OpenAgents

Code

AGENTS.md

58e6347eeb72 · 28 KB

This is a web application written using the Phoenix web framework.

Vocabulary

Read docs/taxonomy.md before writing docs, commit messages, or product copy. It defines what each word means here — forge versus GitHub, push versus deploy, computer versus machine, which receipt, which module — and the naming rules that keep them straight.

Communication style

All text in this repo — docs, README, AGENTS.md, commit messages, and agent responses — follows the Google Developer Documentation Style Guide. When writing or reviewing text, read the google-developer-style skill at .agents/skills/google-developer-style/SKILL.md.

  • Write in active voice and address the reader as you.
  • Use sentence case for all headings and titles.
  • Use code font for code, filenames, class names, HTTP status codes, and placeholders.
  • Use bold for UI elements.
  • Use numbered lists for procedures and bulleted lists for unrelated items.
  • Avoid jargon, buzzwords, metaphors, exclamation marks, and phrases like simply or just.
  • Avoid please in instructions.

Git remotes

Push to the forge, never to GitHub:

git push openagents HEAD:main

The openagents remote is the forge at openagents.com, which records every push in the durable WAL and serves it. The origin remote is the GitHub mirror; pushing to it directly leaves the forge behind a mirror it does not know about, and nothing reports the divergence until a clone disagrees with the site. Production mirrors to GitHub with a force push of every ref, so a direct GitHub push is not merged later — it is overwritten. GitHub holds what the forge last mirrored there. ops/ci/push-remote-check.sh refuses a non-forge push, and mix precommit installs it into the clone you are working in, so running precommit before you push is enough. To install it by hand:

sh ops/dev/install-push-guard.sh

One install covers every worktree of that clone. Where core.hooksPath points at .githooks, its pre-push runs the same guard ahead of the release gate. See INVARIANTS.md, REPOSITORY-002.

Project guidelines

  • Use mix precommit alias when you are done with all changes and fix any pending issues
  • Use the already included and available :req (Req) library for HTTP requests, avoid :httpoison, :tesla, and :httpc. Req is included by default and is the preferred HTTP client for Phoenix apps

Phoenix v1.8 guidelines

  • Always begin your LiveView templates with <Layouts.app flash={@flash} ...> which wraps all inner content
  • The MyAppWeb.Layouts module is aliased in the my_app_web.ex file, so you can use it without needing to alias it again
  • Anytime you run into errors with no current_scope assign:
    • You failed to follow the Authenticated Routes guidelines, or you failed to pass current_scope to <Layouts.app>
    • Always fix the current_scope error by moving your routes to the proper live_session and ensure you pass current_scope as needed
  • Phoenix v1.8 moved the <.flash_group> component to the Layouts module. You are forbidden from calling <.flash_group> outside of the layouts.ex module
  • Render icons only through OpenAgentsWeb.UI.icon/1. Prefer the vendored Apps SDK set. Use a hero-* fallback only when docs/ICONS.md records why the preferred set has no suitable glyph
  • Always use the imported OpenAgentsWeb.UI.input/1 component for form inputs. It accepts Phoenix.HTML.FormField values and unwrapped raw controls
  • If you override the default input classes (<.input class="myclass px-2 py-1 rounded-lg">)) class with your own values, no default classes are inherited, so your custom classes must fully style the input

JS and CSS guidelines

  • Use Tailwind CSS classes and custom CSS rules to create polished, responsive, and visually stunning interfaces.

  • Tailwindcss v4 no longer needs a tailwind.config.js and uses a new import syntax in app.css:

    @import "tailwindcss" source(none);
@source "../css";
@source "../js";
@source "../../lib/my_app_web";

  • Always use and maintain this import syntax in the app.css file for projects generated with phx.new

  • Never use @apply when writing raw css

  • There is exactly one component system: vendored Basecoat plus OpenAgents style pack. Basecoat lives in assets/vendor/basecoat/components/ and carries structure (display, padding, min-height); assets/css/openagents.css carries OpenAgents' identity (motion tokens, radius scale, color, the notched variant, corner frames) and must stay the last import so its declarations win.

