Skip to content

Add Claude Code plugin for AI-assisted Hamilton development #1460

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Open
skrawcz wants to merge 2 commits into
base: main
Choose a base branch
from
Open

Add Claude Code plugin for AI-assisted Hamilton development #1460

skrawcz wants to merge 2 commits into from
+3,606 −0

Conversation

@skrawcz
Copy link
Contributor

@skrawcz skrawcz commented Jan 31, 2026

This commit adds a comprehensive Claude Code plugin/skill that provides AI-powered assistance for Hamilton DAG development.

Features:

  • Create Hamilton modules with proper patterns and decorators
  • Understand and explain existing DAGs
  • Apply function modifiers (@parameterize, @config.when, @check_output, etc.)
  • Convert Python scripts to Hamilton modules
  • Debug issues (circular dependencies, missing nodes, etc.)
  • Optimize pipelines with caching and parallelization
  • Generate tests and documentation
  • LLM/RAG workflow patterns
  • Integration patterns (Airflow, FastAPI, Streamlit, etc.)

Structure:

  • .claude/skills/hamilton/ - Auto-available for Hamilton contributors
  • .claude/plugins/hamilton/ - Installable plugin for external users
  • docs/ecosystem/claude-code-plugin.md - User documentation

Installation for users:
/plugin marketplace add apache/hamilton /plugin install hamilton --scope user

All files include Apache 2.0 license headers.

Contributing: Users can file issues or submit PRs to improve the skill.

Changes

  • adds claude skill
  • adds documentation

How I tested this

  • with claude -- likely can be improved still (need to audit the skill a bit more)

Notes

Checklist

  • PR has an informative and human-readable title (this will be pulled into the release notes)
  • Changes are limited to a single goal (no scope creep)
  • Code passed the pre-commit check & code is left cleaner/nicer than when first encountered.
  • Any change in functionality is tested
  • New functions are documented (with a description, list of inputs, and expected output)
  • Placeholder code is flagged / future TODOs are captured in comments
  • Project documentation has been updated if adding/changing functionality.

skrawcz
skrawcz commented Jan 31, 2026
View reviewed changes
.claude/plugins/hamilton/.claude-plugin/marketplace.json Outdated
"description": "Official Claude Code plugin for the Hamilton framework",
"version": "1.0.0",
"owner": {
"name": "DAGWorks Inc.",
Copy link
Contributor Author

@skrawcz skrawcz Jan 31, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

ugh fixing

@skrawcz skrawcz requested a review from zilto January 31, 2026 19:35
@skrawcz
Copy link
Contributor Author

skrawcz commented Jan 31, 2026

@zilto have you already created something here? I'm still proof reading the AI guide here...

@zilto
Copy link
Contributor

zilto commented Feb 2, 2026 •
edited
Loading

The general workflow that works for me:

  1. problem in natural language -> DOT language graph
  • well understood by LLMs, simple language, extremely token efficient
  1. DOT language -> function signatures + docstring
  • this is simple and can be done programmatically; the challenge is defining type annotations ahead of time; can run ty type checker or force Any when uncertain
  • validate the DAG with the hamilton CLI (I think the command is build ?)
  1. function signatures -> implementations
  • can do test driven development where nodes are implemented with tests 1-by-1. This allows Claude for example to build much more complex code than spaghetti scripts; LLM has easier time refactoring

Misc

  • you can set Any everywhere or use the "don't validate types" adapter on the Driver and then run MonkeyType to annotate functions based on values passed. This is not perfect but very efficient for a first draft

zilto
zilto requested changes Feb 3, 2026
View reviewed changes
Copy link
Contributor

@zilto zilto left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I reviewed the files, but I couldn't figure out local installation after cloning the repo. I believe it will be much simpler once it's merged to main and widely available.

Question: why not make the plugin directly available under .claude-plugin/?

Installation would be simpler if .claude-plugin/marketplace.json was at the top-level instead of .claude/plugins/hamilton/.claude-plugins.

