🧠 Abstract v0.2.0

Abstract is a minimal, agent-first Python framework built to orchestrate intelligent agent systems —
powering Aegix and next-generation AI-native platforms.

Abstract provides a clean, modular foundation to define agents, organize teams, run workflows, and scale dynamic execution.
Built for clarity. Designed for scalability.

Think in Higher Dimension

---

🚀 Core Features (v0.2.0)

• ✅ Agent class — LLM-driven agents with tools, goals, memory, and roles
• ✅ OrgTree — Declarative + object-based team hierarchy builder
• ✅ Workflow system — Sequential and parallel task execution
• ✅ LLMCall and ToolCall — Unified task units for orchestration
• ✅ Plug-and-play design for future RAG, memory, and multi-agent scaling

---

📦 Installation

git clone https://github.com/Aegix-AI/Abstract.git
cd Abstract
pip install -e .

---

🧪 Example Usage

from abstract.core.agent import Agent
from abstract.core.orgtree import OrgTree
from abstract.core.workflow import Workflow, Sequential, Parallel, LLMCall, ToolCall

# 1. Define agents
researcher = Agent("ResearcherX", llm="gpt-4o", tools=["web_search"], goal="Find articles")
writer = Agent("WriterY", llm="deepseek", tools=["notion_writer"], goal="Write blog")

# 2. Build team
structure = {"ResearcherX": {"WriterY": {}}}
team = OrgTree(name="ContentPod", manager="ResearcherX", structure=structure, agents=[researcher, writer])

# 3. Create workflow
class ContentCreationWorkflow(Workflow):
    def __init__(self):
        pipeline = Sequential([
            LLMCall("plan_content", agent="ResearcherX"),
            ToolCall("web_search", agent="ResearcherX"),
            LLMCall("write_blog", agent="WriterY"),
        ])
        super().__init__(pipeline)

workflow = ContentCreationWorkflow()

# 4. Run workflow
context = {"topic": "AI Safety"}
result = workflow.run(context)
print(result)

---

🧰 Project Structure

abstract/
├── __init__.py
├── core/
│   ├── agent.py            # Agent class
│   ├── orgtree.py          # Team and hierarchy system
│   ├── workflow.py         # Workflows, Sequential, Parallel, Calls
│   ├── memory.py           # (Coming soon) Memory module
│   ├── api.py              # (Coming soon) API caller
├── icon/
│   └── Abstract_icon.png   # Branding assets
├── setup.py                # Packaging
└── README.md               # You're reading this

---

🔮 Roadmap

• v0.3.0: Memory system (agent memory, scoped memory pools)

• v0.4.0: Unified LLM + Tool API layer (real execution)

• v0.5.0: Retrieval-Augmented Generation (RAG) modules

• v0.6.0: Meta-agent and dynamic agent orchestration

---

📄 License

MIT License © 2025 Aegix