    • Reach for OpenAgentsWeb.UI first. It wraps that CSS in 22 ready primitives — button/1, card/1, badge/1, alert/1, input/1, textarea/1, label/1, field/1, header/1, table/1, list/1, avatar/1, menu/1, empty/1, kbd/1, and the rest. Fall back to hand-written classes only when no primitive covers the shape.
    • Variants are data attributes, not classes. A control is class="btn" plus data-variant="primary" / data-size="sm" / data-tone="danger". Never invent btn-primary-style variant classes; they defeat the whole point of the split.
    • Never add a second component library — DaisyUI above all. DaisyUI was removed deliberately. It emitted flat .btn rules (setting background, color, border) from a cascade layer that outranked every .btn[data-variant=…] in openagents.css, so all eight UI button variants rendered identically on staging. Any library with the same shape will do the same thing again.
    • Import Basecoat components individually, one @import "../vendor/basecoat/components/<name>.css" per component a surface actually uses. Component CSS lands in @layer components and ships whether or not the class appears in markup, so the list is a budget. Never import basecoat.css, basecoat-base.css, or basecoat-components.css — each pulls in all 39 components at once.
    • The palette is the two owned token ladders in assets/css/openagents.css. The interface supports only light and dark palettes; the system preference selects between them and is not a third palette. Use the Basecoat utility names built on these tokens — bg-background, bg-card, text-foreground, text-muted-foreground, and border-border. The retired compatibility aliases no longer exist

  • Add custom Tailwind only when the design is unique to OpenAgents and no primitive fits.

  • Out of the box only the app.js and app.css bundles are supported

    • You cannot reference an external vendor'd script src or link href in the layouts
    • You must import the vendor deps into app.js and app.css to use them
    • Never write inline <script> tags within templates. The only exception is the synchronous, content-free theme bootstrap in root.html.heex, which must run before the first stylesheet paint and carry the response-scoped CSP nonce. Put every other client behavior in app.js or a colocated LiveView hook

UI and UX design guidelines

  • Produce world-class UI designs with a focus on usability, aesthetics, and modern design principles
  • Implement subtle micro-interactions (e.g., button hover effects, and smooth transitions)
  • Ensure clean typography, spacing, and layout balance for a refined, premium look
  • Focus on delightful details like hover effects, loading states, and smooth page transitions

Elixir guidelines

  • Elixir lists do not support index based access via the access syntax

    Never do this (invalid):

    i = 0
mylist = ["blue", "green"]
mylist[i]

    Instead, always use Enum.at, pattern matching, or List for index based list access, ie:

    i = 0
mylist = ["blue", "green"]
Enum.at(mylist, i)

  • Elixir variables are immutable, but can be rebound, so for block expressions like if, case, cond, etc you must bind the result of the expression to a variable if you want to use it and you CANNOT rebind the result inside the expression, ie:

    # INVALID: we are rebinding inside the `if` and the result never gets assigned
if connected?(socket) do
  socket = assign(socket, :val, val)
end

# VALID: we rebind the result of the `if` to a new variable
socket =
  if connected?(socket) do
    assign(socket, :val, val)
  end

  • Never nest multiple modules in the same file as it can cause cyclic dependencies and compilation errors

  • Never use map access syntax (changeset[:field]) on structs as they do not implement the Access behaviour by default. For regular structs, you must access the fields directly, such as my_struct.field or use higher level APIs that are available on the struct if they exist, Ecto.Changeset.get_field/2 for changesets

  • Elixir's standard library has everything necessary for date and time manipulation. Familiarize yourself with the common Time, Date, DateTime, and Calendar interfaces by accessing their documentation as necessary. Never install additional dependencies unless asked or for date/time parsing (which you can use the date_time_parser package)

  • Don't use String.to_atom/1 on user input (memory leak risk)

  • Predicate function names should not start with is_ and should end in a question mark. Names like is_thing should be reserved for guards

  • Elixir's builtin OTP primitives like DynamicSupervisor and Registry, require names in the child spec, such as {DynamicSupervisor, name: MyApp.MyDynamicSup}, then you can use DynamicSupervisor.start_child(MyApp.MyDynamicSup, child_spec)

  • Use Task.async_stream(collection, callback, options) for concurrent enumeration with back-pressure. The majority of times you will want to pass timeout: :infinity as option

