Sqirvy-cli (Go Implementation)

This directory contains the Go implementation of sqirvy-cli, a command-line tool for interacting with various Large Language Models (LLMs).

Overview

The Go version of sqirvy-cli provides a native executable for querying LLMs from the terminal. It leverages several popular Go libraries to offer a robust and efficient command-line experience.

Key Features

• Native Executable: Compiles to a single binary for easy distribution and execution.
• Multi-Provider Support: Interacts with Anthropic, Google Gemini, OpenAI, and Llama models via the langchaingo library.
• Structured Commands: Uses the cobra library for a clear command structure:
  • query: Sends arbitrary prompts (default command).
  • plan: Requests the LLM to generate a plan.
  • code: Asks the LLM to generate source code.
  • review: Instructs the LLM to review code or text.
  • models: Lists supported models and their providers.
• Flexible Input: Reads prompts from:
  • Standard Input (stdin) for easy piping.
  • File paths.
  • URLs (content is scraped using the colly library).
• Configuration:
  • Command-line flags (-m for model, -t for temperature) managed by cobra.
  • Environment variables for API keys (ANTHROPIC_API_KEY, GEMINI_API_KEY, OPENAI_API_KEY, LLAMA_API_KEY) and base URLs (OPENAI_BASE_URL, LLAMA_BASE_URL).
  • Optional configuration file support via viper (default: $HOME/.config/sqirvy-cli/config.yaml).
• System Prompts: Uses embedded .md files for command-specific system prompts (query.md, plan.md, code.md, review.md).
• Modular Design:
  • cmd/sqirvy-cli: Contains the main application logic, command definitions (cobra), and prompt reading/processing.
  • pkg/sqirvy: Implements the core LLM interaction logic, defining the Client interface and provider-specific implementations (Anthropic, Gemini, OpenAI, Llama) using langchaingo. Manages model-provider mapping and token limits.
  • pkg/util: Provides utility functions for file reading (files.go) and web scraping (scraper.go).

Building

You can build the executable using the standard Go toolchain:

cd go/cmd/sqirvy-cli
go build -o sqirvy-cli main.go

Alternatively, use the provided Makefiles:

make build # Builds the binary in the project root

or

cd go
make build # Builds the binary in go/bin/

Running

Once built, you can run the tool from your terminal:

# Basic query using default model and temperature
echo "What is the capital of France?" | ./sqirvy-cli

# Specify model and temperature, providing a file
./sqirvy-cli -m claude-sonnet-4-20250514 -t 0.7 query my_prompt.txt

# Generate a plan from stdin
cat requirements.txt | ./sqirvy-cli plan -m gpt-4o

# Generate code based on a plan file and a URL
./sqirvy-cli code -m gemini-1.5-pro plan.md https://example.com/api-spec

# List available models
./sqirvy-cli models

Remember to set the required API key environment variables for the models you intend to use.

Testing

Unit tests are included in the pkg/sqirvy and pkg/util directories. Run tests using:

cd go
make test

Or run tests for specific packages:

cd go/pkg/sqirvy
go test ./...

cd go/pkg/util
go test ./...

Note: Some tests in pkg/sqirvy require API keys to be set in the environment and will be skipped otherwise.