determinagent 0.13.0

CLI-First deterministic multi-agent orchestration library

DeterminAgent

CLI-First Deterministic Multi-Agent Orchestration Library

Orchestrate powerful AI workflows at zero extra cost. DeterminAgent controls multiple AI CLI tools (Claude Code, Copilot CLI, Gemini CLI, OpenAI Codex) using LangGraph to create deterministic pipelines powered by your existing flat-rate subscriptions.

---

🚀 First Contact

DeterminAgent is a Python library for developers who want to build complex, multi-agent systems without paying for expensive per-token API calls. By wrapping the CLI tools you already pay for, DeterminAgent allows you to build production-grade workflows for $0 in variable costs.

Key Features

• Library-Only: Full control in pure Python. No proprietary YAML DSL.
• Subscription Arbitrage: Uses your flat-rate CLI subscriptions.
• Deterministic: Powered by LangGraph state machines.
• Zero-Latency: Controls local tools via subprocess.

---

📦 Installation

From PyPI (Recommended)

pip install determinagent

From Source

For the latest development version or to contribute:

# Clone the repository
git clone https://github.com/Experto-AI/determinagent.git
cd determinagent

# Install dependencies and set up environment
poetry install

# Verify installation
poetry run python -c "import determinagent; print(determinagent.__version__)"

Prerequisites

• Python 3.10+
• At least one supported AI CLI tool installed and authenticated:
  • Claude Code (claude)
  • Copilot CLI (copilot)
  • Gemini CLI (gemini)
  • OpenAI Codex (codex)

---

⚡ Quick Start

Library Usage

from determinagent import UnifiedAgent, SessionManager

# Create a deterministic agent
writer = UnifiedAgent(
    provider="claude",
    model="balanced",
    role="Technical Blogger",
    session=SessionManager("claude")
)

# Send a prompt - zero per-token cost!
response = writer.send("Explain LangGraph in 3 sentences.")
print(response)

Template Flows

Don't start from scratch. Use our pre-built Python templates in the flows/ directory:

• flows/blog/: Complete Writer → Editor → Reviewer workflow with human review.

To run the blog flow:

python flows/blog/main.py "My Blog Topic" --writer claude --editor copilot

---

🧩 Compatibility Matrix

Provider	Adapter Status	Session Support	Web Search	Model Aliases
Claude Code	✅ Alpha	✅ Native	✅ Yes	fast, balanced, powerful, reasoning, free
Copilot CLI	✅ Alpha	✅ Native	✅ Yes	fast, balanced, powerful, reasoning, free
Gemini CLI	✅ Alpha	✅ Native	✅ Yes	fast, balanced, powerful, reasoning, free
OpenAI Codex	✅ Alpha	✅ Native	❌ No	fast, balanced, powerful, reasoning, free

---

🎯 Model Alias Map

DeterminAgent resolves model aliases per provider so you can keep flows consistent.

Alias	Claude Code	Gemini CLI	Copilot CLI	OpenAI Codex
fast	haiku	gemini-3-flash-preview	claude-haiku-4.5	gpt-5.1-codex-mini
balanced	sonnet	gemini-3-pro-preview	claude-sonnet-4.5	gpt-5.1-codex
powerful	opus	gemini-3-pro-preview	claude-opus-4.5	gpt-5.1-codex-max
reasoning	opus	gemini-3-pro-preview	gpt-5.2	gpt-5.1-codex
free	haiku	gemini-3-flash-preview	claude-haiku-4.5	gpt-5.1-codex-mini

Notes:

• You can always pass an exact model string to override the alias.
• Availability depends on your provider plan and CLI version.
• Gemini 3 preview models require enabling Preview Features in Gemini CLI; if unavailable, pass gemini-2.5-pro or gemini-2.5-flash.
• Codex CLI does not enumerate models in --help; defaults mirror Codex model names exposed by Copilot CLI.

---

🛠️ Troubleshooting

Common Issues

1. ProviderNotAvailable: CLI command 'claude' not found

   • Ensure the tool is installed and available in your $PATH.
   • Run claude --version manually to verify.

2. Authentication Errors

   • DeterminAgent uses your local sessions. Ensure you are logged in to the CLI tool (e.g., copilot auth status or claude login).

3. Subprocess Timeouts

   • Some agents (like Writer) can take a few minutes for long content. Ensure your environment doesn't kill long-running processes.

Debug Mode

Set LOG_LEVEL=DEBUG to see the full subprocess commands and raw output.

---

📖 Documentation

Core Documentation

• Technical Architecture: Design principles and system internals.
• CLI Reference: Low-level flag mappings for each provider.
• Actionable Plan & Roadmap: Current status and next steps.

API Reference

• UnifiedAgent: Core orchestration class.
• Provider Adapters: Claude, Copilot, Gemini, Codex wrappers.
• SessionManager: Conversation history management.
• Exceptions: Error handling hierarchy.

Tutorials

• Your First Flow: Step-by-step guide to building a workflow.

Community

• Contributing: How to help improve the project.
• Code of Conduct: Community guidelines.
• Security Policy: Reporting vulnerabilities.

---

📜 License

Apache License 2.0 - see LICENSE for details.