Library / SDK
writer/writer-framework avatar
writer/writer-framework

Writer Framework: two manifests with two versions, and tag history that stopped in December 2024

No-code in the front, Python in the back. An open-source framework for creating data apps.

1,450 stars102 forksPythonApache-2.0

At a glance

What is it?
A Python framework for building apps with a visual editor in the browser and Python behind it. The repository ships a Node workspace for the editor, Playwright tests across three browsers, and app templates that only exist because the build copies example folders into the Python package.
Who is it for?
Writer Framework fits teams that want to hand a UI to someone who does not write Python and keep the logic in Python, and the two halves of the build make that split concrete rather than metaphorical. Two things decide whether it is a good fit.
Can I use it commercially?
Yes. Apache-2.0 is a permissive licence: you can use, modify and sell software built on it, as long as you keep its copyright and licence notices.
Is it still maintained?
Yes. The repository last received commits 23 days ago.
What is it written in?
Mainly Python, according to GitHub's language statistics.

Answers come from the project's GitHub data, last synced on November 15, 2026, and from our analysis. They are not legal advice.

From daily snapshots

Star history

  • writer/writer-framework1,450 ★ · +3
Star history
Repository2026-09-112026-11-15
writer/writer-framework1.4k1.5k

3 Stars in the last 65 days, measured from the daily GitHub snapshot. Compare with other projects

Editorial analysis

Two manifests, two versions, and a tag list that stopped

The repository root holds a pyproject.toml and a package.json, and they do not agree about the version. The Python manifest declares version 1.27.0 for a package named writer. The Node manifest is also named writer, declares version 1.0.0, and has an empty description field. The tags tell a third story: the newest release is v0.8.2, dated 2024-12-02, after v0.8.1 on 2024-11-29 and v0.7.4 on 2024-09-03. So the published releases stopped roughly two years before the last commit on the dev branch, dated 2026-10-01, and none of them corresponds to the 1.27.0 in the metadata. At the same time the classifiers in the Python manifest include Development Status :: 5 - Production/Stable and Python versions 3.9 through 3.13. A package can be declared stable and still have no tag covering its own version number, which is the practical thing to remember when you decide what to pin.

The file caps Python at 3.12, the constraint allows anything under 4.0

The installation section states that the framework works on Linux, Mac and Windows and supports Python versions 3.9.2 through 3.12. The packaging metadata is looser in two directions at once. The Python constraint reads >=3.9.2, <4.0, so it accepts 3.13 and every later 3.x release, and the classifier list runs to 3.13 even though the prose stops at 3.12. Nothing in the visible metadata marks an upper bound that matches the documented range. If you are on 3.13, the README tells you the combination is unsupported while pip will happily resolve it. The same pattern shows up in the dependency list, where fastapi is capped below 1, pydantic below 3, pandas between 2.2 and 3, pyarrow between 15 and 19, requests between 2.31.0 and 2.33, websockets between 12 and 16, and numpy is pinned conditionally at <=2.0.2 for Python below 3.10 and ^2.0 above it. Those caps are the real compatibility surface, and they are not the ones the prose describes.

App templates are copies made during the build

The starter apps your projects begin from are not tracked files. A Node script copies them into the Python package:

code
"apps:build": "cp -R ./apps/hello ./src/writer/app_templates/ && cp -R ./apps/default ./src/writer/app_templates/"

So apps/hello and apps/default live at the repository root as example folders, and they land in src/writer/app_templates only when npm run apps:build runs, which the main build script does between the editor build and the codegen step. The Python packaging then ships that directory, listing app_templates as included in both the sdist and the wheel, alongside the static and templates directories. The lint script excludes the same path, mypy ./src/writer --exclude app_templates/*, which means the copied templates are checked by neither type checker nor ruff. Three consequences: the directory is generated output rather than source, two copies land in one destination, and anyone reading the published wheel sees templates that were assembled on the machine that built it.

The Python CLI build is a user-interface codegen step

Read the build chain and the two halves of the project interlock in an unexpected order. npm run build is npm run ui:build and then npm run apps:build and then npm run ui:codegen, so the editor is compiled before the app templates are copied in, and codegen runs last. The script named cli:build, which is the build for the Python command line package, is npm run ui:codegen, so the CLI's build output depends on the web editor's code generation rather than on any Python compilation. A second name, codegen, is a plain alias for the same thing. The Python side has its own two commands: cli:test runs pytest tests with log_cli enabled, and cli:lint runs mypy over src/writer with the template directory excluded and then ruff check. Nothing in the visible scripts runs a Python build step, because the published artifact is produced by poetry-core from the build-system table at the top of pyproject.toml.

