Introduction

Overview of the Sapientia local-first AI orchestration platform, inclusive vision, core architecture, and privacy model.

Sapientia is a local-first agentic AI orchestration platform designed to run Large Language Models (LLMs) natively on your personal computer. All inference computations, vector memory storage, and tool executions operate locally on your hardware without transmitting user conversations or private documents to third-party servers.

Sapientia is built to open AI access for everyone-from students, educators, and creators to software developers and organizations. Whether writing, studying, analyzing data, or building autonomous workflows, anyone can turn their daily laptop or desktop into a private, high-performance AI workspace that remains fully dependable offline without recurring API token bills.


Philosophy & Core Values

The name Sapientia originates from Latin, meaning Wisdom. This principle serves as the primary guiding compass in the development of Sapientia, grounding our commitment to:

  1. User Privacy (Privacy): Operating on strict zero-telemetry standards. Conversation histories, indexed documents, and vector embeddings remain securely on your local storage.
  2. System Security (Security): Code execution, tool invocations, and app prototyping run inside sandboxed boundaries with transparent permission checkpoints.
  3. Granular Control (Control): Users retain absolute sovereignty over chosen model weights, inference parameters, agent graphs, and data lifecycles.
  4. Tangible Utility (Utility): Engineered to run smoothly on standard consumer hardware (from 8GB RAM laptops to dedicated GPU rigs) with practical, day-to-day productivity benefits.
  5. Long-Term Sustainability (Sustainability): Fostering technological independence through cloudless architecture, reducing operational costs, and maintaining dependable full-offline availability.

AI-Generated Multi-Language Support

Sapientia includes dynamic multi-language generation powered directly on-device by the local AI runtime:

  • Intuitive Feature Understanding: Users can explore application capabilities, construct complex workflows, and receive guidance in English, Indonesian, or custom regional dialects.
  • Zero Cloud Translation Dependencies: Contextual translation and multilingual reasoning execute on local model weights, ensuring zero external exposure for international or multilingual data.
  • Adaptive Communication Styles: The local model seamlessly adjusts explanations from concise technical specifications to accessible overviews tailored to beginners.

Core Capabilities

Sapientia unifies several workflow modules into a single local workstation:

1. Visual Workflow Builder

Construct multi-step AI automations on a visual graph-based canvas. Connect Start, Agent, Condition, Decision, Approval, and Loop nodes using interactive text editors with @mention variable referencing.

2. Hybrid RAG (Memories & Knowledge Base)

An adaptive retrieval system combining:

  • Vector RAG: Semantic similarity search over local documents (PDF, DOCX, TXT) indexed into an embedded vector store (LanceDB).
  • Web RAG: On-demand internet search via Google Custom Search API, feeding fresh search results into local model processing.

3. App Builder

Build production-ready interactive forms with drag-and-drop components, reusable templates, and prompt mapping. Turn user inputs into structured workflows without writing frontend code.

4. MCP Tools Management

Adopt the open Model Context Protocol (MCP) standard across Stdio (CLI subprocesses) and Streamable HTTP transports to integrate local models with filesystems, databases, and external APIs.

Direct peer-to-peer (P2P) connection pairing the desktop host with mobile or web clients via QR code without cloud intermediaries, enabling remote local AI interaction and App Form submission directly from mobile devices.

6. Local API Server

Expose a local HTTP server compatible with the OpenAI API format (/v1/chat/completions), simplifying integration with developer tools, IDE extensions, and existing API clients.