abi-core-ai 1.11.3

Agent-Based Infrastructure Core - Runtime and CLI

ABI-Core AI 🤖

Build AI agents with deterministic DAG execution, semantic discovery, and governed autonomy.

ABI-Core is a Python framework for creating agents that run as containerized services with structured execution graphs, inter-agent communication, and policy-driven security. One pip install, one CLI command, and you have a running agent system.

pip install abi-core-ai
abi-core create swarm --name my-system
abi-core run

⚠️ Beta — Pipeline is functional end-to-end. APIs may change between minor versions.

---

Create an Agent in 3 Files

1. Define the DAG (app.py)

from abi_core.agent import AbiCore
from .my_agent import MyAgent

agent = AbiCore()

@agent.step(name="parse_input")
async def parse_input(raw_input: str) -> dict:
    return {"query": raw_input.strip(), "timestamp": time.time()}

@agent.step(name="process", depends_on=["parse_input"], input_map={"data": "$parse_input.result"})
async def process(data: dict) -> dict:
    result = await invoke(config.LLM_CONFIG, data["query"])
    return {"output": result}

@agent.step(name="respond", depends_on=["process"], input_map={"result": "$process.result"})
async def respond(result: dict) -> dict:
    return {"response": result["output"]}

agent.run(MyAgent())

2. Define the Agent Class (my_agent.py)

from abi_core.agent import AbiAgent

class MyAgent(AbiAgent):
    def __init__(self):
        super().__init__(
            agent_name="my-agent",
            description="Processes user queries with structured reasoning",
            llm_config={"provider": "ollama", "model": "qwen3:8b", "temperature": 0.3},
            system_prompt="You are a helpful assistant that...",
        )

3. Configure Identity (config/config.py)

import os

AGENT_NAME = "my-agent"
DESCRIPTION = "Processes user queries with structured reasoning"
LLM_CONFIG = {"provider": "ollama", "model": "qwen3:8b", "temperature": 0.3}
OLLAMA_HOST = os.getenv("OLLAMA_HOST", "http://localhost:11434")

That's it. abi-core run containerizes and starts your agent with A2A protocol, health checks, and semantic registration.

---

Key Concepts

Decorator API

Decorator	Purpose
@agent.step(name, depends_on, input_map)	Deterministic DAG node — compiled at init, executed per request
@agent.tool(name)	DAG node + LangChain tool — LLM can invoke on demand
@agent.mcp_tool(name)	Remote tool via MCP protocol with HMAC auth
@agent.task(name, tools)	Programmatic orchestrator — manual step execution with branching

DAG Execution

Steps registered with @agent.step() compile into a ToolExecutionGraph (LangGraph). Nodes at the same dependency level run in parallel. The LLM never decides execution order — the graph does.

# These two run in parallel (no dependencies between them)
@agent.step(name="classify")
async def classify(raw_input): ...

@agent.step(name="validate")
async def validate(raw_input): ...

# This waits for both
@agent.step(name="decide", depends_on=["classify", "validate"],
            input_map={"cls": "$classify.result", "valid": "$validate.result"})
async def decide(cls, valid): ...

invoke() — Unified LLM Calls

from abi_core.agent import invoke

# Simple LLM call
result = await invoke(config.LLM_CONFIG, "Classify this query...")

# With conversation memory
result = await invoke(config.LLM_CONFIG, "Follow up...", thread_id=session_id)

# With tools
result = await invoke(config.LLM_CONFIG, "Find...", tools=[search_tool, write_tool])

Inter-Agent Communication

Agents talk to each other via A2A protocol using AgentInteractionFlow:

from abi_core.common.workflow import AgentInteractionFlow

flow = AgentInteractionFlow(source_agent="my-agent", target_agent="planner", host=planner_url)
async for event in flow.stream(query="Decompose this task"):
    process(event)

---

CLI

# Scaffold
abi-core create swarm --name <name>          # Full system: agents + services + compose
abi-core create project <name>               # Project only
abi-core add agent <name> --description "…"  # Add agent to existing project
abi-core add semantic-layer                  # Add Weaviate semantic discovery
abi-core add service guardian-native         # Add OPA security gate

# Run
abi-core run                # Detached (status table)
abi-core run --logs         # With container output
abi-core run --build        # Rebuild containers first

---

Built-in Agents

When you create swarm, you get these out of the box:

Agent	Role
Orchestrator	Entry point. Triage + Guardian gate + routes to Planner
Planner	Decomposes queries into structured task plans
Builder	Spawns ephemeral containers from task specs
Zombie	Ephemeral executor — runs task, uploads artifacts, self-destructs

The pipeline: User → Orchestrator → Planner → Builder → Zombie → Artifact → Done.

---

Multi-Provider LLM

# Ollama (local, default)
{"provider": "ollama", "model": "qwen3:8b"}

# OpenAI
{"provider": "openai", "model": "gpt-4o", "api_key": "..."}

# Anthropic
{"provider": "anthropic", "model": "claude-sonnet-4-20250514"}

# AWS Bedrock
{"provider": "bedrock", "model": "anthropic.claude-3-sonnet"}

# Azure OpenAI
{"provider": "azure", "model": "gpt-4o", "endpoint": "..."}

---

Security & Governance

• Guardian Agent — validates every request against OPA policies before execution
• HMAC Authentication — inter-agent calls are signed and verified
• Semantic Access Validation — agents can only call tools within their access_scope
• Audit Trail — all decisions logged with risk scores
• Human Veto — Plan Confirmation allows blocking execution before it starts

---

Project Structure

my-swarm/
├── agents/
│   ├── orchestrator/     # Triage + routing
│   ├── planner/          # Task decomposition
│   ├── builder/          # Container spawning
│   └── my-agent/         # Your custom agents
├── services/
│   ├── semantic_layer/   # Weaviate + MCP tools
│   └── guardian/         # OPA policies
├── compose.yaml
└── .abi/runtime.yaml

---

Documentation

Full docs: https://abi-core.readthedocs.io

---

Contributing

git clone https://github.com/Joselo-zn/abi-core
cd abi-core-ai
uv sync --dev
uv run pytest

---

License

Apache 2.0 — see LICENSE

---

Built by José Luis Martínez