Installation

For Users

uv

```bash
uv add ace-framework
```

With extras

```bash
uv add ace-framework[all]            # All optional features
uv add ace-framework[instructor]     # Structured outputs (Instructor)
uv add ace-framework[langchain]      # LangChain integration
uv add ace-framework[browser-use]    # Browser automation
uv add ace-framework[claude-code]    # Claude Code CLI integration
uv add ace-framework[claude-sdk]     # Anthropic SDK integration steps
uv add ace-framework[observability]  # Opik monitoring + cost tracking
uv add ace-framework[deduplication]  # Skill deduplication (embeddings)
uv add ace-framework[transformers]   # Local model support
```

For Contributors

UV (Recommended)

```bash
git clone https://github.com/kayba-ai/agentic-context-engine
cd agentic-context-engine
uv sync  # Installs everything (10-100x faster than pip)
```

uv

```bash
git clone https://github.com/kayba-ai/agentic-context-engine
cd agentic-context-engine
uv add -e .
```

Requirements

Python 3.12
An API key for your LLM provider

Configure Your LLM

The recommended way to set up your API keys and model selection:

ace setup

This interactive wizard validates your API key and model, then saves config to ace.toml (model names, safe to commit) and .env (API keys, gitignored). See Setup for full details.

Manual alternative

If you prefer not to use the wizard, set environment variables directly:

export OPENAI_API_KEY="sk-..."

Or create a .env file (add to .gitignore):

OPENAI_API_KEY=sk-...

Verify Installation

from ace import ACELiteLLM

# Uses ace.toml + .env from `ace setup`
agent = ACELiteLLM.from_setup()
print(agent.ask("Hello!"))

Or without ace setup:

agent = ACELiteLLM.from_model("gpt-4o-mini")
print(agent.ask("Hello!"))

Set Up Coding Agent Skills (Optional)

If you use Claude Code, install the Kayba pipeline skill:

kayba setup

This installs the evaluation pipeline skill to .claude/skills/ and prints CLI instructions. See Hosted API for details.