Kite v0.2.1 — AI-Powered CLI

🪁 Kite v0.2.1 — AI-Powered CLI

Release date: March 4, 2026
Install: pip install kite-agent==0.2.1

---

What's new

kite generate — describe your agent, get working code

The biggest addition in this release. Give Kite a natural language description, get a complete, runnable multi-agent Python script back.

kite generate "customer support agent that tracks orders and processes refunds" --out support_bot.py
python support_bot.py

The generator uses your configured LLM (Groq, OpenAI, Anthropic, or Ollama) to design the agent architecture, define tools, wire up the pipeline, and produce a script that follows Kite's patterns correctly. Works with any provider in your .env.

kite compile — turn a Markdown spec into working code

Have a skill.md or any structured text file describing what an agent should do? Compile it directly into a Kite workflow.

kite compile skills/research_assistant.md --out research_agent.py

No rigid template format required — the compiler reads the intent and produces the code.

kite init — scaffold a new project in seconds

Three project types supported:

kite init --type=agent    --name=my_bot       # single agent with tool loop
kite init --type=workflow --name=my_pipeline  # multi-agent pipeline
kite init --type=tool     --name=my_tool      # standalone tool module

Each scaffold includes main.py, .env.example, and requirements.txt, ready to run after adding an API key.

---

Changed

• kite/core.py — config loading now correctly prioritises environment variables, allowing the CLI to override provider settings without modifying code
• setup.py — added console_scripts entry point so kite is available as a shell command after pip install
• tests/test_cli.py — 15 new tests covering all three commands and edge cases (existing directory, invalid skill file, malformed templates)

---

Bug fix in this release

When an LLM call fails during kite generate or kite compile, the output file no longer silently contains an error string. The CLI now exits with a non-zero code and prints a clear error message.

---

How to upgrade

pip install --upgrade kite-agent

Verify the CLI is available:

kite --help

---

Notes

• Requires a configured LLM provider. Fastest option: Groq free tier with llama-3.3-70b-versatile
• kite generate quality depends on the model used — llama-3.3-70b and gpt-4o produce the most reliable output
• Generated scripts may need minor edits for domain-specific tool logic (API keys, DB connections, etc.)

---

Full changelog: CHANGELOG.md
Previous release: v0.1.1

v0.1.1

🚀 New Features (LLM Integrations)

• Groq Integration: Added native support for Groq API (Llama 3.1 8B/70B).
• OpenAI Integration: Added support for OpenAI models (GPT-4o, GPT-5-nano).
• Anthropic Integration: Added support for Claude 3.5 Sonnet and Haiku.
• Verification Script: Added examples/verify_providers.py to test connectivity with all providers.

🛠 Improvements

• Updated LLMFactory to support new providers.
• Improved error handling for API quotas and model not found errors.
• Smoother fallback mechanisms in create_agent.

📦 Dependencies

• Added groq, openai, anthropic to requirements.txt and setup.py.

v0.1.0 - Initial PyPI Release

🎉 First Release on PyPI!

Install via pip:

pip install kite-agent

✨ Features

• Production-ready agentic AI framework
• Multi-LLM provider support (OpenAI, Anthropic, Groq, Ollama, Together)
• Advanced routing (LLM Router, Semantic Router, Resource-Aware Router)
• Built-in safety features (circuit breakers, retry logic, idempotency)
• Memory systems (RAG, sessions, graph knowledge)
• Comprehensive examples (5 production use cases)

📦 Quick Start

from kite import Kite

# Initialize
ai = Kite()

# Create agent
agent = ai.create_agent(
    name="Support",
    system_prompt="You are a helpful assistant.",
    tools=[search_tool, refund_tool],
    agent_type="react"
)

# Run
result = await agent.run("Help me with order #123")

📚 Resources

• PyPI: https://pypi.org/project/kite-agent/
• Documentation: https://github.com/thienzz/Kite/blob/main/README.md
• Examples: https://github.com/thienzz/Kite/tree/main/examples

🔧 What's New in v0.1.0

• ✅ Published to PyPI as kite-agent
• ✅ Fixed .env loading for installed package
• ✅ Added missing kite.optimization module
• ✅ Updated README with pip installation
• ✅ Added ResourceAwareRouter configuration (FAST_LLM_MODEL, SMART_LLM_MODEL)
• ✅ All example tests passing (5/5 configured cases)

📊 Test Results

Case	Status	Time	Features
E-commerce Support	✅ PASSED	52.7s	LLM routing + agents
Enterprise Analytics	⏭️ SKIP	-	Needs pandas (optional)
Research Assistant	✅ PASSED	0.5s	Integration test
Multi-Agent Collab	✅ PASSED	39.9s	Resource-aware routing
DevOps Automation	✅ PASSED	22.6s	Tool execution
Reasoning Architectures	✅ PASSED	24.0s	ReAct loop

Full Changelog: https://github.com/thienzz/Kite/commits/v0.1.0