working on voice control
Some checks failed
CI/CD Pipeline / build (push) Failing after 4s

This commit is contained in:
2026-03-24 15:06:53 -06:00
parent f514012396
commit 86ff82a015
16 changed files with 704 additions and 173 deletions

View File

@@ -20,7 +20,7 @@
--color-seafoam-700: hsl(192.92 72.28% 30.98%); --color-seafoam-700: hsl(192.92 72.28% 30.98%);
--color-seafoam-800: hsl(194.38 59.57% 27.06%); --color-seafoam-800: hsl(194.38 59.57% 27.06%);
--color-seafoam-900: hsl(198.18 73.33% 17.65%); --color-seafoam-900: hsl(198.18 73.33% 17.65%);
--color-seafoam-950: hsl(198.26 58.97% 7.65%); --color-seafoam-950: hsl(196.88 72.73% 8.63%);
} }
@variant phx-click-loading (&.phx-click-loading, .phx-click-loading &); @variant phx-click-loading (&.phx-click-loading, .phx-click-loading &);

View File

@@ -92,7 +92,7 @@
} }
.markdown table { .markdown table {
@apply block w-full border-collapse my-4 text-sm overflow-x-auto; @apply block w-full border-collapse my-4 text-sm overflow-x-auto max-w-full;
} }
.markdown thead { .markdown thead {
@apply bg-seafoam-950; @apply bg-seafoam-950;

View File

@@ -30,6 +30,18 @@ const VoiceControl = {
// Button clicks dispatch DOM events to avoid a server round-trip // Button clicks dispatch DOM events to avoid a server round-trip
this.el.addEventListener("voice:start", () => this.startRecording()); this.el.addEventListener("voice:start", () => this.startRecording());
this.el.addEventListener("voice:stop", () => this.stopRecording()); this.el.addEventListener("voice:stop", () => this.stopRecording());
// Handle navigate_to from the server — trigger a live navigation so the
// root layout (and this VoiceLive) is preserved across page changes.
this.handleEvent("navigate_to", ({ path }) => {
let a = document.createElement("a");
a.href = path;
a.setAttribute("data-phx-link", "redirect");
a.setAttribute("data-phx-link-state", "push");
document.body.appendChild(a);
a.click();
a.remove();
});
}, },
destroyed() { destroyed() {

View File

@@ -135,6 +135,11 @@ defmodule ElixirAi.AiUtils.StreamLineUtils do
:ok :ok
end end
def handle_stream_line(server, "proxy error" <> _ = error) when is_binary(error) do
Logger.error("Proxy error in AI stream: #{error}")
send(server, {:stream, {:ai_request_error, error}})
end
def handle_stream_line(server, json) when is_binary(json) do def handle_stream_line(server, json) when is_binary(json) do
case Jason.decode(json) do case Jason.decode(json) do
{:ok, body} -> {:ok, body} ->

View File

@@ -1,7 +1,7 @@
defmodule ElixirAi.ChatRunner do defmodule ElixirAi.ChatRunner do
require Logger require Logger
use GenServer use GenServer
alias ElixirAi.{AiTools, Conversation, Message} alias ElixirAi.{AiTools, Conversation, Message, SystemPrompts}
import ElixirAi.PubsubTopics import ElixirAi.PubsubTopics
import ElixirAi.ChatRunner.OutboundHelpers import ElixirAi.ChatRunner.OutboundHelpers
@@ -94,6 +94,12 @@ defmodule ElixirAi.ChatRunner do
_ -> "auto" _ -> "auto"
end end
system_prompt =
case Conversation.find_category(name) do
{:ok, category} -> SystemPrompts.for_category(category)
_ -> nil
end
server_tools = AiTools.build_server_tools(self(), allowed_tools) server_tools = AiTools.build_server_tools(self(), allowed_tools)
liveview_tools = AiTools.build_liveview_tools(self(), allowed_tools) liveview_tools = AiTools.build_liveview_tools(self(), allowed_tools)
@@ -106,7 +112,7 @@ defmodule ElixirAi.ChatRunner do
ElixirAi.ChatUtils.request_ai_response( ElixirAi.ChatUtils.request_ai_response(
self(), self(),
messages, messages_with_system_prompt(messages, system_prompt),
server_tools ++ liveview_tools, server_tools ++ liveview_tools,
provider, provider,
tool_choice tool_choice
@@ -117,6 +123,7 @@ defmodule ElixirAi.ChatRunner do
%{ %{
name: name, name: name,
messages: messages, messages: messages,
system_prompt: system_prompt,
streaming_response: nil, streaming_response: nil,
pending_tool_calls: [], pending_tool_calls: [],
allowed_tools: allowed_tools, allowed_tools: allowed_tools,

View File

@@ -10,7 +10,7 @@ defmodule ElixirAi.ChatRunner.ConversationCalls do
ElixirAi.ChatUtils.request_ai_response( ElixirAi.ChatUtils.request_ai_response(
self(), self(),
new_state.messages, messages_with_system_prompt(new_state.messages, state.system_prompt),
state.server_tools ++ state.liveview_tools, state.server_tools ++ state.liveview_tools,
state.provider, state.provider,
effective_tool_choice effective_tool_choice

View File

@@ -18,4 +18,7 @@ defmodule ElixirAi.ChatRunner.OutboundHelpers do
message message
end end
def messages_with_system_prompt(messages, nil), do: messages
def messages_with_system_prompt(messages, prompt), do: [prompt | messages]
end end

