RangerIO Guide

2. Features & Capabilities

Overview

RangerIO is an AI Safety Gateway: a local-first workbench that profiles your data, scans for sensitive content, and lets you bring cloud AI in through a governed handoff when you want it.

Data Ingestion

Supported Formats

Structured Data:
- CSV, TSV (any delimiter)
- Excel (.xlsx, .xls)
- JSON, JSONL - Nested objects supported
- Parquet - High-performance columnar format

Databases:
- PostgreSQL
- MySQL / MariaDB
- Microsoft SQL Server
- SQLite

Documents:
- PDF - Text extraction with layout preservation
- Microsoft Word (.docx)
- Plain text files
- Markdown

APIs:
- REST endpoints
- JSON data sources
- Optional bearer-token auth

Import Features

AI-Powered Analysis

Natural Language Queries

Ask questions about your data in plain English:
- "What are the top 10 customers by revenue?"
- "Show me all transactions from last quarter with amounts over $5,000"
- "Summarize the key trends in this dataset"
- "Are there any unusual patterns in the sales data?"

RAG Modes

Basic Mode - Fast, simple queries
- Quick lookups and filtering
- Single-document context
- Best for straightforward questions

Assistant Mode - Balanced analysis
- Multi-document reasoning
- Context-aware responses
- Recommended for most use cases

Deep Search Mode - Comprehensive analysis
- Searches across all data sources
- Multi-hop reasoning
- Best for complex questions requiring synthesis

Supported AI Models

The Full edition bundles a local LLM appropriate for your hardware tier. The Lite edition lets you bring your own — drop any GGUF-compatible llama.cpp model into the models directory. Current families known to run well: Qwen3, Qwen3.5, Phi-3.5-mini, and DeepSeek-R1 distilled variants.

Requirement Details
Format .gguf files (GGUF format)
Backend llama.cpp via llama-cpp-python
Quantization Any GGUF quantization (Q4, Q5, Q8, IQ, F16, etc.)
Min RAM 16 GB system RAM (Lite); 32 GB or Apple Silicon 16 GB+ unified (Full)
Context Window Auto-detected from GGUF metadata
Model Size ~1B to 14B parameters for typical desktops; larger if hardware allows
GPU Optional (Metal on Apple Silicon, CUDA on NVIDIA, Vulkan for AMD on Windows)
Cloud Models Via Stage 2 connector (Claude Desktop today; ChatGPT and Perplexity in named phases)

Data Quality

Automatic Profiling

When you import data, RangerIO automatically generates:

Statistical Profile:
- Row and column counts
- Data types and formats
- Min, max, mean, median for numbers
- Most/least common values
- Distribution histograms

Quality Metrics:
- Completeness (% non-null values)
- Uniqueness (duplicate detection)
- Validity (format compliance)
- Consistency (cross-field checks)

Quality Checks

Completeness Checks:
- Missing value detection
- Required field validation
- Null pattern analysis

Accuracy Checks:
- Format validation (emails, phones, dates)
- Range checks (min/max bounds)
- Referential integrity

Consistency Checks:
- Cross-field validation
- Business rule compliance
- Standardization checks

Uniqueness Checks:
- Duplicate detection
- Primary key validation
- Fuzzy matching for near-duplicates

Privacy Scan

Every imported file runs through an entity-aware scan that finds sensitive content and tiers each finding by confidence:

Detected entity classes:
- Names and personal identifiers
- Addresses (physical and email)
- Phone numbers
- Government IDs (SSN, passport, driver's license, etc.)
- Financial identifiers (credit cards, account numbers)
- Dates, locations, and organizations
- Custom patterns (user-defined)

Confidence tiers:
- Confirmed - High-confidence sensitive value
- Suspected - Likely sensitive; review recommended
- Dismissed - Filtered as a structural false positive

Actions per column:
- Leave as-is
- Mask in place (XXX-XX-1234)
- Replace with stable stand-ins
- Remove the column
- Override a finding with one click on the column dot — overrides persist across rescans

Data Preparation

Prepare Wizard

Guided workflow for data cleaning and transformation:

Step 1: Intent Capture
- Describe what you want to achieve
- AI suggests appropriate transformations

Step 2: Scope Selection
- Choose data sources
- Select specific tables/columns

Step 3: Consolidation
- Merge data from multiple sources
- Join tables with AI assistance

Step 4: Quality Actions
- Review quality issues
- Apply fixes (fill missing, remove duplicates, standardize)

Step 5: Sensitivity & PII
- Review detected PII
- Apply masking or removal

Step 6: Export Configuration
- Choose output format
- Configure export options

Data Transformations

Column Operations:
- Rename, reorder, remove columns
- Change data types
- Split/merge columns
- Extract patterns (regex)

Row Operations:
- Filter by conditions
- Remove duplicates
- Sort and rank
- Sample (random, stratified)

Value Operations:
- Fill missing values (mean, median, forward-fill)
- Standardize formats (dates, phones, addresses)
- Replace values (find & replace)
- Normalize/scale numbers

Advanced Operations:
- Pivot and unpivot
- Group and aggregate
- Window functions
- Custom expressions

Natural Language Transformations

The query layer can drive transformations from plain-English prompts. Examples:
- "Convert all dates to YYYY-MM-DD format"
- "Remove rows where revenue is negative"
- "Create a new column for profit margin"

Export Capabilities

Supported Formats

Structured:
- CSV (configurable delimiter, encoding)
- Excel (.xlsx) - Multiple sheets
- JSON, JSONL
- Parquet
- SQL INSERT statements

Reports:
- PDF reports with charts
- HTML dashboards
- Markdown summaries

Databases:
- Direct export to PostgreSQL, MySQL, SQL Server
- Batch insert optimization

Export Options

Performance & Scalability

Data Size Limits

Hardware Acceleration

GPU Acceleration:
- NVIDIA CUDA - Automatic detection and optimization
- AMD on Windows via Vulkan
- Apple Metal - Optimized for Apple Silicon (M1+)
- CPU fallback always available

CPU Optimization:
- Multi-threading for all operations
- SIMD instructions (AVX2, AVX-512)
- Optimized memory management

Memory Management:
- Streaming for large files
- Incremental processing
- Automatic garbage collection

Security & Privacy

Local-First Architecture

Data Protection

Compliance

RangerIO's architecture is designed for regulated workflows: local-first processing, governed cloud handoff, opt-in connectivity, no telemetry. Common frameworks it can support:
- HIPAA (Healthcare)
- SOX (Financial)
- GDPR / CCPA (Privacy)
- PCI-DSS (Payment data)
- Export-controlled and government workflows

"HIPAA-ready" or "SOC 2-certified" are certifications that belong to your organization and your audit posture, not to a piece of software. Use RangerIO as part of your compliance program; it does not substitute for one.

Integration

Import Methods

Export Targets

User Interface

Modern Web Interface

Key Screens