Mix guidelines

  • Read the docs and options before using tasks (by using mix help task_name)
  • To debug test failures, run tests in a specific file with mix test test/my_test.exs or run all previously failed tests with mix test --failed
  • mix deps.clean --all is almost never needed. Avoid using it unless you have good reason

Test guidelines

  • Always use start_supervised!/1 to start processes in tests as it guarantees cleanup between tests
  • Avoid Process.sleep/1 and Process.alive?/1 in tests

    • Instead of sleeping to wait for a process to finish, always use Process.monitor/1 and assert on the DOWN message:

      ref = Process.monitor(pid) assert_receive {:DOWN, ^ref, :process, ^pid, :normal}

    • Instead of sleeping to synchronize before the next call, always use _ = :sys.get_state/1 to ensure the process has handled prior messages

Phoenix guidelines

  • Remember Phoenix router scope blocks include an optional alias which is prefixed for all routes within the scope. Always be mindful of this when creating routes within a scope to avoid duplicate module prefixes.

  • You never need to create your own alias for route definitions! The scope provides the alias, ie:

    scope "/admin", AppWeb.Admin do
  pipe_through :browser

  live "/users", UserLive, :index
end

    the UserLive route would point to the AppWeb.Admin.UserLive module

  • Phoenix.View no longer is needed or included with Phoenix, don't use it

Ecto Guidelines

  • Always preload Ecto associations in queries when they'll be accessed in templates, ie a message that needs to reference the message.user.email
  • Remember import Ecto.Query and other supporting modules when you write seeds.exs
  • Ecto.Schema fields always use the :string type, even for :text, columns, ie: field :name, :string
  • Ecto.Changeset.validate_number/2 DOES NOT SUPPORT the :allow_nil option. By default, Ecto validations only run if a change for the given field exists and the change value is not nil, so such as option is never needed
  • You must use Ecto.Changeset.get_field(changeset, :field) to access changeset fields
  • Fields which are set programmatically, such as user_id, must not be listed in cast calls or similar for security purposes. Instead they must be explicitly set when creating the struct
  • Always invoke mix ecto.gen.migration migration_name_using_underscores when generating migration files, so the correct timestamp and conventions are applied

Phoenix HTML guidelines

  • Phoenix templates always use ~H or .html.heex files (known as HEEx), never use ~E

  • Always use the imported Phoenix.Component.form/1 and Phoenix.Component.inputs_for/1 function to build forms. Never use Phoenix.HTML.form_for or Phoenix.HTML.inputs_for as they are outdated

  • When building forms always use the already imported Phoenix.Component.to_form/2 (assign(socket, form: to_form(...)) and <.form for={@form} id="msg-form">), then access those forms in the template via @form[:field]

  • Always add unique DOM IDs to key elements (like forms, buttons, etc) when writing templates, these IDs can later be used in tests (<.form for={@form} id="product-form">)

  • For "app wide" template imports, you can import/alias into the my_app_web.ex's html_helpers block, so they will be available to all LiveViews, LiveComponent's, and all modules that do use MyAppWeb, :html (replace "my_app" by the actual app name)

  • Elixir supports if/else but does NOT support if/else if or if/elsif. Never use else if or elseif in Elixir, always use cond or case for multiple conditionals.

    Never do this (invalid):

    <%= if condition do %>
  ...
<% else if other_condition %>
  ...
<% end %>

    Instead always do this:

    <%= cond do %>
  <% condition -> %>
    ...
  <% condition2 -> %>
    ...
  <% true -> %>
    ...
<% end %>

  • HEEx require special tag annotation if you want to insert literal curly's like { or }. If you want to show a textual code snippet on the page in a <pre> or <code> block you must annotate the parent tag with phx-no-curly-interpolation:

    <code phx-no-curly-interpolation>
  let obj = {key: "val"}
</code>

    Within phx-no-curly-interpolation annotated tags, you can use { and } without escaping them, and dynamic Elixir expressions can still be used with <%= ... %> syntax

  • HEEx class attrs support lists, but you must always use list [...] syntax. You can use the class list syntax to conditionally add classes, always do this for multiple class values:

    <a class={[
  "px-2 text-white",
  @some_flag && "py-5",
  if(@other_condition, do: "border-red-500", else: "border-blue-100"),
  ...
]}>Text</a>

    and always wrap if's inside {...} expressions with parens, like done above (if(@other_condition, do: "...", else: "..."))