View File

@@ -159,7 +159,7 @@ defmodule ElixirAi.ChatRunner.StreamHandler do
ElixirAi.ChatUtils.request_ai_response( ElixirAi.ChatUtils.request_ai_response(
self(), self(),
state.messages ++ [new_message], messages_with_system_prompt(state.messages ++ [new_message], state.system_prompt),
state.server_tools ++ state.liveview_tools, state.server_tools ++ state.liveview_tools,
state.provider, state.provider,
state.tool_choice state.tool_choice

View File

@@ -101,6 +101,17 @@ defmodule ElixirAi.Conversation do
end end
end end
def find_category(name) do
sql = "SELECT category FROM conversations WHERE name = $(name) LIMIT 1"
params = %{"name" => name}
case DbHelpers.run_sql(sql, params, "conversations") do
{:error, :db_error} -> {:error, :db_error}
[] -> {:error, :not_found}
[row | _] -> {:ok, row["category"] || "user-web"}
end
end
def find_allowed_tools(name) do def find_allowed_tools(name) do
sql = "SELECT allowed_tools FROM conversations WHERE name = $(name) LIMIT 1" sql = "SELECT allowed_tools FROM conversations WHERE name = $(name) LIMIT 1"
params = %{"name" => name} params = %{"name" => name}

View File

@@ -0,0 +1,14 @@
defmodule ElixirAi.SystemPrompts do
@prompts %{
"voice" =>
"You are responding to voice-transcribed input. Keep replies concise and conversational. The user spoke aloud and their message was transcribed, so minor transcription errors may be present.",
"user-web" => nil
}
def for_category(category) do
case Map.get(@prompts, category) do
nil -> nil
prompt -> %{role: :system, content: prompt}
end
end
end

View File

@@ -260,8 +260,17 @@ defmodule ElixirAiWeb.ChatLive do
def handle_info({:ai_request_error, reason}, socket) do def handle_info({:ai_request_error, reason}, socket) do
error_message = error_message =
case reason do case reason do
%{__struct__: mod, reason: r} -> "#{inspect(mod)}: #{inspect(r)}" "proxy error" <> _ ->
_ -> inspect(reason) "Could not connect to AI provider. Please check your proxy and provider settings."
%{__struct__: mod, reason: r} ->
"#{inspect(mod)}: #{inspect(r)}"
msg when is_binary(msg) ->
msg
_ ->
inspect(reason)
end end
{:noreply, assign(socket, ai_error: error_message, streaming_response: nil)} {:noreply, assign(socket, ai_error: error_message, streaming_response: nil)}

