Local-First AI Data Analysis System
This implementation creates a completely local data analysis workflow where OpenClaw orchestrates multi-step analytics while Ollama provides local LLM reasoning. The system processes CSV datasets and supporting documents entirely on-device without any cloud API calls.
Architecture and Workflow
The system consists of three core components working together:
web_assistant.py): Handles file uploads, creates run directories, and sends slash commands to OpenClawmain.py): Performs actual data processing, column inference, chart generation, and insight creationTechnical Implementation
Users upload CSV files through a web interface, which triggers OpenClaw to execute a comprehensive analysis workflow. The system automatically infers relevant columns, generates trend visualizations, and produces three key outputs: trend_chart.png for visual insights, analysis_report.md with detailed findings, and tool_trace.json for workflow transparency.
The configuration uses Ollama's qwen3:8b model running locally on port 11434, with OpenClaw configured to use this endpoint instead of external APIs. The setup includes Docker-based sandboxing for secure code execution and workspace skills for file management.
Privacy and Security Benefits
This architecture ensures complete data privacy - sensitive financial reports, legal documents, or customer records never leave the local machine. The system works offline once models are downloaded and provides full transparency with inspectable execution traces. All processing, from data ingestion to final report generation, happens locally while maintaining agent-style multi-step execution capabilities.