A sourcemap upload script the tree does not show

One script points at a path that is not in the top-level listing. ld:upload-sourcemaps runs ./scripts/upload-sourcemaps.sh, and build:with-sourcemaps is npm run ui:build followed by that upload. The root entries are .alfred.toml, .devcontainer/, .editorconfig, .github/, .gitignore, .nvmrc, CODE_OF_CONDUCT.md, CONTRIBUTING.md, LICENSE.txt, README.md, alfred/, apps/, favicon.svg, mypy.ini, package-lock.json, package.json, poetry.lock, poetry.toml, pyproject.toml, pytest.ini, src/ and tests/. No scripts directory appears among them, so the upload target of that script is not visible at the top of the repository and the sourcemap build variant cannot be followed from the listing alone. Two other root entries point at tooling outside the framework itself: .alfred.toml and alfred/, an editor launcher configuration, and .devcontainer/, a container definition for the development environment.

Node 22 and npm 10 with x placeholders, three Playwright browsers

The Node side is pinned loosely on purpose. The engines field asks for node 22.x.x and npm 10.x.x, which accepts any 22 and any 10 rather than an exact build, and .nvmrc sits at the root to match. npm workspaces covers two directories, src/ui for the editor and tests/e2e for the browser tests. Playwright is the only entry in devDependencies, at ^1.57.0, and the end-to-end scripts are split by browser, with e2e:firefox, e2e:chromium and e2e:webkit, plus a shared e2e target and an e2e:ci variant. The workspace test and lint commands are npm run --if-present -ws test and the same for lint, so a workspace without a test script is skipped silently instead of failing. The Python configuration lives beside it rather than inside it: mypy.ini, pytest.ini, poetry.lock and poetry.toml all sit at the root, so one repository carries a Python dependency lock and an npm lockfile for the same build.

pip install writer, on a package name that belongs to everyone

The installation section is one command:

sh
pip install writer

That name is as short as it gets, and it is the name the Python manifest uses too. The same repository also publishes a Node package called writer, and the keywords listed in the Python metadata are data apps, gui, ui, framework and writer framework, with the description framing the project as an open-source Python framework for building apps fully integrated with the Writer platform. The README closes with a paragraph about the company rather than the code, describing a full-stack generative AI platform for enterprises with LLM, graph-based RAG and guardrail tooling, and the outbound link to writer.com carries three campaign parameters, utm_source, utm_medium and utm_campaign, in the URL itself. None of that is hidden, but a reader arriving from a package index sees a commercial pitch in the last paragraph of an Apache 2.0 file.

Editorial conclusion

Writer Framework fits teams that want to hand a UI to someone who does not write Python and keep the logic in Python, and the two halves of the build make that split concrete rather than metaphorical. Two things decide whether it is a good fit. Version numbers here are not a reliable guide: one manifest says 1.27.0, the other says 1.0.0 with an empty description, and the newest tag is v0.8.2 from 2024-12-02 while the dev branch was still moving on 2026-10-01. And the app templates your projects start from are generated during the build, so what you get from the editor depends on copies made by a shell command rather than on tracked files. Before you pin anything, check which Python versions the package metadata accepts, since the README stops at 3.12 while the constraint allows anything below 4.0. The license is Apache 2.0.

Frequently asked questions

How do I install Writer Framework?

One pip command in a proper Python environment: pip install writer. The file says the framework works on Linux, Mac and Windows and supports Python 3.9.2 through 3.12, while the packaging metadata allows >=3.9.2 and below 4.0 and lists classifiers through 3.13, so the documented upper bound is tighter than the constraint.

What does the writer edit command do?

It opens the visual editor in your browser, where the interface is built by dragging and dropping components. The quick start pairs it with writer create my_app to make the app, writer hello to generate a demo one, and writer run my_app to serve it.

What version of Writer Framework is current?

The Python manifest declares version 1.27.0 while the Node manifest declares 1.0.0 with an empty description. The newest release tag is v0.8.2 from 2024-12-02, after v0.8.1 and v0.7.4 in 2024, so no tag matches the declared 1.27.0. The default branch is dev and its last commit on record is dated 2026-10-01.

Where do Writer Framework's starter app templates come from?

From a copy step in the Node build: apps:build runs cp -R on the apps/hello and apps/default folders into src/writer/app_templates. The Python packaging then includes that generated directory in both the sdist and the wheel, and the lint script excludes the same path from mypy.

Is Writer Framework the same thing as Writer the AI platform?

It is related but separate. The README describes Writer as a full-stack generative AI platform for enterprises with developer tools integrated with its LLM, graph-based RAG and guardrail tooling, while the framework is the open-source Python layer that builds apps on top of it, with the writer-sdk package as a dependency and Apache 2.0 licensing for the framework itself.

Official sources

  1. License: Apache-2.0
  2. Project website
  3. README
  4. Releases
  5. writer/writer-framework on GitHub
Add this badge to your README

If you maintain this project, the badge below links readers to this analysis and shows its maintenance status from the daily GitHub snapshot. Paste the markdown into your README; add ?metric=license or ?metric=stars to the image URL for a different field.

Add this badge to your README

markdown
[![Hysen Labs](https://hysenlabs.com/badge/writer-writer-framework.svg)](https://hysenlabs.com/projects/writer-writer-framework)

Star history chart for your README

markdown
[![Star history](https://hysenlabs.com/chart/writer-writer-framework.svg)](https://hysenlabs.com/projects/writer-writer-framework)

Was this useful?