View File

@@ -1,8 +1,9 @@
defmodule ElixirAiWeb.ChatMessage do defmodule ElixirAiWeb.ChatMessage do
use Phoenix.Component use Phoenix.Component
alias Phoenix.LiveView.JS alias Phoenix.LiveView.JS
import ElixirAiWeb.JsonDisplay
defp max_width_class, do: "max-w-300" defp max_width_class, do: "max-w-full xl:max-w-300"
attr :content, :string, required: true attr :content, :string, required: true
attr :tool_call_id, :string, required: true attr :tool_call_id, :string, required: true
@@ -38,7 +39,7 @@ defmodule ElixirAiWeb.ChatMessage do
def user_message(assigns) do def user_message(assigns) do
~H""" ~H"""
<div class="mb-2 text-sm text-right"> <div class="mb-2 text-sm text-right">
<div class={"inline-block px-3 py-2 rounded-lg bg-seafoam-950 text-seafoam-50 #{max_width_class()} text-left"}> <div class={"w-fit px-3 py-2 rounded-lg bg-seafoam-950 text-seafoam-50 #{max_width_class()} text-left"}>
{@content} {@content}
</div> </div>
</div> </div>
@@ -78,7 +79,7 @@ defmodule ElixirAiWeb.ChatMessage do
# chunks instead of re-rendering the full markdown on every token. # chunks instead of re-rendering the full markdown on every token.
def streaming_assistant_message(assigns) do def streaming_assistant_message(assigns) do
~H""" ~H"""
<div class="mb-2 text-sm text-left"> <div class="mb-2 text-sm text-left min-w-0">
<!-- Reasoning section — only shown once reasoning_content is non-empty. <!-- Reasoning section — only shown once reasoning_content is non-empty.
The div is always in the DOM so the hook mounts before chunks arrive. --> The div is always in the DOM so the hook mounts before chunks arrive. -->
<div id="stream-reasoning-wrap"> <div id="stream-reasoning-wrap">
@@ -127,7 +128,7 @@ defmodule ElixirAiWeb.ChatMessage do
phx-hook="MarkdownStream" phx-hook="MarkdownStream"
phx-update="ignore" phx-update="ignore"
data-event="md_chunk" data-event="md_chunk"
class={"inline-block px-3 py-2 rounded-lg #{max_width_class()} markdown bg-seafoam-950/50"} class={"w-fit px-3 py-2 rounded-lg #{max_width_class()} markdown bg-seafoam-950/50 overflow-x-auto"}
> >
</div> </div>
</div> </div>
@@ -142,7 +143,7 @@ defmodule ElixirAiWeb.ChatMessage do
defp message_bubble(assigns) do defp message_bubble(assigns) do
~H""" ~H"""
<div class="mb-2 text-sm text-left"> <div class="mb-2 text-sm text-left min-w-0">
<%= if @reasoning_content && @reasoning_content != "" do %> <%= if @reasoning_content && @reasoning_content != "" do %>
<button <button
type="button" type="button"
@@ -191,7 +192,7 @@ defmodule ElixirAiWeb.ChatMessage do
phx-hook="MarkdownRender" phx-hook="MarkdownRender"
phx-update="ignore" phx-update="ignore"
data-md={@content} data-md={@content}
class={"inline-block px-3 py-2 rounded-lg #{max_width_class()} markdown bg-seafoam-950/50"} class={"w-fit px-3 py-2 rounded-lg #{max_width_class()} markdown bg-seafoam-950/50 overflow-x-auto"}
> >
</div> </div>
<% end %> <% end %>
@@ -253,47 +254,49 @@ defmodule ElixirAiWeb.ChatMessage do
) )
~H""" ~H"""
<div class={[ <div
"mb-1 #{max_width_class()} rounded-lg border text-xs font-mono overflow-hidden bg-seafoam-950/40", id={@_id}
@state == :error && "border-red-900/50", class={[
@state == :called && "border-seafoam-900/60", "mb-1 #{max_width_class()} rounded-lg border text-xs font-mono overflow-hidden bg-seafoam-950/40",
@state in [:pending, :success] && "border-seafoam-900" @state == :error && "border-red-900/50",
]}> @state == :called && "border-seafoam-900/60",
<div class={[ @state in [:pending, :success] && "border-seafoam-900"
"flex items-center gap-2 px-3 py-1.5 border-b text-seafoam-400", ]}
@state == :error && "border-red-900/50 bg-red-900/20", >
@state == :called && "border-seafoam-900/60 bg-seafoam-900/20", <div
@state in [:pending, :success] && "border-seafoam-900 bg-seafoam-900/30" class={[
]}> "flex items-center gap-2 px-3 py-1.5 border-b text-seafoam-400",
<.tool_call_icon /> @_truncated && "cursor-pointer select-none",
<span class="text-seafoam-300 font-semibold shrink-0">{@name}</span> @state == :error && "border-red-900/50 bg-red-900/20",
<span :if={@_truncated} class="text-seafoam-600/50 truncate flex-1 min-w-0 ml-1"> @state == :called && "border-seafoam-900/60 bg-seafoam-900/20",
{@_truncated} @state in [:pending, :success] && "border-seafoam-900 bg-seafoam-900/30"
</span> ]}
<span :if={!@_truncated} class="flex-1" /> phx-click={
<button @_truncated &&
:if={@_truncated}
type="button"
phx-click={
JS.toggle_class("hidden", to: "##{@_id}-args") JS.toggle_class("hidden", to: "##{@_id}-args")
|> JS.toggle_class("rotate-180", to: "##{@_id}-chevron") |> JS.toggle_class("rotate-180", to: "##{@_id}-chevron")
} }
class="shrink-0 text-seafoam-700 hover:text-seafoam-400 transition-colors mx-1" >
<.tool_call_icon />
<span class="text-seafoam-400 font-semibold shrink-0">{@name}</span>
<span :if={@_truncated} class="text-seafoam-500 truncate flex-1 min-w-0 ml-1">
<.json_display json={@_truncated} inline />
</span>
<span :if={!@_truncated} class="flex-1" />
<svg
:if={@_truncated}
id={"#{@_id}-chevron"}
xmlns="http://www.w3.org/2000/svg"
viewBox="0 0 16 16"
fill="currentColor"
class="w-3 h-3 shrink-0 mx-1 text-seafoam-700 transition-transform duration-200"
> >
<svg <path
id={"#{@_id}-chevron"} fill-rule="evenodd"
xmlns="http://www.w3.org/2000/svg" d="M4.22 6.22a.75.75 0 0 1 1.06 0L8 8.94l2.72-2.72a.75.75 0 1 1 1.06 1.06l-3.25 3.25a.75.75 0 0 1-1.06 0L4.22 7.28a.75.75 0 0 1 0-1.06Z"
viewBox="0 0 16 16" clip-rule="evenodd"
fill="currentColor" />
class="w-3 h-3 transition-transform duration-200" </svg>
>
<path
fill-rule="evenodd"
d="M4.22 6.22a.75.75 0 0 1 1.06 0L8 8.94l2.72-2.72a.75.75 0 1 1 1.06 1.06l-3.25 3.25a.75.75 0 0 1-1.06 0L4.22 7.28a.75.75 0 0 1 0-1.06Z"
clip-rule="evenodd"
/>
</svg>
</button>
<span :if={@state == :called} class="flex items-center gap-1 text-seafoam-500/50 shrink-0"> <span :if={@state == :called} class="flex items-center gap-1 text-seafoam-500/50 shrink-0">
<svg <svg
xmlns="http://www.w3.org/2000/svg" xmlns="http://www.w3.org/2000/svg"
@@ -376,26 +379,10 @@ defmodule ElixirAiWeb.ChatMessage do
attr :arguments, :any, default: nil attr :arguments, :any, default: nil
defp tool_call_args(%{arguments: args} = assigns) when not is_nil(args) and args != "" do defp tool_call_args(%{arguments: args} = assigns) when not is_nil(args) and args != "" do
assigns =
assign(
assigns,
:pretty_args,
case args do
s when is_binary(s) ->
case Jason.decode(s) do
{:ok, decoded} -> Jason.encode!(decoded, pretty: true)
_ -> s
end
other ->
Jason.encode!(other, pretty: true)
end
)
~H""" ~H"""
<div class="px-3 py-2 border-b border-seafoam-900/50"> <div class="px-3 py-2 border-b border-seafoam-900/50">
<div class="text-seafoam-500 mb-1 uppercase tracking-wider text-[10px]">arguments</div> <div class="text-seafoam-500 mb-1 uppercase tracking-wider text-[10px]">arguments</div>
<pre class="text-seafoam-400 whitespace-pre-wrap break-all">{@pretty_args}</pre> <.json_display json={@arguments} />
</div> </div>
""" """
end end

