aiccel 3.1.3

AIccel is a versatile Python library for building lightweight AI agents with multiple LLM providers

⚡ AICCEL Framework 3.1: The Production-Grade Agentic Library

AICCEL (AI-Accelerated Agentic Library) is a high-performance, security-first framework for building orchestrated AI systems.

---

🚀 Why AICCEL?

Most AI frameworks are "heavy"—they import massive dependencies on startup and offer little control over data privacy. AICCEL is different:

• Extreme Performance: Uses advanced Lazy Loading to keep startup times under 100ms.
• Decoupled Architecture: Planning and Execution are separated for infinite scalability.
• Security by Default: FIPS-compliant encryption, automated PII masking, and jailbreak protection built-in.
• Developer First: Dedicated health-check CLI and comprehensive observability tracing.

---

📦 Installation

AICCEL is modular. Install only what you need.

pip install aiccel             # Core (OpenAI, Gemini, Groq support)
pip install aiccel[safety]     # Jailbreak Guard (Transformers)
pip install aiccel[privacy]    # PII Masking (GLiNER)
pip install aiccel[all]        # Full Suite (Security, Data, RAG)

Verify your environment:

aiccel check

---

🏗️ 1. Building High-Performance Agents

The Agent is the core building block. It uses an internal ExecutionPlanner to determine strategies before acting.

from aiccel import Agent, GeminiProvider, SearchTool

# 1. Setup Provider & Tools
provider = GeminiProvider(api_key="...", model="gemini-2.0-flash")
search = SearchTool(api_key="...")

# 2. Build the Agent
agent = Agent(
    provider=provider,
    tools=[search],
    name="ResearchAgent",
    instructions="You are a precise technical researcher."
)

# 3. Execution
result = agent.run("What are the latest breakthroughs in battery tech?")

print(f"Thought: {result['thinking']}")
print(f"Answer: {result['response']}")

⚙️ Feature: Thinking & Planning

AICCEL Agents have a "Think-before-you-act" mode. When thinking_enabled=True, the agent uses a separate planning pass to structure its tool usage, resulting in much higher accuracy for complex tasks.

---

🛡️ 2. Enterprise Security Suite

🕵️ PII Masking (aiccel.privacy)

Automatically detects and masks sensitive data (Emails, Phones, Names) using GLiNER before it hits the LLM.

from aiccel.privacy import mask_text

# Mask data
result = mask_text("Contact John at [e-mail]")
# Output: "Contact [PERSON_1] at [EMAIL_1]"

# Mapping is kept locally to unmask the response later.

🛑 Jailbreak Guard

Protect your system from malicious prompt injections.

from aiccel import AgentConfig
config = AgentConfig(safety_enabled=True) # Blocks attacks automatically

🔐 Secure Vault & Encryption

Military-grade AES-256-GCM encryption for managing API keys and secrets.

from aiccel.encryption import encrypt, decrypt
encrypted = encrypt("my-secret-key", "strong-password")

---

🎼 3. Multi-Agent Orchestration

🤝 AgentManager

Coordinate specialist agents to solve problems that are too big for one LLM.

from aiccel.manager import AgentManager

# Create a expert team
manager = AgentManager(
    llm_provider=provider,
    agents=[research_agent, math_agent, writer_agent]
)

# Collaborative reasoning
response = await manager.collaborate_async("Analyze this fiscal report and summarize findings.")

⛓️ Workflow DAGs

Build deterministic pipelines with loops and conditional routing.

from aiccel import WorkflowBuilder

workflow = (WorkflowBuilder("pipeline")
    .add_agent("research", researcher)
    .add_agent("write", writer)
    .chain("research", "write")
    .build())

---

🔬 4. Advanced Utilities

• Neural Reranking: Advanced semantic sorting for RAG applications.
• MCP Support: Native Model Context Protocol client to connect to thousands of external tools.
• Autonomous Goal Agents: Agents that can break down a high-level goal into a dynamic task list.

---

📖 Full Documentation

Visit our Documentation Directory for specialized guides:

• 🚀 Quickstart Guide
• 🤖 Deep Dive: Agents
• 🛡️ Security & Privacy Full Guide
• ⛓️ Building Complex Workflows