    and never do this, since it's invalid (note the missing [ and ]):

    <a class={
  "px-2 text-white",
  @some_flag && "py-5"
}> ...
=> Raises compile syntax error on invalid HEEx attr syntax

  • Never use <% Enum.each %> or non-for comprehensions for generating template content, instead always use <%= for item <- @collection do %>

  • HEEx HTML comments use <%!-- comment --%>. Always use the HEEx HTML comment syntax for template comments (<%!-- comment --%>)

  • HEEx allows interpolation via {...} and <%= ... %>, but the <%= %> only works within tag bodies. Always use the {...} syntax for interpolation within tag attributes, and for interpolation of values within tag bodies. Always interpolate block constructs (if, cond, case, for) within tag bodies using <%= ... %>.

    Always do this:

    <div id={@id}>
  {@my_assign}
  <%= if @some_block_condition do %>
    {@another_assign}
  <% end %>
</div>

    and Never do this – the program will terminate with a syntax error:

    <%!-- THIS IS INVALID NEVER EVER DO THIS --%>
<div id="<%= @invalid_interpolation %>">
  {if @invalid_block_construct do}
  {end}
</div>

Phoenix LiveView guidelines

  • Never use the deprecated live_redirect and live_patch functions, instead always use the <.link navigate={href}> and <.link patch={href}> in templates, and push_navigate and push_patch functions LiveViews
  • Avoid LiveComponent's unless you have a strong, specific need for them
  • LiveViews should be named like AppWeb.WeatherLive, with a Live suffix. When you go to add LiveView routes to the router, the default :browser scope is already aliased with the AppWeb module, so you can just do live "/weather", WeatherLive

LiveView streams

  • Always use LiveView streams for collections for assigning regular lists to avoid memory ballooning and runtime termination with the following operations:

    • basic append of N items - stream(socket, :messages, [new_msg])
    • resetting stream with new items - stream(socket, :messages, [new_msg], reset: true) (e.g. for filtering items)
    • prepend to stream - stream(socket, :messages, [new_msg], at: -1)
    • deleting items - stream_delete(socket, :messages, msg)

  • When using the stream/3 interfaces in the LiveView, the LiveView template must 1) always set phx-update="stream" on the parent element, with a DOM id on the parent element like id="messages" and 2) consume the @streams.stream_name collection and use the id as the DOM id for each child. For a call like stream(socket, :messages, [new_msg]) in the LiveView, the template would be:

    <div id="messages" phx-update="stream">
  <div :for={{id, msg} <- @streams.messages} id={id}>
    {msg.text}
  </div>
</div>

  • LiveView streams are not enumerable, so you cannot use Enum.filter/2 or Enum.reject/2 on them. Instead, if you want to filter, prune, or refresh a list of items on the UI, you must refetch the data and re-stream the entire stream collection, passing reset: true:

    def handle_event("filter", %{"filter" => filter}, socket) do
  # re-fetch the messages based on the filter
  messages = list_messages(filter)

  {:noreply,
   socket
   |> assign(:messages_empty?, messages == [])
   # reset the stream with the new messages
   |> stream(:messages, messages, reset: true)}
end

  • LiveView streams do not support counting or empty states. If you need to display a count, you must track it using a separate assign. For empty states, you can use Tailwind classes:

    <div id="tasks" phx-update="stream">
  <div class="hidden only:block">No tasks yet</div>
  <div :for={{id, task} <- @streams.tasks} id={id}>
    {task.name}
  </div>
</div>

    The above only works if the empty state is the only HTML block alongside the stream for-comprehension.

