OpenIntelligence

Private neural document answers on Apple Silicon. Import PDFs, get cited answers, and sync across iPhone, iPad, and Mac.

OpenIntelligence is a dedicated, on-device document search and citation-backed Q&A engine for iPhone, iPad, and Mac.

IMPORTANT: This is not a conversational chatbot, creative writing assistant, or internet-connected search agent. It is a professional productivity tool engineered to index, search, and answer questions about your specific offline documents, files, and transcripts.

Why OpenIntelligence is unique:

• 100% Offline: Indexing, OCR, transcription, and reasoning run entirely on Apple Silicon. Functions in Airplane Mode with no cloud dependencies, subscriptions, or data leaving your device.
• Citation-Backed Answers: Every response is generated from matching passages. Claims are verified against source texts; if a fact is missing, the engine states it cannot verify instead of hallucinating. Tapping a citation jumps to the exact page and highlight.
• Isolated Libraries: Keep projects, clients, or classes completely isolated in distinct local or iCloud-synced containers.

Engineering & Architecture (Docs Reference): Built from the ground up to leverage the local hardware capabilities of Apple Silicon:

• Lexical & Semantic Hybrid Search: Combines local BM25 keyword matching with semantic vector databases. Vector operations utilize Apple's Accelerate framework (BNNS and vDSP_mmul matrix computations) directly on the Neural Engine and GPU.
• Local CoreML Inference: Runs on-device LLM reasoning using CoreML, optimizing memory and latency on Apple Silicon without cloud API endpoints.
• Verification Gate: A strict math-based validation layer inspects retrieved text fragments, strips out ungrounded assertions, and cancels speculative answers before displaying them.
• Hardware Telemetry HUD: Real-time system monitoring overlays SoC thermal outlines, active Neural Engine/GPU utilization, and memory layout configurations.
• Private Cloud Compute Compatibility: Securely integrates with Apple's Private Cloud Compute architecture for advanced context lengths exceeding on-device chip capacity (handled by the OS).
• Resilient iCloud Sync: Custom multi-device sync logic utilizing deletion tombstones, conflict resolution merging, and concurrent background file downloading using Swift TaskGroups.

How the workflow works:

1. Setup: No sign-up, logins, or cloud accounts.
2. Ingestion: Import files from iCloud/other apps or scan pages. OpenIntelligence extracts digital text, performs on-device OCR on scanned pages, transcribes audio/video locally, and builds vector indexes.
3. Querying: Query in natural language using local reasoning modes:
   • Standard: Fast, direct lookups (names, dates, figures).
   • Deep Think: Multi-step reasoning across multiple files or separate pages.
   • Maximum: Exhaustive CPU and Neural Engine power for complex synthesis.

Supported Formats:

• Documents: PDF, DOCX, XLSX, PPTX, Markdown, CSV, RTF, TXT, HTML, JSON, XML, YAML, and code.
• Images: JPEG, PNG, HEIC, TIFF, GIF.
• Audio & Video: MP3, M4A, WAV, AIFF, MP4, MOV.

Performance & Limitations:

• Strengths: Finding facts, specifications, code references, and citations across manuals, legal documents, and research papers offline.
• Limitations: Parsing corrupted/legacy PDFs, poor handwriting in scans, iWork formats. Converting these to clean PDFs or TXT files first is recommended.

Subscription Tiers:

• Free: Evaluation mode with limited document ingestion.
• Pro: Unlocks unlimited documents, larger libraries, and deeper context-packing.
• Lifetime: One-time purchase to permanently unlock all features on all your devices.

Privacy: No user tracking, no telemetry, and no remote server connections. Your data remains strictly yours.

Terms of Use (EULA): https://www.apple.com/legal/internet-services/itunes/dev/stdeula/ Privacy Policy: https://gunzino.me/openintelligence/privacy