View File

@@ -0,0 +1,132 @@
defmodule ElixirAiWeb.JsonDisplay do
use Phoenix.Component
attr :json, :any, required: true
attr :class, :string, default: nil
attr :inline, :boolean, default: false
def json_display(%{json: json, inline: inline} = assigns) do
formatted =
case json do
nil ->
""
"" ->
""
s when is_binary(s) ->
case Jason.decode(s) do
{:ok, decoded} -> Jason.encode!(decoded, pretty: !inline)
_ -> s
end
other ->
Jason.encode!(other, pretty: !inline)
end
assigns = assign(assigns, :_highlighted, json_to_html(formatted))
~H"""
<pre
:if={!@inline}
class={["whitespace-pre-wrap break-all text-xs font-mono leading-relaxed", @class]}
><%= @_highlighted %></pre>
<span :if={@inline} class={["text-xs font-mono truncate", @class]}>{@_highlighted}</span>
"""
end
@token_colors %{
key: "text-sky-300",
string: "text-emerald-400/80",
keyword: "text-violet-400",
number: "text-orange-300/80",
colon: "text-seafoam-500/50",
punctuation: "text-seafoam-500/50",
quote: "text-seafoam-500/50"
}
# Converts a plain JSON string into a Phoenix.HTML.safe value with
# <span> tokens coloured by token type.
defp json_to_html(""), do: Phoenix.HTML.raw("")
defp json_to_html(str) do
# Capture groups (in order):
# 1 string literal "..."
# 2 keyword true | false | null
# 3 number -?digits with optional frac/exp
# 4 punctuation { } [ ] , :
# 5 whitespace spaces / newlines / tabs
# 6 fallback any other single char
token_re =
~r/(".(?:[^"\\]|\\.)*")|(true|false|null)|(-?(?:0|[1-9]\d*)(?:\.\d+)?(?:[eE][+-]?\d+)?)|([{}\[\],:])|(\s+)|(.)/s
tokens = Regex.scan(token_re, str, capture: :all_but_first)
{parts, _, _} =
Enum.reduce(tokens, {[], :val, []}, fn groups, {acc, state, ctx} ->
[string_tok, keyword_tok, number_tok, punct_tok, whitespace_tok, fallback_tok] =
pad_groups(groups, 6)
cond do
string_tok != "" ->
{color, next_state} =
if state == :key,
do: {@token_colors.key, :after_key},
else: {@token_colors.string, :after_val}
content = string_tok |> String.slice(1..-2//1) |> html_escape()
quote_span = color_span(@token_colors.quote, "&quot;")
{[quote_span <> color_span(color, content) <> quote_span | acc], next_state, ctx}
keyword_tok != "" ->
{[color_span(@token_colors.keyword, keyword_tok) | acc], :after_val, ctx}
number_tok != "" ->
{[color_span(@token_colors.number, number_tok) | acc], :after_val, ctx}
punct_tok != "" ->
{next_state, next_ctx} = advance_state(punct_tok, state, ctx)
color = if punct_tok == ":", do: @token_colors.colon, else: @token_colors.punctuation
{[color_span(color, punct_tok) | acc], next_state, next_ctx}
whitespace_tok != "" ->
{[whitespace_tok | acc], state, ctx}
fallback_tok != "" ->
{[html_escape(fallback_tok) | acc], state, ctx}
true ->
{acc, state, ctx}
end
end)
Phoenix.HTML.raw(parts |> Enum.reverse() |> Enum.join())
end
# State transitions driven by punctuation tokens.
# State :key → we are about to read an object key.
# State :val → we are about to read a value.
# State :after_key / :after_val → consumed the token; awaiting : or ,.
defp advance_state("{", _, ctx), do: {:key, [:obj | ctx]}
defp advance_state("[", _, ctx), do: {:val, [:arr | ctx]}
defp advance_state("}", _, [_ | ctx]), do: {:after_val, ctx}
defp advance_state("}", _, []), do: {:after_val, []}
defp advance_state("]", _, [_ | ctx]), do: {:after_val, ctx}
defp advance_state("]", _, []), do: {:after_val, []}
defp advance_state(":", _, ctx), do: {:val, ctx}
defp advance_state(",", _, [:obj | _] = ctx), do: {:key, ctx}
defp advance_state(",", _, ctx), do: {:val, ctx}
defp advance_state(_, state, ctx), do: {state, ctx}
defp pad_groups(list, n), do: list ++ List.duplicate("", max(0, n - length(list)))
defp color_span(class, content), do: ~s|<span class="#{class}">#{content}</span>|
defp html_escape(str) do
str
|> String.replace("&", "&amp;")
|> String.replace("<", "&lt;")
|> String.replace(">", "&gt;")
end
end