  • When updating an assign that should change content inside any streamed item(s), you MUST re-stream the items along with the updated assign:

    def handle_event("edit_message", %{"message_id" => message_id}, socket) do
  message = Chat.get_message!(message_id)
  edit_form = to_form(Chat.change_message(message, %{content: message.content}))

  # re-insert message so @editing_message_id toggle logic takes effect for that stream item
  {:noreply,
   socket
   |> stream_insert(:messages, message)
   |> assign(:editing_message_id, String.to_integer(message_id))
   |> assign(:edit_form, edit_form)}
end

    And in the template:

    <div id="messages" phx-update="stream">
  <div :for={{id, message} <- @streams.messages} id={id} class="flex group">
    {message.username}
    <%= if @editing_message_id == message.id do %>
      <%!-- Edit mode --%>
      <.form for={@edit_form} id="edit-form-#{message.id}" phx-submit="save_edit">
        ...
      </.form>
    <% end %>
  </div>
</div>

  • Never use the deprecated phx-update="append" or phx-update="prepend" for collections

LiveView JavaScript interop

  • Remember anytime you use phx-hook="MyHook" and that JS hook manages its own DOM, you must also set the phx-update="ignore" attribute
  • Always provide an unique DOM id alongside phx-hook otherwise a compiler error will be raised

LiveView hooks come in two flavors, 1) colocated js hooks for "inline" scripts defined inside HEEx, and 2) external phx-hook annotations where JavaScript object literals are defined and passed to the LiveSocket constructor.

Inline colocated js hooks

Never write raw embedded <script> tags in heex as they are incompatible with LiveView. Instead, always use a colocated js hook script tag (:type={Phoenix.LiveView.ColocatedHook}) when writing scripts inside the template:

<input type="text" name="user[phone_number]" id="user-phone-number" phx-hook=".PhoneNumber" />
<script :type={Phoenix.LiveView.ColocatedHook} name=".PhoneNumber">
  export default {
    mounted() {
      this.el.addEventListener("input", e => {
        let match = this.el.value.replace(/\D/g, "").match(/^(\d{3})(\d{3})(\d{4})$/)
        if(match) {
          this.el.value = `${match[1]}-${match[2]}-${match[3]}`
        }
      })
    }
  }
</script>
  • colocated hooks are automatically integrated into the app.js bundle
  • colocated hooks names MUST ALWAYS start with a . prefix, i.e. .PhoneNumber

External phx-hook

External JS hooks (<div id="myhook" phx-hook="MyHook">) must be placed in assets/js/ and passed to the LiveSocket constructor:

const MyHook = {
  mounted() { ... }
}
let liveSocket = new LiveSocket("/live", Socket, {
  hooks: { MyHook }
});

Pushing events between client and server

Use LiveView's push_event/3 when you need to push events/data to the client for a phx-hook to handle. Always return or rebind the socket on push_event/3 when pushing events:

# re-bind socket so we maintain event state to be pushed
socket = push_event(socket, "my_event", %{...})

# or return the modified socket directly:
def handle_event("some_event", _, socket) do
  {:noreply, push_event(socket, "my_event", %{...})}
end

Pushed events can then be picked up in a JS hook with this.handleEvent:

mounted() {
  this.handleEvent("my_event", data => console.log("from server:", data));
}

Clients can also push an event to the server and receive a reply with this.pushEvent:

mounted() {
  this.el.addEventListener("click", e => {
    this.pushEvent("my_event", { one: 1 }, reply => console.log("got reply from server:", reply));
  })
}

Where the server handled it via:

def handle_event("my_event", %{"one" => 1}, socket) do
  {:reply, %{two: 2}, socket}
end

LiveView tests

  • Phoenix.LiveViewTest module and LazyHTML (included) for making your assertions

  • Form tests are driven by Phoenix.LiveViewTest's render_submit/2 and render_change/2 functions

  • Come up with a step-by-step test plan that splits major test cases into small, isolated files. You may start with simpler tests that verify content exists, gradually add interaction tests

  • Always reference the key element IDs you added in the LiveView templates in your tests for Phoenix.LiveViewTest functions like element/2, has_element/2, selectors, etc

  • Never tests again raw HTML, always use element/2, has_element/2, and similar: assert has_element?(view, "#my-form")

  • Instead of relying on testing text content, which can change, favor testing for the presence of key elements

  • Focus on testing outcomes rather than implementation details

  • Be aware that Phoenix.Component functions like <.form> might produce different HTML than expected. Test against the output HTML structure, not your mental model of what you expect it to be

  • When facing test failures with element selectors, add debug statements to print the actual HTML, but use LazyHTML selectors to limit the output, ie:

    html = render(view)
document = LazyHTML.from_fragment(html)
matches = LazyHTML.filter(document, "your-complex-selector")
IO.inspect(matches, label: "Matches")

Form handling

Creating a form from params

If you want to create a form based on handle_event params:

def handle_event("submitted", params, socket) do
  {:noreply, assign(socket, form: to_form(params))}
end

When you pass a map to to_form/1, it assumes said map contains the form params, which are expected to have string keys.

