Changelog

LEX AI 1.2.1 and 1.2.2

This is the first LEX AI entry in this changelog. It covers two releases, 1.2.1
and 1.2.2, because both are available now. Upgrading is one command and it reaches
both. Contact us if you would like help with the update.

Two things define these releases: more than one LEX AI run can work on a project at the
same time, and LEX AI can now build dashboards with Reflex as well as with Streamlit.

If you work in a single chat and do not use Reflex, nothing you already do changes. Run
the update and carry on.


1.2.2 — Dashboards with Reflex

Reflex and Streamlit are two different tools for building a web page from Python code.
Lex App has supported both for a while. LEX AI only knew Streamlit, so an agent asked for
a Reflex dashboard wrote a Streamlit one instead.

Every mode that touches a dashboard now handles both

  • A new project chooses the framework before it writes anything. The choice is made
    once, at the start, and the rest of the run follows it.
  • A targeted change offers a Reflex task where the project already uses Reflex.
  • A documentation run detects a Reflex dashboard and writes it into the wiki.
  • A deployment run treats the Reflex process as part of what has to be started, beside
    the application itself.

Streamlit remains the default

This is not a preference. There are three reasons:

  • Only a Streamlit dashboard opens from the record page's Analytics tab and from the
    grid's dashboard toggle. A Reflex dashboard is reached by its own web address.
  • lex_view and lex_widgets, which are Lex App's own pages and controls inside a
    dashboard, are Streamlit components. A Reflex dashboard links to Lex App's pages, or
    calls the API, instead.
  • Reflex sign-in runs through Reflex Enterprise. That is free while you are developing,
    but it needs a paid Reflex plan to run in production.

Your existing Streamlit dashboards are not affected. LEX AI never moves a dashboard
from one tool to the other unless you ask it to.

A note on running one. The command to run a Reflex dashboard is lex reflex, which
belongs to the LEX App package and not to LEX AI. If that command is not available on your
machine, update LEX App as shown below.


1.2.1 — More than one run at a time

LEX AI used to assume one chat per project. A second chat — a fix during a test run, or a
review beside an edit — meant both were writing the same files, and neither could tell
whose work was whose. These releases make a second run supported instead of dangerous.

Added

  • Parallel runs on one project. lex ai-worktree prepares a separate copy of the
    project on its own branch and prints the folder to open the new chat in. That chat
    starts by asking what the run is for, then hands over to whichever mode the answer
    needs.
  • Automatic handover. Start a second chat without knowing that command exists, and
    LEX AI notices the project is already busy, prepares the same folder itself, and gives
    you the path.
  • A check before every run, in all ten modes. Several edits, or a test run beside an
    edit, each get their own copy. Building, documenting, input changes and deployment stay
    one at a time, and tell you which run to wait for instead of colliding with it. A run
    whose chat you have closed stops blocking after a while.
  • Edit mode now version-controls its own work. One branch per run, one commit per
    task, then a pull request. It does not merge for you, because with several runs open at
    once that would put one run's work underneath another that is still building on it.
  • A record for each run at .lex/briefs/<run-id>.md, so a pull request carries a file
    saying what the run was for.
  • Clearer failure messages — an inspecting proxy, an expired token, no disk space, not
    a git repository — and the build version on every event. No code, business data, prompts
    or passwords are collected.

Changed

  • The project contract now accumulates instead of being replaced. A short interview no
    longer deletes the domain rules and data agreements an earlier build established.
    Anything carried forward from an earlier run is marked as carried and counted, so an old
    decision is not mistaken for a new one. A run's own goal and acceptance criteria are
    never inherited.

Fixed

  • Windows: LEX AI now reads your repository the way your own git does. A clean copy of
    a project was being read as hundreds of changed files.
  • The check that protects your LEX AI files was silently skipping on many
    installations. It now runs properly.
  • Run identifiers could clash when two chats were opened a few seconds apart.
  • Host and user information is now recorded for all ten modes rather than only some.

How to upgrade

Open a terminal in your project folder and run:

python -m lex ai-update

That is the whole upgrade. It updates the LEX AI package, refreshes the Lex handbook in
your docs/ folder, rewrites the generated agent files, and re-registers LEX AI in your
IDE's configuration. Your own files are not changed.

Restart your IDE afterwards. Your IDE keeps its own copy of the LEX AI command list
and only reads the new one when it starts.

If you want to use Reflex dashboards, you also need a recent version of LEX App, which is
a different package:

py -3.12 -m pip install lex-app -U      # Windows
python3.12 -m pip install lex-app -U    # macOS and Linux

Everything else here works without updating LEX App.