View File

@@ -0,0 +1,90 @@
defmodule ElixirAiWeb.Voice.Recording do
use Phoenix.Component
alias Phoenix.LiveView.JS
attr :state, :atom, required: true
def recording(assigns) do
~H"""
<div class="p-4 flex flex-col gap-3">
<div class="flex items-center justify-between">
<div class="flex items-center gap-3">
<%= if @state == :idle do %>
<svg
xmlns="http://www.w3.org/2000/svg"
class="h-4 w-4 text-seafoam-500 shrink-0"
viewBox="0 0 24 24"
fill="currentColor"
>
<path d="M12 1a4 4 0 0 1 4 4v7a4 4 0 0 1-8 0V5a4 4 0 0 1 4-4zm0 2a2 2 0 0 0-2 2v7a2 2 0 1 0 4 0V5a2 2 0 0 0-2-2zm-7 9a7 7 0 0 0 14 0h2a9 9 0 0 1-8 8.94V23h-2v-2.06A9 9 0 0 1 3 12H5z" />
</svg>
<span class="text-seafoam-400 font-semibold text-sm">Voice Input</span>
<% end %>
<%= if @state == :recording do %>
<span class="relative flex h-3 w-3 shrink-0">
<span class="animate-ping absolute inline-flex h-full w-full rounded-full bg-red-500 opacity-75">
</span>
<span class="relative inline-flex rounded-full h-3 w-3 bg-red-500"></span>
</span>
<span class="text-seafoam-50 font-semibold text-sm">Recording</span>
<% end %>
<%= if @state == :processing do %>
<span class="relative flex h-3 w-3 shrink-0">
<span class="animate-ping absolute inline-flex h-full w-full rounded-full bg-seafoam-400 opacity-75">
</span>
<span class="relative inline-flex rounded-full h-3 w-3 bg-seafoam-400"></span>
</span>
<span class="text-seafoam-50 font-semibold text-sm">Processing…</span>
<% end %>
</div>
<button
phx-click="minimize"
title="Minimize"
class="p-1 rounded-lg text-seafoam-600 hover:text-seafoam-300 hover:bg-seafoam-800/50 transition-colors"
>
<svg
xmlns="http://www.w3.org/2000/svg"
class="h-4 w-4"
viewBox="0 0 24 24"
fill="none"
stroke="currentColor"
stroke-width="2"
stroke-linecap="round"
stroke-linejoin="round"
>
<path d="M5 12h14" />
</svg>
</button>
</div>
<%= if @state in [:recording, :processing] do %>
<div id="voice-viz-wrapper" phx-update="ignore">
<canvas id="voice-viz-canvas" height="72" class="w-full rounded-lg bg-seafoam-950 block">
</canvas>
</div>
<% end %>
<%= if @state == :idle do %>
<button
phx-click={JS.dispatch("voice:start", to: "#voice-control-hook")}
class="w-full flex items-center justify-between px-3 py-1.5 rounded-lg bg-seafoam-700 hover:bg-seafoam-600 text-seafoam-50 text-xs font-medium transition-colors"
>
<span>Start Recording</span>
<kbd class="text-seafoam-300 bg-seafoam-800 border border-seafoam-600 px-1.5 py-0.5 rounded font-mono">
Ctrl+Space
</kbd>
</button>
<% end %>
<%= if @state == :recording do %>
<button
phx-click={JS.dispatch("voice:stop", to: "#voice-control-hook")}
class="w-full flex items-center justify-between px-3 py-1.5 rounded-lg bg-seafoam-800 hover:bg-seafoam-700 text-seafoam-50 text-xs font-medium transition-colors border border-seafoam-700"
>
<span>Stop Recording</span>
<kbd class="text-seafoam-300 bg-seafoam-900 border border-seafoam-700 px-1.5 py-0.5 rounded font-mono">
Space
</kbd>
</button>
<% end %>
</div>
"""
end
end

View File

@@ -0,0 +1,97 @@
defmodule ElixirAiWeb.Voice.VoiceConversation do
use Phoenix.Component
alias Phoenix.LiveView.JS
import ElixirAiWeb.ChatMessage
import ElixirAiWeb.Spinner
attr :messages, :list, required: true
attr :streaming_response, :any, default: nil
attr :ai_error, :string, default: nil
def voice_conversation(assigns) do
~H"""
<div class="flex flex-col flex-1 overflow-hidden">
<div class="flex items-center justify-between px-4 pt-4 pb-2">
<span class="text-seafoam-300 font-semibold text-sm">Voice Chat</span>
<button
phx-click="minimize"
title="Minimize"
class="p-1 rounded-lg text-seafoam-600 hover:text-seafoam-300 hover:bg-seafoam-800/50 transition-colors"
>
<svg
xmlns="http://www.w3.org/2000/svg"
class="h-4 w-4"
viewBox="0 0 24 24"
fill="none"
stroke="currentColor"
stroke-width="2"
stroke-linecap="round"
stroke-linejoin="round"
>
<path d="M5 12h14" />
</svg>
</button>
</div>
<%= if @ai_error do %>
<div class="mx-4 mt-1 px-3 py-2 rounded text-sm text-red-400 bg-red-950/40" role="alert">
AI error: {@ai_error}
</div>
<% end %>
<div
id="voice-chat-messages"
phx-hook="ScrollBottom"
class="flex-1 overflow-y-auto px-4 py-2 space-y-1"
>
<%= for msg <- @messages do %>
<%= cond do %>
<% msg.role == :user -> %>
<.user_message content={Map.get(msg, :content) || ""} />
<% msg.role == :tool -> %>
<.tool_result_message
content={Map.get(msg, :content) || ""}
tool_call_id={Map.get(msg, :tool_call_id) || ""}
/>
<% true -> %>
<.assistant_message
content={Map.get(msg, :content) || ""}
reasoning_content={Map.get(msg, :reasoning_content)}
tool_calls={Map.get(msg, :tool_calls) || []}
/>
<% end %>
<% end %>
<%= if @streaming_response do %>
<.streaming_assistant_message
content={@streaming_response.content}
reasoning_content={@streaming_response.reasoning_content}
tool_calls={@streaming_response.tool_calls}
/>
<.spinner />
<% end %>
</div>
<div class="px-4 pb-3 pt-2 flex items-center justify-between gap-2">
<button
phx-click="dismiss_transcription"
class="text-xs text-seafoam-500 hover:text-seafoam-300 transition-colors"
>
Dismiss
</button>
<button
phx-click={JS.dispatch("voice:start", to: "#voice-control-hook")}
title="Voice input (Ctrl+Space)"
class="flex items-center gap-1.5 px-3 py-1.5 rounded-lg bg-seafoam-700 hover:bg-seafoam-600 text-seafoam-50 text-xs font-medium transition-colors"
>
<svg
xmlns="http://www.w3.org/2000/svg"
class="h-3.5 w-3.5"
viewBox="0 0 24 24"
fill="currentColor"
>
<path d="M12 1a4 4 0 0 1 4 4v7a4 4 0 0 1-8 0V5a4 4 0 0 1 4-4zm0 2a2 2 0 0 0-2 2v7a2 2 0 1 0 4 0V5a2 2 0 0 0-2-2zm-7 9a7 7 0 0 0 14 0h2a9 9 0 0 1-8 8.94V23h-2v-2.06A9 9 0 0 1 3 12H5z" />
</svg>
<span>Record</span>
</button>
</div>
</div>
"""
end
end

