Skip to content

Configuration

Configure QUASAR with API keys and options.

API Keys

QUASAR requires API keys from at least one provider. You can use multiple keys for automatic rotation on rate limits.

Create a .env file in your project directory:

# Groq (recommended - fast inference)
GROQ_API_KEY_1=gsk_your_key_here
GROQ_API_KEY_2=gsk_your_second_key_here

# Cerebras
CEREBRAS_API_KEY_1=csk_your_key_here
CEREBRAS_API_KEY_2=csk_your_second_key_here

# Ollama runs locally - no API key needed
# Default: http://localhost:11434

# Web Search (optional but recommended)
# QUASAR has multiple web tools with fallback support
# Get free key at: https://tavily.com
TAVILY_API_KEY=tvly_your_key_here

Option 2: Environment Variables

# Groq
export GROQ_API_KEY_1="gsk_your_key_here"
export GROQ_API_KEY_2="gsk_your_second_key_here"

# Cerebras
export CEREBRAS_API_KEY_1="csk_your_key_here"

# Temporary (current session only)
set GROQ_API_KEY_1=gsk_your_key_here
set CEREBRAS_API_KEY_1=csk_your_key_here

# Permanent (restart terminal after)
setx GROQ_API_KEY_1 "gsk_your_key_here"
setx CEREBRAS_API_KEY_1 "csk_your_key_here"

# Temporary (current session only)
$env:GROQ_API_KEY_1 = "gsk_your_key_here"
$env:CEREBRAS_API_KEY_1 = "csk_your_key_here"

# Permanent
[System.Environment]::SetEnvironmentVariable("GROQ_API_KEY_1", "gsk_your_key_here", "User")

Get Free API Keys

Provider Free Tier Link
Groq ✅ Yes console.groq.com
Cerebras ✅ Yes cloud.cerebras.ai
Ollama ✅ Free (Local) ollama.ai
Tavily ✅ Yes tavily.com

Multiple Keys

You can add multiple keys per provider:

GROQ_API_KEY_1=gsk_first_key
GROQ_API_KEY_2=gsk_second_key
GROQ_API_KEY_3=gsk_third_key

Benefits:

  • Automatic rotation on rate limits
  • No interruptions during heavy usage
  • Seamless failover

Provider Priority

QUASAR uses this default priority:

  1. Cerebras - Best for tool calling
  2. Ollama - Local, good fallback
  3. Groq - Fast, but tool calling issues with some models

You can override with --model flag:

quasar --model groq/openai/gpt-oss-120b "your request"

Ollama Configuration

QUASAR uses cloud models via Ollama for fast execution.

Setup

  1. Install Ollama from ollama.ai
  2. Start the server: ollama serve
  3. Pull the required cloud models:
# Required for Auto mode
ollama pull glm-4.7:cloud
ollama pull deepseek-v3.1:671b-cloud
ollama pull qwen3-coder:480b-cloud

Default Models

Model Description
glm-4.7:cloud Default, balanced
deepseek-v3.1:671b-cloud Code intelligence
gpt-oss:120b-cloud Large, powerful
qwen3-coder:480b-cloud Coding focused

Custom/Local Models

You can also use any local model with the --model flag:

# Use a custom local model
quasar --model ollama/qwen2.5-coder:7b "your request"
quasar --model ollama/codellama:7b "your request"

Default Ollama URL: http://localhost:11434

Next Steps