What it is
DataJev is a Python CLI backend for data analysis agents. An LLM analyst writes and interprets Python analysis steps, while Jev controls the analytical trajectory. It is for developers building inspectable data analysis loops.
How it uses Jev
Jev reads a compressed analytical state (goal, schema, recent results, retained insights, trajectory) and returns typed decisions: CONTINUE, SWITCH, VERIFY, or STOP. Each response also carries an insight-retention probability, a verification probability, and an analysis-completeness score, recorded alongside the action and its probability distribution.
Primitives:choicescore
Technique worth stealing
Separate analytical execution from trajectory control: let the LLM analyze, let Jev decide what to analyze next.
Try it
uv run datajev analyze examples/temporal_regimes.csv --goal "..." --controller jev --analyst llm
Evidence
Each line is one question put to Jev about the README. ≥ 0.60 reads as yes, ≤ 0.40 as no; in between Jev is not making a call.
- Jev-centricyes0.61
- Shows a System One patternyes0.70
- Handles uncertaintyno0.04
- Measuredno0.05
- Runnableunclear0.41
- Worth recommendingunclear0.43
- Model replicano0.39
- Problem scopescore on a 0–2 scale1.38
- About Jevyes0.94
Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 21 Sept 2026.