View File

@@ -2,15 +2,29 @@ defmodule ElixirAiWeb.VoiceLive do
use ElixirAiWeb, :live_view use ElixirAiWeb, :live_view
require Logger require Logger
alias ElixirAiWeb.Voice.Recording
alias ElixirAiWeb.Voice.VoiceConversation
alias ElixirAi.{AiProvider, ChatRunner, ConversationManager}
import ElixirAi.PubsubTopics
def mount(_params, _session, socket) do def mount(_params, _session, socket) do
{:ok, assign(socket, state: :idle, transcription: nil, expanded: false), layout: false} {:ok,
assign(socket,
state: :idle,
transcription: nil,
expanded: false,
conversation_name: nil,
messages: [],
streaming_response: nil,
runner_pid: nil,
ai_error: nil
), layout: false}
end end
def render(assigns) do def render(assigns) do
~H""" ~H"""
<div id="voice-control-hook" phx-hook="VoiceControl"> <div id="voice-control-hook" phx-hook="VoiceControl">
<%= if not @expanded do %> <%= if not @expanded do %>
<%!-- Collapsed: semi-transparent mic button, still listens to Ctrl+Space via hook --%>
<button <button
phx-click="expand" phx-click="expand"
title="Voice input (Ctrl+Space)" title="Voice input (Ctrl+Space)"
@@ -26,99 +40,18 @@ defmodule ElixirAiWeb.VoiceLive do
</svg> </svg>
</button> </button>
<% else %> <% else %>
<%!-- Expanded panel --%> <div class={[
<div class="fixed top-4 right-4 w-72 bg-seafoam-950/95 border border-seafoam-800 rounded-2xl shadow-2xl z-50 p-4 flex flex-col gap-3 backdrop-blur"> "fixed top-4 right-4 z-50 bg-seafoam-900 border border-seafoam-800 rounded-2xl shadow-2xl flex flex-col backdrop-blur",
<div class="flex items-center justify-between"> if(@state == :transcribed, do: "w-96 max-h-[80vh]", else: "w-72")
<div class="flex items-center gap-3"> ]}>
<%= if @state == :idle do %>
<svg
xmlns="http://www.w3.org/2000/svg"
class="h-4 w-4 text-seafoam-500 shrink-0"
viewBox="0 0 24 24"
fill="currentColor"
>
<path d="M12 1a4 4 0 0 1 4 4v7a4 4 0 0 1-8 0V5a4 4 0 0 1 4-4zm0 2a2 2 0 0 0-2 2v7a2 2 0 1 0 4 0V5a2 2 0 0 0-2-2zm-7 9a7 7 0 0 0 14 0h2a9 9 0 0 1-8 8.94V23h-2v-2.06A9 9 0 0 1 3 12H5z" />
</svg>
<span class="text-seafoam-400 font-semibold text-sm">Voice Input</span>
<% end %>
<%= if @state == :recording do %>
<span class="relative flex h-3 w-3 shrink-0">
<span class="animate-ping absolute inline-flex h-full w-full rounded-full bg-red-500 opacity-75">
</span>
<span class="relative inline-flex rounded-full h-3 w-3 bg-red-500"></span>
</span>
<span class="text-seafoam-50 font-semibold text-sm">Recording</span>
<% end %>
<%= if @state == :processing do %>
<span class="relative flex h-3 w-3 shrink-0">
<span class="animate-ping absolute inline-flex h-full w-full rounded-full bg-seafoam-400 opacity-75">
</span>
<span class="relative inline-flex rounded-full h-3 w-3 bg-seafoam-400"></span>
</span>
<span class="text-seafoam-50 font-semibold text-sm">Processing…</span>
<% end %>
<%= if @state == :transcribed do %>
<span class="text-seafoam-300 font-semibold text-sm">Transcription</span>
<% end %>
</div>
<%!-- Minimize button --%>
<button
phx-click="minimize"
title="Minimize"
class="p-1 rounded-lg text-seafoam-600 hover:text-seafoam-300 hover:bg-seafoam-800/50 transition-colors"
>
<svg
xmlns="http://www.w3.org/2000/svg"
class="h-4 w-4"
viewBox="0 0 24 24"
fill="none"
stroke="currentColor"
stroke-width="2"
stroke-linecap="round"
stroke-linejoin="round"
>
<path d="M5 12h14" />
</svg>
</button>
</div>
<%= if @state in [:recording, :processing] do %>
<div id="voice-viz-wrapper" phx-update="ignore">
<canvas id="voice-viz-canvas" height="72" class="w-full rounded-lg bg-seafoam-950 block">
</canvas>
</div>
<% end %>
<%= if @state == :transcribed do %> <%= if @state == :transcribed do %>
<.transcription_display transcription={@transcription} /> <VoiceConversation.voice_conversation
<% end %> messages={@messages}
<%= if @state == :idle do %> streaming_response={@streaming_response}
<button ai_error={@ai_error}
phx-click={JS.dispatch("voice:start", to: "#voice-control-hook")} />
class="w-full flex items-center justify-between px-3 py-1.5 rounded-lg bg-seafoam-700 hover:bg-seafoam-600 text-seafoam-50 text-xs font-medium transition-colors" <% else %>
> <Recording.recording state={@state} />
