Show HN: fenic – LLMs as dataframe operators, query meaning and structure (github.com)

The newly announced fenic platform revolutionizes data processing by integrating AI operators directly into DataFrame pipelines, allowing users to handle both structured and unstructured data seamlessly. By utilizing operations like extract, classify, and summarize as first-class citizens within query models, fenic transforms documents and logs into structured, reusable workflows. This approach not only eliminates the need for brittle regex scripts and one-off prompts but also ensures that all data work results in durable artifacts that are easily inspectable and rerunnable. Fenic stands out by embedding inference within its query model, employing typed operators with schemas rather than manual orchestration. Users can leverage familiar PySpark/SQL syntax alongside semantic features, streamlining processes from data exploration to artifact creation. The platform's capabilities include row-level lineage tracking, per-query metrics, and automated integration with AI coding agents like Claude Code and OpenAI's Codex, enhancing collaboration between human and machine contributions to data tasks. By providing a structured methodology for working with diverse data types, fenic addresses the challenges of reproducibility and inspection, setting a new standard for AI-assisted data workflows in the AI/ML community.