SuperAPI AI Directory
Partner with us中文AI newsFavorites
All toolsAI codingPrecis

Precis

Local-first data quality workbench for Excel/CSV: validate against schemas and 10 zero-code constraint types (uniqueness, ranges, foreign keys, allowed values and more), with row-level error reports. Also ships a v

Open website

About this tool

Purpose

Precis is a quality validation tool for Excel / CSV / TSV tabular data, positioned as a local-first visual data quality workbench. Its core goal is to help users complete the full validation workflow from data ingestion and cleaning/transformation to multi-dimensional quality checks without uploading data to any external service. The project is currently in Beta; the official description states that the core features are stable, the V2 configuration format and CLI/API contracts are frozen (additions only, no removals), and it is recommended to keep backups of production-critical data.

Main Capabilities

Visual Orchestration

Build validation models on a canvas by dragging nodes and connecting flows, with no coding required. The flow covers data ingestion, cleaning/transformation, and multi-dimensional quality checks.

Check Rules and Data Transformations

Provides 10 check rules: required validation, uniqueness, referential integrity, allowed value lists, numeric ranges, conditional logic, custom scripts, character sets, date logic, and multi-rule combinations. Provides 22 data transformations: string splitting, pattern extraction, mathematical calculations, grouping and aggregation, filtering, sorting, and more.

Large File Support

Very large files are automatically processed in chunks.

AI Assistant

Can connect to domestic and international large models or local Ollama, infer schemas and generate constraints in natural language during conversation, and after confirmation automatically write them in and sync to the canvas. Built-in presets for providers such as OpenAI, Anthropic, Gemini, DeepSeek, Tongyi Qianwen, Zhipu, Kimi, MiniMax, and Xiaomi MiMo, and local Ollama is also supported. Before each write to disk, a change list is presented, and after confirmation it is applied to the project and synced to the canvas. It is also available in the CLI (ai chat / ai generate). AI features are optional: when no provider is configured, validation and other features are unaffected, and all validation still runs locally.

MCP Server

The repository implements a Model Context Protocol (MCP) server based on the official MCP Python SDK and using stdio transport. AI coding assistants that support MCP (Claude Code, Cursor, Kimi Code, etc.) can directly call the validation engine after configuration.

Local Operation and Bilingual Interface

Data never leaves the local machine; a Chinese-English bilingual interface is provided; the desktop application has a built-in runtime environment and is ready to use after installation.

Usage

The officialprovides three usage modes:

  1. Desktop application (recommended): graphical interface, ready to use out of the box.
  2. Command line: suitable for batch execution, or for invocation by AI coding assistants (Kimi Code / Claude Code, etc.). You can directly pip install precis-cli get precis commands; for how to connect AI coding assistants, see integrations/.
  3. HTTP interface: for integration into your own systems.

Regarding installation, if the system default python3 is lower than 3.12 (macOS ships with 3.9), it can be installed via Homebrew. The one-click script will sequentially complete: Python environment detection, virtual environment creation, dependency installation, and frontend and desktop application builds.

Note: for any npm script that calls python (dev / backend:dev / cli / electron:dev), you must first run source backend/.venv/bin/activatebefore execution, otherwise the system's built-in lower-version Python will be used. For more startup methods and script descriptions, see scripts/README.md and tui-rust/README.md. Normal output of validation results indicates that the environment is ready.

In the CLI, you can first configure the provider, then execute ai chat(Agent mode) or ai generate data.xlsx Generate configuration directly from data files.

The MCP client configuration uses the mcpServers format, which is common to Claude Code / Cursor and similar tools.

Use Cases

  • Scenarios that require quality validation of Excel / CSV / TSV tabular data;
  • Users who want to orchestrate validation workflows in a visual, zero-code way;
  • Scenarios that require processing sensitive data locally and do not want data to be sent out;
  • Developers who need to run validation in batches, or want an AI programming assistant to call the validation engine;
  • Scenarios that require integrating validation capabilities into their own systems (HTTP interface).

Limitations and Notes

  • The project is in Beta, and the official recommendation is to keep backups of production-critical data.
  • AI features are optional and require configuring a provider yourself; if not configured, validation and other features are unaffected.
  • Regarding fees, pricing, specific platform support scope, performance metrics, etc., the materials do not currently provide information.
  • The project uses the Apache-2.0 license; see LICENSE_NOTICE.md for details.

Project Structure

Details for each subproject: backend, frontend, electron, tui-rust, scripts, e2e. Developers can refer to CONTRIBUTING.md (development commands and packaging instructions), AGENTS.md and docs/ARCHITECTURE.md (architecture details), CHANGELOG.md (changelog), SECURITY.md (security notes).

Related tools

Browse category