Skip to main content

Yuclid

Combinatorially explode your experiments

Yuclid is a CLI + Web UI tool for orchestrating and visualizing experiments in N-dimensional irregular spaces of parameters. It collects custom metrics in a single JSONL file for easy post-processing. Yuclid builds the Cartesian product of the dimensions you defined, and runs an experiment per point in that space. Experiments can be monitored and steered in real-time via CLI or Web UI.

Check out the examples.

Installation

Requires python >= 3.10

Development head:

pip install git+https://github.com/fsossai/yuclid.git

Stable release:

pip install yuclid

Overview

The main workflow is to run an experiment, watch or steer it while it is in progress, then inspect and visualize the resulting dataset.

Run and monitor experiments

  • yuclid run builds the configured space, executes every selected point, and writes its dimensions and metrics to JSON Lines or CSV.
  • yuclid serve opens the browser interface for watching, launching, and steering runs recorded in the current directory.
  • yuclid status reports the progress of a live run; --watch keeps the report current until it ends.
  • yuclid runs lists recent recorded runs and their result files.

Steer a live run

  • yuclid pause and yuclid resume temporarily stop and continue scheduling work.
  • yuclid drop removes matching points from the remaining plan; yuclid add restores them or extends the plan with new values.
  • yuclid repeat changes the repetition count for remaining points, and yuclid order changes their traversal order.
  • yuclid kill abandons a point or repetition currently in flight; yuclid stop ends the run and interrupts its active commands.

Revisit recorded runs

  • yuclid finish measures the points a previous run left without results.
  • yuclid replay optimistically runs a previous run again, preserving its intended points and repetitions while retrying failures.

Inspect and visualize results

  • yuclid describe summarizes a result file and reports missing points.
  • yuclid plot explores slices of a dataset in a graphical interface.
  • yuclid tplot provides the interactive plotter in a terminal.
  • yuclid stats plots the distribution of a selected metric.

Agent integration

  • yuclid skills installs or uninstalls Yuclid's configuration and plotting skills for Codex, Claude, or a custom Agent Skills directory.

Run yuclid <command> --help for the complete options of any subcommand.

Configuration

See Configuration.

Skills

Yuclid includes skills for agents that support the Agent Skills format. A few examples of how to install them:

yuclid skills install --agent codex                 # Codex, user-wide
yuclid skills install --agent claude                # Claude, user-wide
yuclid skills install --directory .agents/skills    # custom or project directory
  • yuclid-config writes and fixes a configuration: the space, the trials, and the commands that scrape the numbers out of them.
  • yuclid-plot reads a result file and suggests what is worth looking at, as yuclid plot / tplot / stats commands that you can paste.

Reproducible scripts

yuclid run --compile experiment.sh writes a shell script instead of running anything. Every point of the space is unrolled, so the script contains no loops and no branches — just the commands, in the order yuclid would have run them:

yuclid run -p quick --compile experiment.sh
sh experiment.sh                    # no yuclid, no configuration needed

Metadata

Release files for yuclid 0.4.4

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for yuclid 0.4.4
File Size Uploaded
yuclid-0.4.4.tar.gz 150.8 kB Details

Built distribution (wheel)

Table of built distributions (wheels) for yuclid 0.4.4
File Interpreter ABI Platform
yuclid-0.4.4-py3-none-any.whl Python 3 none any Details

Total release size: 308.0 kB

Release files / yuclid-0.4.4.tar.gz

Download URL yuclid-0.4.4.tar.gz
Size 150.8 kB
Tags Source
SHA-256 checksum
How to use checksums
c193189196ac74fbb6614f1d6f20bcf866dc50ec7a660d830b6dbc221fde7859
BLAKE2b-256 checksum
How to use checksums
33cc7923d792807cc117a192bd5c12ef0a793289e1ffb5853dbc3363d7fd0dbb
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/6.1.0 CPython/3.12.7

Release files / yuclid-0.4.4-py3-none-any.whl

Download URL yuclid-0.4.4-py3-none-any.whl
Size 157.2 kB
Tags Python 3
SHA-256 checksum
How to use checksums
0fe466f243bc28568df653ea5c326da75da7fee489828a2b8bea15392d0ed82c
BLAKE2b-256 checksum
How to use checksums
28ef2ea5c260081845ae5c47e3924ca6cebd204cbd51ccb52cf3726dc24a275e
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/6.1.0 CPython/3.12.7

Release history Release notifications | RSS feed

0.4.7

2 release files

0.4.5

2 release files

This release

0.4.4 This release

2 release files

0.4.3

2 release files

0.4.2

2 release files

0.4.1

2 release files

0.4.0

2 release files

0.3.0

2 release files

0.2.5

2 release files

0.2.4

2 release files

0.2.3

2 release files

0.2.1

2 release files

0.2.0

2 release files

0.1.17

2 release files

0.1.12

2 release files

0.1.9

2 release files

0.1.8

2 release files

0.1.7

2 release files

0.1.5

2 release files

0.1.4

2 release files

0.1.3

2 release files

0.1.2

2 release files

0.1.1

2 release files

0.1.0

2 release files

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page