Skip to main content

​
Overview

Dakora provides a comprehensive CLI for managing templates, executing prompts, and comparing model outputs without writing any Python code.

​
Installation

The CLI is automatically available after installing Dakora: 
pip install dakora
Verify installation: 
dakora --help

​
Commands

​
dakora init

Initialize a new Dakora project in the current directory. 
dakora init
Creates:
  • dakora.yaml - Project configuration file
  • prompts/ - Directory for template files
  • prompts/summarizer.yaml - Example template to get started

​
dakora list

List all available templates in the project. 
dakora list
Output: 
summarizer
greeting
code-reviewer

​
dakora get

Display the raw template content for a specific template. 
dakora get <template-id>
Example: 
dakora get summarizer
Output: 
Summarize the following into exactly 3 bullet points:

{{ input_text }}

​
dakora run

Execute a template against a single LLM model. 
dakora run <template-id> --model <model> [OPTIONS]
Required Arguments:
  • <template-id> - ID of the template to execute
  • --model, -m - Model to use (e.g., gpt-4, claude-3-opus)
  • Template-specific inputs as flags (e.g., --input-text "...")
Optional Arguments:
  • --config - Config file path (default: dakora.yaml)
  • --temperature, -t - Sampling temperature (0.0-2.0)
  • --max-tokens - Maximum tokens to generate
  • --top-p - Nucleus sampling probability
  • --json - Output raw JSON result
  • --quiet, -q - Only output response text
Examples: Basic execution: 
dakora run summarizer \
  --model gpt-4 \
  --input-text "Long article text here..."
With LLM parameters: 
dakora run summarizer \
  --model gpt-4 \
  --input-text "Article..." \
  --temperature 0.7 \
  --max-tokens 150
JSON output: 
dakora run summarizer \
  --model gpt-4 \
  --input-text "Article..." \
  --json
Output (default): 
╭─────────────────────────────────────╮
│ Model: gpt-4 (openai)               │
│ Cost: $0.0045 USD                   │
│ Latency: 1,234 ms                   │
│ Tokens: 150 → 80                    │
╰─────────────────────────────────────╯

The article discusses the recent advances...

​
dakora compare

Compare template execution across multiple LLM models in parallel. 
dakora compare <template-id> --models <model1>,<model2>,... [OPTIONS]
Required Arguments:
  • <template-id> - ID of the template to execute
  • --models, -m - Comma-separated list of models
  • Template-specific inputs as flags
Optional Arguments:
  • --config - Config file path (default: dakora.yaml)
  • --temperature, -t - Sampling temperature (applies to all models)
  • --max-tokens - Maximum tokens to generate
  • --top-p - Nucleus sampling probability
  • --json - Output raw JSON result
  • --verbose, -v - Show full responses (not truncated)
Examples: Basic comparison: 
dakora compare summarizer \
  --models gpt-4,claude-3-opus,gemini-pro \
  --input-text "Article to summarize..."
With temperature control: 
dakora compare summarizer \
  --models gpt-4,claude-3-opus \
  --input-text "Article..." \
  --temperature 0.3
Verbose output (full responses): 
dakora compare summarizer \
  --models gpt-4,claude-3-opus \
  --input-text "Article..." \
  --verbose
JSON output: 
dakora compare summarizer \
  --models gpt-4,claude-3-opus \
  --input-text "Article..." \
  --json
Output (default): 
─────────────────────────────────────────────────────────────────────────
 Model           │ Response (preview)                    │ Cost    │ ...
─────────────────────────────────────────────────────────────────────────
✅gpt-4           │ The article discusses recent AI...    │ $0.0045 │ ...
✅claude-3-opus   │ This article explores cutting-edge... │ $0.0038 │ ...
✅gemini-pro      │ A comprehensive overview of modern... │ $0.0012 │ ...
─────────────────────────────────────────────────────────────────────────

Total Cost: $0.0095 | Total Tokens: 450 → 237 | Success: 3/3
Output (verbose): 
✅ gpt-4 (openai)
Cost: $0.0045 | Latency: 1,234 ms | Tokens: 150 → 80

The article discusses the recent advances in artificial intelligence...
(full response text)

────────────────────────────────────────────────────────────────────

✅ claude-3-opus (anthropic)
Cost: $0.0038 | Latency: 856 ms | Tokens: 150 → 75

This article explores cutting-edge developments in AI and ML...
(full response text)

────────────────────────────────────────────────────────────────────

Total Cost: $0.0083
Total Tokens: 300 → 155
Success Rate: 2/2
The compare command handles partial failures gracefully. If one model fails (e.g., API key missing), others continue execution.

