KikoLens¶
Private, local-first data analysis. Load a dataset, let the autopilot take a first pass, clean it, explore it, test hypotheses, compare many models and understand them, forecast, study customers and the effect of a change, watch a data source over time, and keep what you built: a recipe you can replay, a model you can score new data with, a panel and a report. Nothing leaves your machine; the optional language-model features use Ollama running locally.
Start here¶
pip install "kikolens @ git+https://github.com/Francisco-B-O/kikolens.git"
kikolens doctor # check the installation
kikolens start # open the dashboard, then pick "Or try an example" or load your own file
| If you want to... | Read |
|---|---|
| Click through the app | Dashboard guide |
| Script it or run it on a schedule | Command line and pipelines |
| Use it from a notebook | Python API and Jupyter |
| Follow a complete example | Workflows |
| Know how a result is computed and when not to trust it | How it works |
| Change settings (Ollama, port, language) | Configuration |
| Contribute | Development |
What is in the box¶
- Data: home page, data story, chart gallery, quality and cleaning (undoable, as replayable recipes), privacy tools, record matching, transformations, an autopilot and an analyst that chats with your data.
- Analysis: statistical tests with effect sizes, sequential and Bayesian A/B tests, anomalies, drift, segmentation, "why did it change", causal effect and survival.
- Models: a model lab (cross-validation, tuning, explanations, error analysis, conformal ranges), regression, neural networks, time series, saved models, model health and monitoring.
- Learn: a concept lab of simulations and a bilingual glossary with reading guides.
- Share: panels, HTML and PDF reports, Excel, notebooks and single-file projects.
- Advanced: text, customers, impact, recommendations, surveys, networks, geography, model governance and synthetic data.
