🏋️ KaiserLift
A smarter way to choose your next workout: data-driven progressive overload
🎯 Never guess your next workout again — KaiserLift analyzes your training history and tells you exactly which exercise, weight, and rep range will give you the easiest PR.
✨ Quick Start
Installation
pip install kaiserlift
Or with uv (recommended):
uv pip install kaiserlift
Run the Web Interface
kaiserlift-cli
Then open http://localhost:8000 in your browser and upload your FitNotes CSV export.
Use as a Python Library
from kaiserlift import pipeline
# Generate interactive HTML with workout recommendations
html = pipeline(["your_fitnotes_export.csv"])
with open("workout_plan.html", "w") as f:
f.write(html)
CSV Data Format
Your CSV should follow the FitNotes export format:
Date,Exercise,Category,Weight,Reps
2022-09-14,Flat Barbell Bench Press,Chest,45.0,10
2022-09-14,Dumbbell Curl,Biceps,35.0,10
Why keep doing 10 rep sets? Are you pushing in a smart way?
The core idea I’m exploring is simple: I want a science-based system for determining what’s the best workout to do next if your goal is muscle growth. The foundation of this is progressive overload—the principle that muscles grow when you consistently challenge them beyond what they’re used to.
One of the most effective ways to apply progressive overload is by taking sets to failure. But doing that intelligently requires knowing what you’ve done before. If you have a record of your best sets on a given exercise, you can deliberately aim to push past those PRs. That history becomes your benchmark.
Now, there’s another dimension to this: rep range adaptation. If you always train for, say, 10 reps, your muscles can get used to that rep range—even if you’re still pushing for PRs. Switching things up and going for, say, 20-rep maxes (even with lighter weight) can stimulate growth by forcing the muscle to adapt to new challenges. Then, when you go back to 10 reps, you might find you’ve blown through a plateau.
To make this system more precise, I propose using a one-rep max (1RM) equivalence formula—a way of mapping rep and weight combinations onto a single curve. It gives you a way to compare different PRs across rep ranges. Using that, you can identify which rep range you’re weakest in—meaning, which PR has the lowest 1RM equivalent. That’s where your next opportunity lies.
For the 1 rep max we use The Epley Formula:
$$
\text{estimated_1rm} = \text{weight} \times \left(1 + \frac{\text{reps}}{30.0}\right)
$$
Here's how to operationalize it:
- Collect your full workout history for a given exercise—every weight and rep combo you’ve done.
- Calculate the Pareto front of that data. This is the set of “non-dominated” performances: the heaviest weights at each rep range that can’t be beaten in both weight and reps at the same time.
- For each point on the Pareto front, compute the 1RM equivalent using a decay-style formula.
- Identify the Pareto front point with the lowest 1RM equivalent—that’s your weakest spot.
- Now, generate a “next step” PR target: a new set that just barely beats that weakest point, by the smallest reasonable margin (e.g., +1 rep or +5 lbs). That becomes your next workout goal.
This method gives you a structured, data-driven way to chase the easiest possible PR—which is still a PR. That keeps you progressing without burning out.
You can extend this concept across exercises too. Let’s say it’s biceps day. Instead of defaulting to the same curl variation you always do, you can rotate in a bicep exercise you haven’t done recently in order to assure a new PR. This introduces variability, which is another powerful way to drive adaptation while still targeting the same muscle group.
The end goal here is simple: use your data to intelligently apply progressive overload, break through plateaus, and train more effectively—with zero guesswork.
To this end I also made an HTML page that can organize these in a text searchable way and can be accessed from my phone at the gym. This table of taget sets can be ordered by 1RPM and so easilly parsed.
Advanced Usage
Development
Set up a local environment with uv:
uv venv
uv sync --extra dev
# or, with pip:
pip install ".[dev]"
Run the full test suite (includes benchmarks via pytest-benchmark):
uv run pytest
Before committing, run the pre-commit hooks locally to mirror CI and avoid formatting failures:
uvx pre-commit run --all-files
This installs the pinned ruff-format version and rewrites files if needed, so
rerun the command until it reports no changes.
Import data and run the pareto calculations:
from kaiserlift import (
process_csv_files,
highest_weight_per_rep,
df_next_pareto,
)
csv_files = glob.glob("*.csv")
df = process_csv_files(csv_files)
df_pareto = highest_weight_per_rep(df)
df_targets = df_next_pareto(df_pareto)
Plotting the data:
from kaiserlift import plot_df
# Simple view of all data (only the blue dots)
fig = plot_df(df, Exercise="Dumbbell Curl")
fig.savefig("build/Dumbbell_Curl_Raw.png")
# View with pareto front plotted (the red line)
fig = plot_df(df, df_pareto=df_pareto, Exercise="Dumbbell Curl")
fig.savefig("build/Dumbbell_Curl_Pareto.png")
# View with pareto and targets (the green x's)
fig = plot_df(df, df_pareto=df_pareto, df_targets=df_targets, Exercise="Dumbbell Curl")
fig.savefig("build/Dumbbell_Curl_Pareto_and_Targets.png")
Generate views:
from kaiserlift import (
print_oldest_exercise,
gen_html_viewer,
)
# Console print out with optional args
output_lines = print_oldest_exercise(df, n_cat=2, n_exercises_per_cat=2, n_target_sets_per_exercises=2)
with open("your_workout_summary.txt", "w") as f:
f.writelines(output_lines)
# Print in HTML format for ease of use
full_html = gen_html_viewer(df)
with open("your_interactive_table.html", "w", encoding="utf-8") as f:
f.write(full_html)
# Console print of the HTML
from IPython.display import display, HTML
display(HTML(full_html))
Example HTML generation in CI
An example dataset and helper script live in tests/example_use. You can
generate the interactive HTML table locally with:
python tests/example_use/generate_example_html.py
The "Generate example HTML" job in the CI workflow runs the same script. When run on the main branch, the generated pages are deployed to GitHub Pages:
- Landing page:
index.html - Lifting demo:
lifting/index.html - Running demo:
running/index.html
The HTML is also available as a downloadable artifact for all branches.
1 maintainer
Metadata
1 maintainer
Release files for kaiserlift 0.1.100
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| kaiserlift-0.1.100.tar.gz | 415.5 kB | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| kaiserlift-0.1.100-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 460.2 kB
Release files / kaiserlift-0.1.100.tar.gz
| Download URL | kaiserlift-0.1.100.tar.gz |
|---|---|
| Size | 415.5 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
5475d363ba065a762f400799a9bd2729764e45ef95183c710a3a89ed6d090156
|
|
BLAKE2b-256 checksum How to use checksums |
bb04c537ae358f8c4c8a96846565c6e7501a920a9e8344ece6c93111af0b83a2
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
uv/0.10.4 {"installer":{"name":"uv","version":"0.10.4","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Ubuntu","version":"24.04","id":"noble","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":true}
|
Release files / kaiserlift-0.1.100-py3-none-any.whl
| Download URL | kaiserlift-0.1.100-py3-none-any.whl |
|---|---|
| Size | 44.7 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
830c8a7e2b9a25a825f38afe6b15b86c6172814f4c70858d25715b9f3569f3d0
|
|
BLAKE2b-256 checksum How to use checksums |
21dbf1c21c063b67c02eefa9300d2da511f51e8ded0b6296439baf7f39cbdcdb
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
uv/0.10.4 {"installer":{"name":"uv","version":"0.10.4","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Ubuntu","version":"24.04","id":"noble","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":true}
|
