polar\_llama
Polars library for parallel LLM inference that also asks Jev per-row Noul, Choice and Score questions through a native Rust layer.
Overview
Polar Llama is a Python library for making parallel inference calls from Polars dataframes. It supports OpenAI, Anthropic, Gemini, Groq and AWS Bedrock, multi-message conversations, Pydantic-defined structured outputs returned as Polars structs, Rust-powered vector similarity and HNSW search, a DSPy-style prompt optimiser, dataframe-native tool calling against MCP servers or Python callables, and rule assessment with plausibility checks where Polars does arithmetic and the model finds unexpected edge cases.
It includes a native Rust inference layer for the TypeSafe System One API: typed, calibrated yes/no, choice and score questions answered per row, or, per the list summary, as one typed contract over a document. This lets data teams label, classify or check rows at scale inside a dataframe pipeline. It is MIT licensed.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Yes / no | Yes/no question answered for each row | yes · no |
| Choice | Choice question answered for each row | — |
| Score | Score question answered for each row | — |
How it works
- Load data into a Polars dataframe.
- Define Noul, Choice or Score questions.
- Run them per row in parallel through the Rust layer.
- Use typed probabilities as new columns.
Strengths
- Row-level decisions inside dataframe pipelines.
- Parallel Rust-backed calls.
- Many other providers in one library.
Caveats
- System One is one feature of a broad library.
- Requires a TypeSafe key for Jev calls.
At a glance
- Type
- Library
- Status
- beta
- Language
- Python
- Runtime
- library (Polars, Rust extension)
- Has code
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
A Polars library for parallel provider inference that also calls Jev per row as Noul, Choice, and Score questions, or as one typed contract over a document.
logicrw/awesome-jev-projects
Frequently asked questions
How does Polar Llama use Jev?
Through a native Rust layer that asks typed questions per row or over a document.
Which other providers does it support?
OpenAI, Anthropic, Gemini, Groq and AWS Bedrock.
Is it a dataframe tool?
Yes, it is built around Polars.
Related projects
dbt package whose macros run Jev Choice, Noul and Score over SQL values in DuckDB and ClickHouse models.
ProjectbetaSQL (Jinja)DuckDB extension that asks Jev about every row in SQL and returns answers as ENUM, DOUBLE or STRUCT columns.
PluginbetaC++Python and Node library that classifies Hugging Face dataset rows on typed Choice, Score and Noul dimensions using Jev.
ProjectbetaPython, TypeScriptTabular classifier that learns which slices of labeled rows and columns to send Jev or LAYA, stopping when it has enough.
ProjectresearchPythonPython library with a TypeScript port that turns Jev questions into functions over lists and DataFrames, with caching and escalation.
LibrarybetaPython, TypeScriptSutro's active-learning CLI that finds rows Jev is least sure about, asks you to label them, and uses GEPA to improve the definition.
CLIbetaPython