<span>Start Recording</span>
<kbd class="text-seafoam-300 bg-seafoam-800 border border-seafoam-600 px-1.5 py-0.5 rounded font-mono">
Ctrl+Space
</kbd>
</button>
<% end %>
<%= if @state == :recording do %>
<button
phx-click={JS.dispatch("voice:stop", to: "#voice-control-hook")}
class="w-full flex items-center justify-between px-3 py-1.5 rounded-lg bg-seafoam-800 hover:bg-seafoam-700 text-seafoam-50 text-xs font-medium transition-colors border border-seafoam-700"
>
<span>Stop Recording</span>
<kbd class="text-seafoam-300 bg-seafoam-900 border border-seafoam-700 px-1.5 py-0.5 rounded font-mono">
Space
</kbd>
</button>
<% end %>
<%= if @state == :transcribed do %>
<button
phx-click="dismiss_transcription"
class="text-xs text-seafoam-500 hover:text-seafoam-300 transition-colors text-center w-full"
>
Dismiss
</button>
<% end %> <% end %>
</div> </div>
<% end %> <% end %>
@@ -126,14 +59,6 @@ defmodule ElixirAiWeb.VoiceLive do
""" """
end end
defp transcription_display(assigns) do
~H"""
<div class="rounded-xl bg-seafoam-900/60 border border-seafoam-700 px-3 py-2">
<p class="text-sm text-seafoam-50 leading-relaxed">{@transcription}</p>
</div>
"""
end
def handle_event("expand", _params, socket) do def handle_event("expand", _params, socket) do
{:noreply, assign(socket, expanded: true)} {:noreply, assign(socket, expanded: true)}
end end
@@ -168,15 +93,254 @@ defmodule ElixirAiWeb.VoiceLive do
end end
def handle_event("dismiss_transcription", _params, socket) do def handle_event("dismiss_transcription", _params, socket) do
{:noreply, assign(socket, state: :idle, transcription: nil, expanded: false)} name = socket.assigns.conversation_name
if name do
if socket.assigns.runner_pid do
try do
GenServer.call(socket.assigns.runner_pid, {:session, {:deregister_liveview_pid, self()}})
catch
:exit, _ -> :ok
end
end
Phoenix.PubSub.unsubscribe(ElixirAi.PubSub, chat_topic(name))
end
{:noreply,
assign(socket,
state: :idle,
transcription: nil,
expanded: false,
conversation_name: nil,
messages: [],
streaming_response: nil,
runner_pid: nil,
ai_error: nil
)}
end end
# Transcription received — open conversation and send as user message
def handle_info({:transcription_result, {:ok, text}}, socket) do def handle_info({:transcription_result, {:ok, text}}, socket) do
{:noreply, assign(socket, state: :transcribed, transcription: text)} socket = start_voice_conversation(socket, text)
{:noreply, socket}
end end
def handle_info({:transcription_result, {:error, reason}}, socket) do def handle_info({:transcription_result, {:error, reason}}, socket) do
Logger.error("VoiceLive: transcription failed: #{inspect(reason)}") Logger.error("VoiceLive: transcription failed: #{inspect(reason)}")
{:noreply, assign(socket, state: :idle)} {:noreply, assign(socket, state: :idle)}
end end
# --- Chat PubSub handlers (same pattern as ChatLive) ---
def handle_info({:user_chat_message, message}, socket) do
{:noreply,
socket
|> update(:messages, &(&1 ++ [message]))
|> push_event("scroll_to_bottom", %{})}
end
def handle_info(
{:start_ai_response_stream,
%{id: _id, reasoning_content: "", content: ""} = starting_response},
socket
) do
{:noreply, assign(socket, streaming_response: starting_response)}
end
def handle_info(
{:reasoning_chunk_content, reasoning_content},
%{assigns: %{streaming_response: nil}} = socket
) do
base = get_snapshot(socket) |> Map.update!(:reasoning_content, &(&1 <> reasoning_content))
{:noreply, assign(socket, streaming_response: base)}
end
def handle_info({:reasoning_chunk_content, reasoning_content}, socket) do
updated_response = %{
socket.assigns.streaming_response
| reasoning_content:
socket.assigns.streaming_response.reasoning_content <> reasoning_content
}
{:noreply,
socket
|> assign(streaming_response: updated_response)
|> push_event("reasoning_chunk", %{chunk: reasoning_content})}
end
def handle_info(
{:text_chunk_content, text_content},
%{assigns: %{streaming_response: nil}} = socket
) do
base = get_snapshot(socket) |> Map.update!(:content, &(&1 <> text_content))
{:noreply, assign(socket, streaming_response: base)}
end
def handle_info({:text_chunk_content, text_content}, socket) do
updated_response = %{
