OVERVIEW

Verdaca docs

DESIGN PREVIEW

This page describes the planned verdaca Python package. That package is not built yet, so the code samples on this page will not run. For what runs today, see the real Getting started page and the recorded receipts in docs/receipts/.

This documentation covers Verdaca's deliberation platform — how it is architected, how to run your first session, and how to extend it with custom agents and adapter substrates. It is written for practitioners: consultants, engineers, and technical leads who need to understand the system well enough to configure and deploy it.

ADVISORY

Start with the Conceptual model before touching configuration — the two-axis pluggability diagram is the key to understanding every other section.

What's in these docs

Quick start

Three lines to run a deliberation session. Install the package, initialize a Studio, and pass a workflow name.

# Install Verdaca
pip install verdaca

# Initialise a session
from verdaca import Studio
session = Studio.create()
result  = session.run("[workflow-placeholder]")

Architecture at a glance

Verdaca is organized on two independent axes: swap the agent personas horizontally, swap the substrate adapters vertically. Neither axis requires changing the other.

Every request reaches the kernel through one channel-neutral gateway, after an auth-first spine of four checks — signature, identity, replay, and budget. See the front door in the conceptual model.

Verdaca two-axis pluggability architecture Plug-in personas — UX Expert, Data Scientist, Legal Counsel, Security Architect — connect into the Verdaca Kernel. The kernel connects to substrate-neutral adapter groups: Memory (Mem0, Letta), LLM Provider (Anthropic, OpenAI), and Compaction and Output (Forge, Caveman). HORIZONTAL AXIS — PLUG-IN PERSONAS UX Expert Data Scientist Legal Counsel Security Architect VERDACA KERNEL Deliberation loop · synthesis · hashed receipt VERTICAL AXIS — SUBSTRATE-NEUTRAL ADAPTERS MEMORY LLM PROVIDER COMPACTION & OUTPUT Mem0 Letta Anthropic OpenAI Forge Caveman