AI Sparkle
AI that knows your notebooks, your schema, and your SQL.
Not a generic code assistant bolted onto an editor — AI Sparkle is built against SAMSKARA's own execution model, catalog, and API surface.
Notebook-aware
AI Sparkle sees the current cell, live session variables, and prior cells in the session — suggestions reuse the variables you already defined instead of re-deriving them.
SQL-aware
Text-to-SQL in SQL Workbench is DuckDB-dialect aware and understands ArrowLake's catalog structure, so generated queries target the right profile and namespace.
Schema-aware
Your ArrowLake table schemas are part of every request's context — no need to paste column names into a prompt.
Learns from previous successful executions
Successful notebook runs and SQL queries are indexed per-org into ChromaDB, so the model picks up your team's own naming conventions and patterns over time.
Local LLM support
Point AI_BASE_URL at any OpenAI-compatible endpoint — including a local Ollama instance — to keep every prompt and completion inside your own network.