← Back to App
LISA

LISA Documentation

Semantic compression for governed, portable AI knowledge

What is LISA?

LISA (Language Intelligence Semantic Anchoring) is a semantic compression platform that transforms documents and conversations into machine-optimized JSON format that any AI can interpret identically. LISA achieves 10-20× token reduction while preserving complete semantic fidelity — with cryptographic proof of integrity.

How LISA Resolves Ambiguity

AI systems process raw text probabilistically, burning compute cycles resolving ambiguities before reasoning can even begin. LISA pre-resolves these ambiguities through semantic translation, resulting in deterministic interpretation rather than probabilistic guessing.

The 10 categories of ambiguity LISA resolves:

  1. Referential Ambiguity — "Fix it" becomes explicit entity references
  2. Temporal Ambiguity — "Earlier" becomes specific timestamps
  3. Priority Ambiguity — Implicit importance becomes explicit weighting
  4. Completion Ambiguity — Open loops get explicit state (done/ongoing/planned)
  5. Social/Emotional Noise — "Thanks!" and pleasantries stripped entirely
  6. Reasoning Collapse — The "why" behind decisions is captured, not just "what"
  7. Role Ambiguity — "I'll do this" becomes explicit ownership assignment
  8. Dependency Ambiguity — Linear conversation becomes relationship graph
  9. State Ambiguity — Stream of changes becomes snapshot of current state
  10. Relevance Ambiguity — Only semantically significant content survives
The result: AI spends 95% of its compute on reasoning, not 40% resolving ambiguity. This is why LISA-compressed content enables better AI performance, not just better storage.

The Problem LISA Solves

Every time you feed a conversation or document to an AI, it spends 40% of its compute cycles resolving ambiguities — figuring out what "it" refers to, when "later" means, which "file" you mentioned. This happens every single time, across every platform, wasting tokens and producing inconsistent interpretations.

The problem isn't the AI. It's the format.

Human language is optimized for humans — rich with context, pronouns, implied causality, and social lubricant. But AI systems process this probabilistically, leading to:

LISA's Solution: Translate Before Inference

LISA doesn't compress text — it translates human language into machine-executable semantics. Before any AI sees your content, LISA resolves all 10 categories of ambiguity and produces a deterministic JSON structure that every AI platform (Claude, GPT-4, Gemini, Grok, Mistral, DeepSeek, Copilot, Perplexity) reconstructs identically.

The result: Cross-platform portability with 10-20× token savings, 95% compute focused on reasoning (not ambiguity resolution), and cryptographic audit trails proving exactly what context the AI received.

Think of it as professional translation for AI. Just as a skilled translator doesn't just swap words between French and English — they stabilize meaning across linguistic boundaries — LISA stabilizes meaning across the human-AI boundary.

How It Works

1. Upload Your Content

Paste text or upload files (.txt, .md, .json, .log, .docx) — AI conversations, technical documents, meeting notes, research papers, or any unstructured text.

2. Select Semantic Anchors

Choose how many semantic anchors to extract (6-24 depending on your tier). More anchors = higher fidelity for complex documents.

3. AI Compression

LISA's semantic engine extracts decisions, insights, entities, relationships, and dependencies — compressing 60,000 words into ~700 words (10-20× reduction) while preserving complete meaning.

4. Download Governed JSON

Get a platform-agnostic JSON file with:

Choosing Your Anchor Count

Semantic anchors are the building blocks of LISA's compression. Each anchor captures a discrete concept with its context, relationships, and significance.

How many anchors do you need?

Note: LISA generates anchors automatically — you never write them manually. The AI analyzes your content and extracts the most semantically significant elements based on the count you specify.
Pro tip: More anchors = higher fidelity, but also slightly larger file size. For most conversations, 6-12 anchors is optimal.

Using Focus Areas

Focus Areas let you guide LISA's compression toward specific themes in your content.

How it works

When compressing, optionally specify what to emphasize (e.g., "technical decisions", "action items", "compliance requirements"). LISA will prioritize semantic anchors related to your focus while still capturing overall context.

Common focus areas:

Leave blank for balanced compression — LISA will extract anchors proportionally across all topics.

Using Your LISA Exports

LISA exports aren't just for continuity — they're queryable knowledge bases.

