> ## Documentation Index
> Fetch the complete documentation index at: https://hillock.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Hillock — Local Neuro-Symbolic Memory for AI Agents

> Hillock is a 100% local memory engine for AI agents. Zero hallucinations, under 1.2 GB VRAM, fully offline. No cloud, no GPU cluster required.

Hillock is a lightweight, three-tier memory engine for AI agents that runs entirely on your local hardware — no cloud APIs, no telemetry, no GPU cluster required. It replaces bloated vector databases with a relational SQLite Knowledge Graph, Hebbian synaptic memory, and Hyperdimensional Computing (HDC), extracting structured facts from your documents and blocking unanswerable questions mathematically before they ever reach the LLM. Built for AI engineers, local LLM developers, and privacy-focused researchers who need deterministic, auditable memory on consumer hardware.

<CardGroup cols={2}>
  <Card title="Quick Start" icon="rocket" href="/quickstart">
    Install Hillock, pull a local model, and run your first document query in under 5 minutes.
  </Card>

  <Card title="Architecture" icon="layer-group" href="/architecture/three-tier-system">
    Understand the three-tier design: SQLite Knowledge Graph, Hebbian plasticity, and HDC reservoir.
  </Card>

  <Card title="Python Library" icon="python" href="/guides/python-library">
    Embed Hillock directly in your Python application using the `IntegratedHillock` class.
  </Card>

  <Card title="API Reference" icon="code" href="/reference/api-reference">
    Explore the OpenAI-compatible REST API and connect any OpenAI client to Hillock's memory.
  </Card>
</CardGroup>

## What makes Hillock different

<CardGroup cols={3}>
  <Card title="Zero Hallucinations" icon="shield-check">
    An HDC cosine-similarity gate (threshold 0.55) mathematically blocks questions that don't match stored knowledge. Hillock refuses honestly instead of guessing.
  </Card>

  <Card title="Under 1.2 GB VRAM" icon="microchip">
    The entire pipeline — TALON extractor, HDC reservoir, and Hebbian memory — fits in less than 1.2 GB VRAM. CPU-only mode is supported for truly minimal hardware.
  </Card>

  <Card title="100% Offline" icon="lock">
    No data ever leaves your machine. Zero cloud dependencies, zero telemetry. Your documents and knowledge graph stay local at all times.
  </Card>

  <Card title="TALON Extraction" icon="bolt">
    The Tensor-Accelerated Local Ontology Network uses O(1) matrix classification to extract Subject-Predicate-Object triples from documents — no LLM pass required for ingestion.
  </Card>

  <Card title="OpenAI-Compatible API" icon="plug">
    Drop Hillock behind any OpenAI client, Open-WebUI, or AnythingLLM instance. Point your base URL to `http://localhost:8000/v1` and chat normally.
  </Card>

  <Card title="Dual-Licensed" icon="scale-balanced">
    Free under AGPL-3.0 for open-source use. Commercial licenses available for proprietary applications — starting at \$49/month.
  </Card>
</CardGroup>


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.