I think .claude/ is reserved for your claude configuration

@skrawcz
Split Hamilton Claude plugin into 5 focused skills
00ea548 
Replace single monolithic skill with 5 specialized skills for better organization:

- **core**: Basic Hamilton patterns (DAG creation, decorators, testing, debugging)
- **scale**: Performance & parallelization (async, Spark, Ray, Dask, caching, multithreading)
- **llm**: AI/ML workflows (RAG pipelines, embeddings, vector DBs, prompt engineering)
- **observability**: Hamilton UI & SDK (tracking, monitoring, lineage, debugging)
- **integrations**: Orchestrators & frameworks (Airflow, FastAPI, Streamlit, Jupyter)

Each skill is ~200-400 lines, focused, and independently invocable via:
- /hamilton-core
- /hamilton-scale
- /hamilton-llm
- /hamilton-observability
- /hamilton-integrations

Benefits:
- Faster skill loading (only load what's needed)
- Better context relevance (focused expertise)
- Easier maintenance (separate concerns)
- Scales with Hamilton's growth (can add more skills)

Updated plugin.json to v2.0.0 with array of skill paths. (+6 squashed commits)
Squashed commits:
[137a106] Remove duplicate skill directory and ignore local settings

Remove `.claude/skills/hamilton/` directory to eliminate duplication.
Keep only `.claude/plugins/hamilton/skills/hamilton/` as the single
source of truth. Update all references in documentation to point to
the plugin location. Also remove `.claude/settings.local.json` from
tracking and add it to `.gitignore` as it contains personal settings.
[f5ca918] Enhance Claude skill with async, Spark, caching, and parallelism documentation

Expand the Hamilton Claude Code skill to comprehensively cover I/O-bound
parallelization patterns and distributed processing. Adds detailed examples
and best practices for AsyncDriver (RAG pipelines, dependent API chains),
Apache Spark integration (@with_columns decorator, pandas UDFs), caching
strategies (LLM call optimization, automatic invalidation), and
MultiThreadingExecutor for synchronous I/O concurrency. These additions
help users leverage Hamilton's full parallelization capabilities for
modern data and AI workflows.
[e09e968] Fix Sphinx toctree warning for Claude Code plugin documentation
[263e092] Update Claude Code plugin to use Apache Hamilton branding

Update all references in the Claude Code plugin and skills to properly
use "Apache Hamilton" as the project name throughout documentation,
descriptions, and user-facing text. This ensures consistent branding
and recognition of the project as part of the Apache Software Foundation.

Changes include:
- Plugin description and metadata
- Skill descriptions and documentation
- README files for both plugin and skills
- Code examples and patterns documentation
- Installation and contribution guidelines
[b68f77b] Update Claude Code plugin references from dagworks-inc to apache

Change all plugin and skill references to use Apache organization:
- GitHub URLs: dagworks-inc/hamilton → apache/hamilton
- Documentation URLs: hamilton.dagworks.io → hamilton.apache.org
- Contact email: support@dagworks.io → dev@hamilton.apache.org
- Marketplace owner: DAGWorks Inc. → Apache Software Foundation
[f290bc7] Add Claude Code plugin for AI-assisted Hamilton development

This commit adds a comprehensive Claude Code plugin/skill that provides
AI-powered assistance for Hamilton DAG development.

Features:
- Create Hamilton modules with proper patterns and decorators
- Understand and explain existing DAGs
- Apply function modifiers (@parameterize, @config.when, @check_output, etc.)
- Convert Python scripts to Hamilton modules
- Debug issues (circular dependencies, missing nodes, etc.)
- Optimize pipelines with caching and parallelization
- Generate tests and documentation
- LLM/RAG workflow patterns
- Integration patterns (Airflow, FastAPI, Streamlit, etc.)

Structure:
- .claude/skills/hamilton/ - Auto-available for Hamilton contributors
- .claude/plugins/hamilton/ - Installable plugin for external users
- docs/ecosystem/claude-code-plugin.md - User documentation

Installation for users:
  /plugin marketplace add dagworks-inc/hamilton
  /plugin install hamilton --scope user

All files include Apache 2.0 license headers.

Contributing: Users can file issues or submit PRs to improve the skill.
@skrawcz
Adds workflow pattern and moves claude plugin directory
2b571aa
zilto
zilto approved these changes Feb 5, 2026
View reviewed changes
Copy link
Contributor

@zilto zilto left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM

@zilto zilto marked this pull request as ready for review February 6, 2026 12:19
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@zilto zilto zilto approved these changes

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

2 participants

@skrawcz @zilto