ~/.pi/agent/models.json with the SambaNova provider configuration:

{
  "providers": {
    "sambanova": {
      "baseUrl": "https://api.sambanova.ai/v1",
      "api": "openai-completions",
      "models": [
        { "id": "MiniMax-M2.7" },
        { "id": "DeepSeek-V3.1" },
        { "id": "Meta-Llama-3.3-70B-Instruct" },
        { "id": "gpt-oss-120b" },
        { "id": "DeepSeek-V3.2" },
        { "id": "gemma-4-31B-it" }
      ]
    }
  }
}

Configure your SambaNova API key

Create ~/.pi/agent/auth.json with your SambaNova API key:

{
  "sambanova": {
    "type": "api_key",
    "key": "your-sambanova-api-key"
  }
}

Replace your-sambanova-api-key with the key from your SambaCloud portal.

Start using SambaNova models

Run Pi with your configured SambaNova provider and model:

pi --provider sambanova --model MiniMax-M2.7

You can use any model configured in ~/.pi/agent/models.json, for example:

pi --provider sambanova --model gpt-oss-120b

pi --provider sambanova --model Meta-Llama-3.3-70B-Instruct

Available models

The SambaNova provider configuration above includes the following models:

MiniMax-M2.7
DeepSeek-V3.1
Meta-Llama-3.3-70B-Instruct
gpt-oss-120b
DeepSeek-V3.2
gemma-4-31B-it

To use a different model, add its model ID to the models array in ~/.pi/agent/models.json, then run Pi with that model:

pi --provider sambanova --model <model-id>

Troubleshooting

If pi isn’t found after installation, verify that the npm global installation directory is on your PATH:

npm prefix -g

Then verify the installation:

pi --version

If ~/.pi/agent/models.json doesn’t exist, create the configuration directory first:

mkdir -p ~/.pi/agent

If the SambaNova API key isn’t being recognized, verify that ~/.pi/agent/auth.json exists and contains the correct provider name and API key:

cat ~/.pi/agent/auth.json

Make sure the provider is named exactly sambanova:

{
  "sambanova": {
    "type": "api_key",
    "key": "your-sambanova-api-key"
  }
}

If requests fail because the configured model isn’t available, verify that the model ID is listed in ~/.pi/agent/models.json and matches the model ID passed to Pi:

pi --provider sambanova --model <model-id>

Overview

Developer tools

Frameworks

Interfaces and low-code

Data and infrastructure

Multimodal

Evaluation and monitoring

Pi integration guide

Prerequisites

Setup

Available models

Troubleshooting

Additional resources

​Prerequisites

​Setup

​Available models

​Troubleshooting

​Additional resources

Prerequisites

Setup

Available models

Troubleshooting

Additional resources