Developer & Analytics Tools
Tools that accelerate development and analytics, plus open models for safer AI. We publish technology we build ourselves and support integration, tuning, and system development.
plusq
BetaThe SQL client with an AI agent built in
plusq is a macOS SQL client for PostgreSQL, BigQuery, MySQL, Snowflake, and DuckDB. An AI agent that knows your schema sits beside you — drafting queries from plain language, explaining errors, and proposing optimizations. Bring your own API key (BYOK): your data never touches plusq servers.
An AI Agent That Knows Your Schema
- The agent reads your tables and drafts queries from plain language
- Error explanations and query optimizations, all in chat
- BYOK with Claude, OpenAI, Gemini, Azure, Bedrock, or local Ollama — your data stays in your stack
Multi-Database, One Workspace
- PostgreSQL, BigQuery, MySQL, Snowflake, and DuckDB in a single app
- Smooth scrolling through 1M+ rows, with per-column types, null rates, and histograms
- One-click charts and export to Excel, JSON, or Pandas
Built for Safe Operations
- Git-backed .sql files, with connections and secrets kept outside the repo
- Automatic BigQuery dry-runs surface scan size and cost before execution
- Read-only mode blocks destructive operations on production connections
PII-JA NER
Open ModelDetect Japanese PII inside your own environment
A BERT-based NER model that extracts 18 types of personal information from Japanese text, including names, addresses, phone numbers, email addresses, and IDs. We publish a quantized ONNX model under Apache-2.0 and support evaluation, tuning, integration, and system development.
Choose an architecture that keeps input text away from external APIs.
The quantized ONNX model can run in browsers, mobile apps, and edge environments as well as on servers.
- FORMAT
- ONNX · INT8
- SIZE
- 105.4 MiB
- LABELS
- 37 BIO
- LICENSE
- Apache-2.0
18 PII Entity Types
- Extract names, organizations, addresses, and contact details at token level
- Detect My Number, passport, driver's license, insurance, and other IDs
- Identify entity spans and types with 37 BIO labels
Run in Your Environment
- Local inference with ONNX Runtime or ONNX Runtime Web
- Architectures that keep input away from external APIs
- INT8-quantized distribution at approximately 105.4 MiB
Integration & Development
- Accuracy validation and evaluation design using business data
- Fine-tuning, additional training, and model optimization
- Masking, review workflows, and integration into existing systems