⚡ 100% CPU NATIVE • ZERO-GPU • MATHEMATICALLY GROUNDED
🪟 Windows 11 Build In Development — Early Buyers Receive Free Upgrade

Document Knowledge Discovery Without Hallucinations or GPUs

Docula is a deterministic Bayesian NLP workstation. Ingest thousands of PDFs, Word files, and spreadsheets into a high-throughput Compressed Sparse Row knowledge graph with exact line-level citation proof.

Download Docula Workstation →
✔ Linux ELF Standalone Ready Now ✔ Free Windows 11 Key Upon Launch ✔ 100% Offline & Air-Gapped ✔ One-Time Purchase
docula.py — Q&A Inference Engine & Deep Grounding
Docula Bayesian NLP & Document Synthesis Workstation Main Interface

🪟 Windows 11 Standalone Edition COMING SOON

We are currently packaging the native 64-bit Windows 11 distribution—featuring native threading, zero-install execution, and full multi-sheet Excel & PDF ingestion. Purchase the Linux release today and receive full access to the Windows 11 installer immediately upon launch at zero extra cost.

Reserve Both Platforms →

Deterministic Rigor vs. Black-Box Probabilities

Why quantitative systems engineers, researchers, and enterprise analysts choose Docula over stochastic tensor embeddings.

❌

Transformer LLMs & Vector Databases

  • ✕ Hallucinations & Fabrications: Generative models guess words based on soft weights, producing false facts with confident phrasing.
  • ✕ Severed Grounding: Vector similarity cosine distance cannot tell you the exact sentence, line number, or file offset where the proof resides.
  • ✕ Heavy Hardware Demands: Requires high-end NVIDIA GPUs with massive VRAM, complex CUDA toolchains, or expensive cloud API tokens.
  • ✕ Privacy Leaks: Proprietary internal documents, contracts, and trade secrets are uploaded to third-party endpoints.
✔

Docula Bayesian CSR Engine

  • ✓ 100% Deterministic Mathematical Grounding: Calculations derive from directed bigram syntax models and Positive Pointwise Mutual Information (PPMI).
  • ✓ Click-to-Line Source Citations: Every inference is backed by exact file coordinates. One click jumps directly to the cited line in the built-in reader.
  • ✓ Bare-Metal CPU Multi-Threading: Zero GPU requirement. Ingests 100k+ tokens/sec on standard Intel/AMD processors using lock-free parallel execution.
  • ✓ 100% Air-Gapped & Private: Runs entirely on your host OS. No telemetry, no API keys, no external network calls, and zero subscription rent.
DIRECT SOURCE VERIFICATION

Interactive Context Viewer with Line-Exact Navigation

Never take a summary at face value. Docula couples its Bayesian evidence scoring directly with an interactive document viewer.

Clicking any ranked citation immediately opens the full original file, displays line gutters, and jumps straight to the target line with synchronized bi-directional search (Next / Prev).

Line-Numbered Gutters Bi-directional Search Jump to Cited Line Sub-string Highlighting
Docula Viewer — gatsby.txt (Line 1727)
Docula Context Document Viewer with Yellow Cited Line Highlight
ENTERPRISE KNOWLEDGE MANAGEMENT

Structured Project Isolation & Q&A Audit History

Organize your investigations into independent project workspaces. Switching projects instantly resets the view and loads a fresh, unbranded suggestion pool mined directly from that project's graph.

Every submitted inquiry, synthesized conclusion, and score-ranked citation set is committed to a local JSON history database for auditability and rapid recall.

Multi-Project Workspaces Timestamped Query Audits Full Citation Logging Clean Workspace Transitions
Docula — Q&A History Database
Docula Q&A History Database Audit Log

Architected for Production Systems

Engineered from first principles without bloated dependencies or brittle cloud APIs.

01

Multi-Tier PDF Engine

Ingests PDFs from modern PDF 1.7 back to legacy 1990s PDF 1.1. Combines Poppler utilities, pypdf non-strict parsing, stream decompressors, and raw PostScript operator tokenizers.

02

Multi-Hop Causal Traversal

Identifies indirect causal links (A → B → C) across disconnected documents where concepts never appear in the same sentence, maximizing joint transition path weights.

03

Contrastive Differential Analysis

Evaluates set differences across statistical neighborhoods (A vs. B) to clearly articulate how competing entities, products, or clauses differ in practice.

04

Word & Excel Normalization

Processes Word (.docx) through direct XML run parsing and translates complex multi-sheet Excel (.xlsx, .xls) workbooks into normalized flat CSV with sheet annotations.

05

Compressed Sparse Row (CSR)

Stores the entire Bayesian transition graph in flat contiguous memory arrays without pointer indirection, unlocking multi-threaded CPU parallel processing at hardware bandwidth limits.

06

Zero Telemetry • Air-Gapped

Engineered for mission-critical legal discovery, financial forensics, defense analysis, and private IP research. No data packets ever leave your host system.

$ docula --system-architecture-profile
Operating Systems:Linux x86_64 (Ready) • Windows 11 (Soon)
Inference Engine:C++20 Parallel Lock-Free Ingestion
Graph Formulation:Pointerless CSR / Symmetric PPMI
Acceleration:AVX2 / Multi-Core Threaded (No CUDA)
Supported Formats:PDF (1.1-1.7), DOCX, XLSX, XLS, CSV, TXT
Licensing:Commercial Perpetual Workstation License
Docula Startup Engine Splash Screen

Own Your Document Intelligence Stack

Get immediate access to the compiled Docula workstation distribution, automated installation scripts, and lifetime air-gapped execution on Linux and Windows 11.

Purchase Docula on Gumroad →
Instant Linux download today • Free upgrade to Windows 11 build upon release • Zero subscriptions