​
dakora bump

Increment the semantic version of a template. 
dakora bump <template-id> [OPTIONS]
Options:
  • --patch - Increment patch version (default: 1.0.0 → 1.0.1)
  • --minor - Increment minor version (1.0.0 → 1.1.0)
  • --major - Increment major version (1.0.0 → 2.0.0)
Examples: 
dakora bump greeting --patch   # 1.0.0 → 1.0.1
dakora bump greeting --minor   # 1.0.0 → 1.1.0
dakora bump greeting --major   # 1.0.0 → 2.0.0

​
dakora watch

Watch the prompts directory for changes and automatically reload templates. 
dakora watch
Useful during development to see template changes immediately without restarting your application.
Press Ctrl+C to stop watching.

​
dakora config

Check which API keys are configured for LLM providers. 
dakora config [OPTIONS]
Options:
  • --provider - Check specific provider (e.g., openai, anthropic)
Examples: 
# Check all providers
dakora config

# Check specific provider
dakora config --provider openai
Output: 
✓ OPENAI_API_KEY found
✗ ANTHROPIC_API_KEY not set
✗ GOOGLE_API_KEY not set
See the API Keys Guide for setting up provider authentication.

​
dakora playground

Launch the interactive web playground for developing and testing templates. 
dakora playground [OPTIONS]
Options:
  • --port - Port to run on (default: 3000)
  • --host - Host to bind to (default: localhost)
  • --config - Config file path (default: dakora.yaml)
  • --dev - Development mode with auto-reload
  • --demo - Demo mode (read-only, session isolation)
  • --no-build - Skip building the UI
  • --no-browser - Don’t open browser automatically
Examples: 
# Start playground (opens browser)
dakora playground

# Custom port
dakora playground --port 8080

# Development mode
dakora playground --dev

# Demo mode (read-only)
dakora playground --demo
The playground automatically:
  1. 🔨 Builds the React UI (first run only)
  2. 🚀 Starts the FastAPI server
  3. 🌐 Opens your browser
You can also use the online playground at playground.dakora.io without any installation.

​
Tips & Best Practices

​
Dynamic Arguments

Both run and compare commands support dynamic template inputs: 
# Any --<input-name> flag is passed to the template
dakora run my-template \
  --model gpt-4 \
  --custom-input "value" \
  --another-input "value2"

​
LLM Parameters

Pass any LiteLLM-supported parameter directly: 
dakora run summarizer \
  --model gpt-4 \
  --input-text "..." \
  --temperature 0.7 \
  --max-tokens 100 \
  --top-p 0.9 \
  --frequency-penalty 0.5

​
Error Handling

The CLI provides clear error messages: 
❌ API key error: OPENAI_API_KEY not found
💡 Set the required environment variable (e.g., OPENAI_API_KEY)

❌ Template 'nonexistent' not found
💡 Run 'dakora list' to see available templates

​
Piping and Scripting

Use --json for programmatic access: 
# Capture JSON output
result=$(dakora run summarizer --model gpt-4 --input-text "..." --json)

# Extract specific fields with jq
echo $result | jq '.output'
echo $result | jq '.cost_usd'
Use --quiet for simple text output: 
# Only the response text
dakora run summarizer --model gpt-4 --input-text "..." --quiet

​
Next Steps

API Keys Setup

Configure API keys for LLM providers

Model Comparison

Learn more about comparing models

LLM Execution

Deep dive into template execution

Examples

See real-world usage examples