Organization repository on GitHub
Plugin for interacting with LLMs in Polars
- Forks
- 2
- Last pushed
- Updated Oct 2026
Repo stats from the GitHub API, cached Oct 2026.Homepage
Jev directory
CommunityJev Integrations & Adapters
Polar Llama
A Python library that runs LLM and Jev calls as Polars expressions: parallel inference across providers, Pydantic-validated outputs as struct columns, and a native TypeSafe System One layer for per-row decision questions.
- Stars
- 30
- Language
- Python
- License
- MIT
- Reading time
- 2 min
- Published by
- Community
- Author
- pnthn-ai
- Last verified
- Oct 2026
Used inJev for Data & Analytics
- Capabilities
- Open sourceRequires API key
Highlights
- Call pl.col('Questions').llama.inference_async(...) to fan parallel model calls out inside a Polars pipeline.
- The TypeSafe System One layer answers calibrated yes/no, choice, and score questions per row, landing as typed columns with confidence.
- Providers include OpenAI, Anthropic, Gemini, Groq, and AWS Bedrock; Pydantic models validate outputs as Polars Structs.
- Ships Rust-powered cosine, dot-product, and Euclidean similarity, HNSW approximate nearest-neighbor search, and provider-native prompt caching.
- execute_tool_calls runs every tool call of every row in parallel against an MCP server or a Python callable.
Quickstart
python
import polars as pl
from polar_llama import Provider
df = pl.DataFrame({'Questions': ['What is the capital of France?']})
df = df.with_columns(
answer=pl.col('Questions').llama.inference_async(
provider=Provider.OPENAI, model='gpt-4o-mini'
)
)Watch out
Copy reviewed against the repo README on 2026-10-15; it states no pricing, token limits, or benchmark numbers. The Jev path needs a TypeSafe API key, parallel inference bills against whichever provider key you configure, and setup details live in docs/TYPESAFE.md beyond what was captured.
Questions about this entry
Short answers drawn from the linked project or page.
A Python library for making parallel LLM inference calls with Polars. It has grown past that start: multi-message conversations, structured Pydantic outputs, vector search, prompt optimization, tool calling, and a TypeSafe System One layer.
Through TypeSafe System One, a native Rust inference layer for the TypeSafe API: typed, calibrated yes/no, choice, and score questions answered per row, or one Pydantic contract applied to every line of a document in a single request, returning typed columns with confidence you can threshold on.
pip install polar-llama for normal use, or maturin develop from source for development. Then import polars as pl and polar_llama, set your provider keys in the environment, and call methods through the fluent .llama namespace.
OpenAI, Anthropic, Gemini, Groq, and AWS Bedrock models, per the README, plus AWS credentials for Bedrock and provider-native prompt caching for Anthropic's 5-minute and 1-hour cache windows.
They arrive as ordinary typed columns: calibrated probabilities or choices you can threshold, or Pydantic-validated structs whose fields you reach directly with struct.field(...), so downstream Polars code keeps working.
Rust-powered similarity metrics, HNSW approximate nearest-neighbor search, a DSPy-style optimizer with Signature, Predict, BootstrapFewShot, and InstructionOptimizer, plus dataframe-native tool calling against an MCP server or a Python callable.
No pricing, token limits, or benchmarks are given in the captured README, and the Jev path requires a TypeSafe API key. The README points to docs/TYPESAFE.md for setup details it does not reproduce.
More Jev resources like this
A dbt package that evaluates SQL values with Jev through TypeSafe's hosted API or OpenRouter: macros expose Choice as a nullable label, Noul as a match probability and Score as a numeric rating, on DuckDB and ClickHouse.
Jev Integrations & Adapters#community#catalog-import#logicrw
Community
smithclay/dbt_jevPython
smithclay/dbt_jevPythonAlibaba's multi-agent platform, which implements native TypeSafe Jev classification through JevClassifierModel for binary, choice and score routing across agent pipelines. Python 3.11+, Apache-2.0, with a FastAPI agent service and web UI.
Jev Integrations & Adapters#community#multi-agent#framework
Community
agentscope-ai/agentscopePython
agentscope-ai/agentscopePythonA Jev harness for PostgreSQL: 41 semantic operators plus a native Rust extension evaluate row-level predicates with Noul questions and keep the returned probabilities for explicit decision thresholds, leaving joins and aggregation to SQL.
Jev Integrations & Adapters#community#postgresql#nl2sql
Community
Sheltercosmo/jev4pgPython
Sheltercosmo/jev4pgPythonA GitHub App that routes pull-request review work through Jev: it ingests the webhook, normalizes the PR context, gets a LOW, NORMAL or SPECIALIST decision, and a deterministic policy layer maps it to auto-approval, a review request or a check-run update.
Jev Integrations & Adapters#community#catalog-import#logicrw
Community
hexyedev/jevprPython
hexyedev/jevprPythonCustom ComfyUI nodes that use Jev for text interpretation and judgments — select candidates, evaluate conditions, score text and extract numbers — plus a node for picking local SKILL.md files.
Jev Integrations & Adapters#community#catalog-import#logicrw
Community
hndrr/comfyui-jevPython
hndrr/comfyui-jevPythonA governed execution layer and drop-in proxy between typed-decision models like Jev and callers such as Claude Code, Cursor/MCP and SDKs: raw probabilities become deterministic policy actions from ALLOW to DENY with a DuckDB audit trail.
Jev Integrations & Adapters#community#catalog-import#logicrw
Community
0xshikhar/jev-fusePython
0xshikhar/jev-fusePythonFrom the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.