GenAI Text Generation Skill

Generate text using the genai-cli text command with Gemini models.

Quick Start

# Basic text generation
uv run genai-cli text "What is the capital of France?"

# With custom model
uv run genai-cli text "Explain quantum computing" --model gemini-3-pro-preview

# Creative writing with temperature
uv run genai-cli text "Write a poem about the ocean" --temperature 0.9

# With system instruction
uv run genai-cli text "Explain this code" --system "You are a senior software engineer"

# Streaming output
uv run genai-cli text "Tell me a long story" --stream

# Save to file
uv run genai-cli text "Generate documentation" --output ./docs.md

CLI Reference

uv run genai-cli text [OPTIONS] PROMPT

Options:
  --model, -m        Model: gemini-2.5-flash (default), gemini-2.5-flash-lite,
                     gemini-3-pro-preview, gemini-2.0-flash
  --temperature, -t  Randomness (0.0-2.0, default: 0.7)
  --max-tokens, -M   Maximum output tokens
  --system, -s       System instruction/persona
  --top-p            Nucleus sampling threshold (0.0-1.0)
  --top-k            Top-k token sampling (1-100)
  --stop             Stop sequences (can be specified multiple times)
  --seed             Random seed for reproducibility
  --stream           Stream output in real-time
  --output, -o       Save response to file
  --json             Output as JSON

Available Models

| Model | Use Case | Notes | |-------|----------|-------| | gemini-2.5-flash | General text/multimodal | Default - balanced speed/quality | | gemini-2.5-flash-lite | Low latency, high volume | Faster, cheaper | | gemini-3-pro-preview | Complex reasoning/coding | Most capable | | gemini-2.0-flash | Alternative general use | Stable |

Parameters

Temperature

Controls randomness. Lower = more deterministic, higher = more creative.

# Factual, deterministic response
uv run genai-cli text "What is 2+2?" --temperature 0.0

# Creative writing
uv run genai-cli text "Write a creative story" --temperature 1.5

System Instruction

Set a persona or context for the model:

# Technical expert
uv run genai-cli text "Review this code" --system "You are a senior security engineer"

# Creative writer
uv run genai-cli text "Write about sunset" --system "You are a poet who loves nature"

# Specific format
uv run genai-cli text "Explain REST APIs" --system "Explain like I'm 5 years old"

Token Control

Limit or control output length:

# Short response
uv run genai-cli text "Summarize this topic" --max-tokens 100

# Long form content
uv run genai-cli text "Write an essay" --max-tokens 2000

Sampling Parameters

Fine-tune generation with advanced sampling:

# Nucleus sampling (top-p)
uv run genai-cli text "Generate ideas" --top-p 0.9

# Top-k sampling
uv run genai-cli text "Complete this sentence" --top-k 40

# Combined for fine control
uv run genai-cli text "Creative brainstorm" --temperature 0.8 --top-p 0.95 --top-k 50

Stop Sequences

Stop generation at specific text:

# Stop at markers
uv run genai-cli text "Write a story" --stop "THE END" --stop "---"

Reproducibility

Use seed for consistent outputs:

# Same seed = same output
uv run genai-cli text "Generate a name" --seed 42
uv run genai-cli text "Generate a name" --seed 42  # Same result

Output Modes

Standard Output

uv run genai-cli text "Hello"
# Output: Hello! How can I assist you today?

Streaming

Real-time output as tokens are generated:

uv run genai-cli text "Tell me a story" --stream
# Output appears word by word

JSON Output

Structured output for automation:

uv run genai-cli text "Hello" --json

{
  "success": true,
  "command": "text",
  "data": {
    "response": "Hello! How can I assist you today?",
    "model": "gemini-2.5-flash"
  },
  "metadata": {
    "temperature": null,
    "max_tokens": null,
    "stream": false,
    "output_file": null
  }
}

File Output

Save directly to file:

uv run genai-cli text "Generate documentation" --output ./docs.md

# With streaming
uv run genai-cli text "Write a report" --stream --output ./report.txt

Use Cases

Code Generation

uv run genai-cli text "Write a Python function to calculate fibonacci" \
  --model gemini-3-pro-preview \
  --system "You are an expert Python developer"

Documentation

uv run genai-cli text "Document this API endpoint" \
  --system "You are a technical writer" \
  --output ./api-docs.md

Creative Writing

uv run genai-cli text "Write a short story about space exploration" \
  --temperature 1.2 \
  --max-tokens 1000 \
  --stream

Data Analysis

uv run genai-cli text "Analyze this dataset and provide insights" \
  --model gemini-3-pro-preview \
  --system "You are a data scientist"

Translation

uv run genai-cli text "Translate to Spanish: Hello, how are you?" \
  --temperature 0.3

Prerequisites

API key must be configured:

uv run genai-cli auth set-key