V2 Alpha Guide

CogniCoreAI V2 starts as an alpha track focused on additive runtime architecture. The primary objective is to introduce multi-agent foundations without breaking existing V1 usage.

Version

Current alpha version: 2.0.0a1

Compatibility Model

  • V1 behavior remains default.

  • V2 behavior is opt-in by passing a runtime mode to Agent.

V2 Runtime Modes

V2 introduces three orchestration modes:

  • supervisor: single coordinator role with optional tool calls.

  • planner: planner -> executor -> critic sequence.

  • blackboard: collaborative iterative roles (researcher, analyst, synthesizer).

Quick Example

from cognicoreai import Agent, OpenAI_LLM, VolatileMemory, CalculatorTool

agent = Agent(
    llm=OpenAI_LLM(),
    memory=VolatileMemory(),
    tools=[CalculatorTool()],
    mode="planner",  # opt-in V2 mode
)

print(agent.chat("What is 12 * 7?"))

V2 Components Added

  • protocol.py: runtime event and lifecycle envelopes.

  • policy.py: policy interfaces and decisions.

  • tool_runtime.py: policy-aware tool execution outcomes.

  • orchestrator.py: V2 multi-agent runtime implementation.

Testing Expectations

  • Keep V1 tests green as compatibility baseline.

  • Run V2 tests for mode behavior and policy/tool interactions.

Recommended commands:

# from repository root
set PYTHONPATH=src
pytest -q

Documentation Build

Windows:

cd docs
make.bat html

Unix/Linux:

make -C docs html