Once you have a LISA JSON export, you can:

1. Continue Conversations

Upload to any AI and say: "Reconstruct this and continue where we left off"

2. Query Compressed Content

Ask questions about the original content without re-uploading it:

3. Cross-Reference Multiple Exports

Upload several LISA files and ask:

The Magic: Your 60,000-word document is now a 700-word semantic map that any AI can query instantly — 10-20× faster and 10-20× cheaper than processing the full text.

Works with: Claude, ChatGPT, Gemini, Grok, Mistral, Ollama (local models)

Reconstruction fidelity: >95% semantic preservation across all platforms

Subscription Tiers & Limits

LISA offers four tiers to match your compression needs:

Tier Monthly Annual Daily Limit Max Anchors
Free $0 — 5/day 6
Pro $19 $79/year (Save $149!) 50/day 12
Team $59 $468/year (Save $240!) 80/day 12
Enterprise $199 $1,188/year (Save $1,200!) 200/day 24

What counts as a compression?

One semantic analysis of your conversation or document. Each compression generates a governed JSON export with cryptographic signatures.

All tiers include:

Report Types

LISA generates governance-verified reports from your compressed content:

FREE+ Quick Summary

Brief overview of key topics and main points. Perfect for quick reference.

PRO+ Executive Summary

High-level overview focusing on decisions, outcomes, and action items. Perfect for stakeholders who need the "what" without the "how."

PRO+ Technical Detailed

In-depth technical breakdown including implementation details, code snippets, architecture decisions, and dependencies. Ideal for developers and engineers.

PRO+ Detailed Analysis

Comprehensive report covering all semantic anchors, relationships, context, reasoning, and recommendations.

ENTERPRISE Custom Reports

Tailored report formats for specific business needs. Contact support to configure custom templates for compliance, audit, or domain-specific requirements.

All reports include: Dual-layer verification (anchor hashes + Merkle root), cryptographic proof of integrity, tamper-evident audit trails, and export as PDF or JSON.

API Access

Programmatic access to LISA's compression engine is available for Team and Enterprise tiers.

Authentication

All API requests require your license key in the Authorization header:

curl -X POST https://api.sat-chain.com/compress \
  -H "Authorization: Bearer YOUR_LICENSE_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "content": "Your document or conversation text here",
    "anchor_count": 12,
    "focus_areas": "technical decisions"
  }'

Core Endpoints

POST /compress — Compress content into LISA format

Request body:

Response: LISA JSON export with semantic anchors, action vectors, reconstruction protocol

POST /reconstruct — Expand LISA export back to full context

Request body:

Response: Reconstructed full-context text

GET /usage — Check your compression usage and limits

Response: Daily limit, remaining compressions, tier info

Rate Limits

Privacy & Security

Your data security and privacy are foundational to LISA.

Data Storage Architecture

PostgreSQL Database (Server-Side)

We store only what's necessary for the platform to function:

Your Browser (localStorage)

Your actual content lives here — under your control:

Key principle: We store hashes and metadata, not your actual content. Your conversations and documents stay in your browser unless you explicitly sync them.

What We DO NOT Store

Data Handling

Cryptographic Integrity

Every LISA export includes:

Compliance

Encryption

Chrome Extension — Available Now (v0.49.1)

The LISA Chrome Extension is live and makes compression effortless — capture conversations with one click, no copy-pasting required.

What It Does

Supported Platforms

Coming soon: Ollama local model support (offline compression with your own LLMs)

How to Use

  1. Install the extension from Chrome Web Store
  2. Visit any supported AI platform — A floating LISA button appears
  3. Click "Capture Conversation" — Extension saves the current chat
  4. Compress immediately OR sync to web app for reports
  5. Access your library — View all captured snapshots in the extension popup or web app
Privacy Note: Snapshots are stored locally in your browser by default. Using "Send to App" uploads snapshot metadata (not full content) to our server for report generation — you control when this happens.

Troubleshooting

Common issues and solutions:

"Rate limit exceeded" error

Cause: You've hit your daily compression limit for your tier.

Solution:

"AI can't read my LISA file"

Cause: The AI platform may have size limits or JSON parsing issues.

Solution:

"Can I share snapshots with team members?"

Yes! Use the Share feature:

"How long are snapshots stored?"

Forever (under your control):

"Can I export my data?"

Absolutely:

Extension not appearing on AI platforms

Solution:

Still need help?

Roadmap

What's coming next for LISA:

v0.50 (Q2 2026) — Enhanced Compression

v0.60 (Q3 2026) — Enterprise Features

v1.0 (Q4 2026) — Full Platform

On the Horizon