data-science-agent-loop

What problem does it solve?

Data scientists often need short-lived, disposable loops that batch data, dispatch tool calls to a cloud LLM, and collect per-tool audit trails, without building full agents or servers.

Core Features & Use Cases

• Single-folder runnable scripts with optional helpers for HTTP calls, batching, and optional parallelism.
• Batch data rows, route tool outputs back to the main process, and emit JSONL audit logs for reproducibility.
• Use cases include data cleaning, enrichment, geospatial routing, and dataset compilation with batched tool calls.

Quick Start

Run the main loop script in the folder, then supply an Ollama model and batch settings via .env and watch batches produce outputs and JSONL audits.