OpenClaw Data Importer Skill

The OpenClaw Data Importer is an AI-driven tool for ingesting unstructured or semi-structured data (CSV, JSON, XLSX) into Supabase. It uses an LLM to automatically infer relationships, generate relational schemas, and map data types, removing the need for manual ETL scripting.

Features

• Auto-Schema Generation: Reads a sample of your file and proposes a SQL schema.
• Intelligent Mapping: Transforms raw keys (e.g., "First Name") to database columns (e.g., "first_name") and casts types (strings to dates/booleans).
• Streaming & Batching: Handles large datasets efficiently by streaming files and uploading in batches.
• Supabase Integration: Direct UPSERT support via PostgREST.

Installation

npm install flexible-data-importer

Usage

CLI

Ensure you have a .env file with:

SUPABASE_URL=...
SUPABASE_KEY=...
OPENAI_API_KEY=...

Run the importer:

npx data-importer ./path/to/my-data.csv

Programmatic

import { UniversalImporter, NodeFileAdapter, OpenAILLMAdapter, SupabaseAdapter } from 'flexible-data-importer';

const importer = new UniversalImporter(
  new NodeFileAdapter(),
  new OpenAILLMAdapter(process.env.OPENAI_API_KEY),
  new SupabaseAdapter(process.env.SUPABASE_URL, process.env.SUPABASE_KEY)
);

await importer.execute('./large-dataset.xlsx');

Architecture

The system uses a UniversalImporter core that orchestrates three adapters:

1. FileAdapter: Abstracts filesystem access (Node streams vs OpenClaw sandbox).
2. LLMAdapter: Abstracts the intelligence provider (OpenAI vs OpenClaw Internal).
3. DatabaseAdapter: Abstracts the storage layer (Supabase vs Postgres).

This architecture allows the skill to be run standalone or injected with OpenClaw's native capabilities when running inside the agent runtime.