You can also specify a name to nest the params:

def handle_event("submitted", %{"user" => user_params}, socket) do
  {:noreply, assign(socket, form: to_form(user_params, as: :user))}
end

Creating a form from changesets

When using changesets, the underlying data, form params, and errors are retrieved from it. The :as option is automatically computed too. E.g. if you have a user schema:

defmodule MyApp.Users.User do
  use Ecto.Schema
  ...
end

And then you create a changeset that you pass to to_form:

%MyApp.Users.User{}
|> Ecto.Changeset.change()
|> to_form()

Once the form is submitted, the params will be available under %{"user" => user_params}.

In the template, the form form assign can be passed to the <.form> function component:

<.form for={@form} id="todo-form" phx-change="validate" phx-submit="save">
  <.input field={@form[:field]} type="text" />
</.form>

Always give the form an explicit, unique DOM ID, like id="todo-form".

Avoiding form errors

Always use a form assigned via to_form/2 in the LiveView, and the <.input> component in the template. In the template always access forms this:

<%!-- ALWAYS do this (valid) --%>
<.form for={@form} id="my-form">
  <.input field={@form[:field]} type="text" />
</.form>

And never do this:

<%!-- NEVER do this (invalid) --%>
<.form for={@changeset} id="my-form">
  <.input field={@changeset[:field]} type="text" />
</.form>
  • You are FORBIDDEN from accessing the changeset in the template as it will cause errors
  • Never use <.form let={f} ...> in the template, instead always use <.form for={@form} ...>, then drive all form references from the form assign as in @form[:field]. The UI should always be driven by a to_form/2 assigned in the LiveView module that is derived from a changeset

Test-driven development workflow

We are driving the GitHub-compatible Issues and Projects API through tests. This section applies while we build that surface.

Principles

  • One endpoint at a time. Each new endpoint starts with a failing test before the route, controller, context, or schema exists.
  • Tests are the contract. The GitHub REST API description in docs/github-api-issues-projects-assessment.md is the source of truth for paths, status codes, and JSON shape.
  • Use OpenAgentsWeb.ConnCase for controller and JSON API tests, and OpenAgents.DataCase for domain tests.
  • Use Req for outbound fixtures we mirror. Prefer captured GitHub API responses stored in test/fixtures/github/*.json.
  • Focus each test on one thing: status, a few key fields, and the link between the request and the database.
  • Avoid Process.sleep/1 in tests. Use monitors and assert_receive when waiting for process teardown.

Order of work

Build the endpoints in this order:

  1. GET /api/v1/repos/{owner}/{repo}/issues — list
  2. GET /api/v1/repos/{owner}/{repo}/issues/{issue_number} — get
  3. POST /api/v1/repos/{owner}/{repo}/issues — create
  4. PATCH /api/v1/repos/{owner}/{repo}/issues/{issue_number} — update
  5. GET/POST /api/v1/repos/{owner}/{repo}/issues/{issue_number}/comments — comments
  6. GET/POST/DELETE /api/v1/repos/{owner}/{repo}/issues/{issue_number}/assignees — assignees
  7. GET/POST/DELETE /api/v1/repos/{owner}/{repo}/issues/{issue_number}/labels — labels
  8. GET/POST/DELETE /api/v1/repos/{owner}/{repo}/milestones — milestones
  9. /api/v1/users/{username}/projectsV2/* — project read and write

Test anatomy

Each endpoint test file follows this shape:

defmodule OpenAgentsWeb.IssueControllerTest do
  use OpenAgentsWeb.ConnCase

  test "GET /api/v1/repos/:owner/:repo/issues lists open issues", %{conn: conn} do
    conn = get(conn, ~p"/api/v1/repos/OpenAgentsInc/openagents.com/issues")
    assert json_response(conn, 200)["issues"] != nil
  end
end

Running tests

  • Run one file: mix test test/openagents_web/controllers/issue_controller_test.exs
  • Run a failed set: mix test --failed
  • Before a final commit: run mix precommit and fix all compile warnings, formatting, and test failures.