Skip to content

configure-providers.md

Latest commit

 

History

History
115 lines (82 loc) · 3.06 KB

configure-providers.md

File metadata and controls

115 lines (82 loc) · 3.06 KB

How to Configure Providers

This guide walks you through setting up AI providers and creating model aliases for easy access.

Overview

gpt2099 uses a two-step provider setup:

  1. Enable providers - Store API keys for authentication
  2. Create aliases - Set up friendly names pointing to specific models

Step-by-Step Setup

1. Enable a Provider

Store your API key and verify connectivity:

gpt provider enable

This will:

  • Prompt you to select from available providers
  • Ask for your API key
  • Test the connection by querying available models
  • Store the key for future use

2. Create Model Aliases

Set up friendly aliases pointing to specific models:

gpt provider ptr milli --set

This will:

  • Show your enabled providers
  • Let you select a provider and model
  • Create an alias you can use with -p milli

Alias Scheme

Aliases reflect relative model capabilities:

nano  < milli  < kilo  < giga

Add .r for reasoning-optimized variants:

  • milli.r - reasoning-optimized lightweight model
  • kilo.r - reasoning-optimized full-capability model

Provider Capabilities

Feature Anthropic Cerebras Cohere Gemini OpenAI
Text conversations yes yes yes yes yes
PDF analysis yes no no yes yes
Image analysis yes no yes* yes yes
Web search yes no no yes no
MCP tools yes yes yes yes yes
Tools + search together yes no no no no
Document caching yes no no yes (auto) yes

Key limitations:

  • Cerebras: No vision/PDF support; JSON schema fields (format, minimum, nullable, $schema) recursively stripped from tools
  • Cohere: Vision and tools mutually exclusive (vision models support images but not tools; other models support tools but not images); no PDF support; no built-in web search
  • Gemini: Cannot use custom MCP tools and web search in the same conversation
  • OpenAI: No built-in web search support

Example Configuration

After setup, your aliases might look like:

{
  "milli": {
    "provider": "anthropic",
    "model": "claude-3-5-haiku-20241022"
  },
  "kilo": {
    "provider": "openai",
    "model": "gpt-4.1"
  },
  "milli.r": {
    "provider": "openai",
    "model": "o4-mini"
  }
}

Usage

Once configured, use aliases in your prompts:

"Hello world" | gpt -p milli
"Complex analysis task" | gpt -p kilo
"Step-by-step reasoning" | gpt -p milli.r

Managing Providers

See the commands reference for the complete list of provider management commands, including:

  • gpt provider models - List available models
  • gpt provider ptr - View current aliases
  • gpt provider enable - Add more providers