socket.assigns.streaming_response
| content: socket.assigns.streaming_response.content <> text_content
}
{:noreply,
socket
|> assign(streaming_response: updated_response)
|> push_event("md_chunk", %{chunk: text_content})}
end
def handle_info(:tool_calls_finished, socket) do
{:noreply, assign(socket, streaming_response: nil)}
end
def handle_info({:tool_request_message, tool_request_message}, socket) do
{:noreply, update(socket, :messages, &(&1 ++ [tool_request_message]))}
end
def handle_info({:one_tool_finished, tool_response}, socket) do
{:noreply, update(socket, :messages, &(&1 ++ [tool_response]))}
end
def handle_info({:end_ai_response, final_message}, socket) do
{:noreply,
socket
|> update(:messages, &(&1 ++ [final_message]))
|> assign(streaming_response: nil)}
end
def handle_info({:ai_request_error, reason}, socket) do
error_message =
case reason do
"proxy error" <> _ ->
"Could not connect to AI provider. Please check your proxy and provider settings."
%{__struct__: mod, reason: r} ->
"#{inspect(mod)}: #{inspect(r)}"
msg when is_binary(msg) ->
msg
_ ->
inspect(reason)
end
{:noreply, assign(socket, ai_error: error_message, streaming_response: nil)}
end
def handle_info({:db_error, reason}, socket) do
Logger.error("VoiceLive: db error: #{inspect(reason)}")
{:noreply, socket}
end
def handle_info({:liveview_tool_call, "navigate_to", %{"path" => path}}, socket) do
{:noreply, push_event(socket, "navigate_to", %{path: path})}
end
def handle_info({:liveview_tool_call, _tool_name, _args}, socket) do
{:noreply, socket}
end
def handle_info(:sync_streaming, %{assigns: %{runner_pid: pid}} = socket)
when is_pid(pid) do
case GenServer.call(pid, {:conversation, :get_streaming_response}) do
nil ->
{:noreply, assign(socket, streaming_response: nil)}
%{content: content, reasoning_content: reasoning_content} = snapshot ->
socket =
socket
|> assign(streaming_response: snapshot)
|> then(fn s ->
if content != "", do: push_event(s, "md_chunk", %{chunk: content}), else: s
end)
|> then(fn s ->
if reasoning_content != "",
do: push_event(s, "reasoning_chunk", %{chunk: reasoning_content}),
else: s
end)
{:noreply, socket}
end
end
def handle_info(:sync_streaming, socket), do: {:noreply, socket}
def handle_info(:recovery_restart, socket) do
{:noreply, assign(socket, streaming_response: nil, ai_error: nil)}
end
# --- Private helpers ---
defp start_voice_conversation(socket, transcription) do
name = "voice-#{System.system_time(:second)}"
case AiProvider.find_by_name("default") do
{:ok, provider} ->
case ConversationManager.create_conversation(name, provider.id, "voice") do
{:ok, _pid} ->
case ConversationManager.open_conversation(name) do
{:ok, conv} ->
connect_and_send(socket, name, conv, transcription)
{:error, reason} ->
assign(socket,
state: :transcribed,
ai_error: "Failed to open voice conversation: #{inspect(reason)}"
)
end
{:error, reason} ->
assign(socket,
state: :transcribed,
ai_error: "Failed to create voice conversation: #{inspect(reason)}"
)
end
{:error, reason} ->
assign(socket,
state: :transcribed,
ai_error: "No default AI provider found: #{inspect(reason)}"
)
end
end
defp connect_and_send(socket, name, conversation, transcription) do
runner_pid = Map.get(conversation, :runner_pid)
if connected?(socket) do
Phoenix.PubSub.subscribe(ElixirAi.PubSub, chat_topic(name))
if runner_pid,
do: GenServer.call(runner_pid, {:session, {:register_liveview_pid, self()}})
send(self(), :sync_streaming)
end
if runner_pid do
GenServer.cast(runner_pid, {:conversation, {:user_message, transcription, nil}})
else
ChatRunner.new_user_message(name, transcription)
end
assign(socket,
state: :transcribed,
transcription: transcription,
conversation_name: name,
messages: conversation.messages,
streaming_response: conversation.streaming_response,
runner_pid: runner_pid,
ai_error: nil
)
end
defp get_snapshot(%{assigns: %{runner_pid: pid}}) when is_pid(pid) do
case GenServer.call(pid, {:conversation, :get_streaming_response}) do
nil -> %{id: nil, content: "", reasoning_content: "", tool_calls: []}
snapshot -> snapshot
end
end
defp get_snapshot(_socket) do
%{id: nil, content: "", reasoning_content: "", tool_calls: []}
end
end end