Ingestion: Home Assistant WebSocket push path

HomeAssistantWebSocketWorker holds a persistent state_changed subscription per HA endpoint that opts in via the connector's WebSocket toggle (HaEndpointConfig.UseWebSocket): auth handshake, subscribe, ingest in real time, capped-backoff reconnect. The REST poll worker skips WS endpoints so each is served once. HaWebSocketProtocol holds the pure handshake/parse logic. Verified by 11 protocol unit tests + a live integration test against an in-process fake HA server. CLAUDE.md updated.

Claude-Session: https://claude.ai/code/session_01Kib2MniVFbD95fkgLgBBnB

CLAUDE.md

@@ -6,7 +6,7 @@ This file provides guidance to Claude Code (claude.ai/code) when working with co
 
 MeterVault is a self-hosted, local-first energy & utility metering platform: it ingests meter data from Home Assistant, Tasmota and MQTT on a schedule, stores every reading timestamped and immutable, normalizes it into consumption, and turns it into cost dashboards. Energy types (electricity, water, heating oil, gas, …) and meters are **user-defined, never hardcoded**.
 
-**Status: implemented (M0–M7) + SDD §8 panels.** The full solution is built and green — five projects, ~108 tests, working Docker deploy. `docs/SDD.md` remains the design reference; the milestone map (§12) matches the git history (M0…M7 commits). The dedicated **PV/Solar** (`/solar`), **Oil/consumable** (`/consumables`) and **meter-detail** (`/meters/{id}`) views (SDD §8.4–§8.6) are implemented as read models in `Infrastructure/Dashboard` (`SolarService`, `ConsumableService`, `MeterDetailService`) — PV meters are found by `Mode == GenerationCounter` and grid/load meters by a `role` tag in `Meter.Meta` (`MeterRoles`/`MeterMeta`), so nothing is hardcoded by name. **Admin write-CRUD** (SDD §8.7) is implemented as MudBlazor inline-dialog pages: energy types, meters (+ recompute on mode/baseline change), a meter's ingest sources (meter-detail Sources tab), tariffs, cost categories + members, and connectors (`ingestion_endpoint`, secrets by env-var reference only). `/admin/settings` is a read-only effective-config view (settings are env-driven and reproducible, not DB-stored). **Home Assistant reading** is configured here: an HA connector (`BaseUrl` + `TokenEnv`) + an HA source (entity id) drives `HomeAssistantWorker`'s REST poll; `HaConnectionTester` powers the connector "Test connection" button. **Meter topology & flow**: `MeterLink` (a directed `from→to` edge; a downstream meter is a *subsection* of an upstream one, multi-parent allowed) drives a per-energy-type page `/energy/{id}` with a hand-rolled SVG **Sankey** (`SankeyChart.razor`, since ApexCharts has no Sankey type) computed by `FlowService` (link value = downstream consumption, split proportionally across multiple parents; unaccounted remainder → an "Other" node). Upstream meters are wired cycle-safely in the meter editor; the nav lists a link per energy type. Remaining refinements (HA WebSocket *push* — REST poll works today; commit-arbitrary-CSV-from-UI needs a meter-mapping wizard; full de-DE UI-string localization) are noted at their commits. Set `MeterVault__SeedReferenceData=true` (compose: `METERVAULT_SEED=true`) for a one-command populated demo.
+**Status: implemented (M0–M7) + SDD §8 panels.** The full solution is built and green — five projects, ~108 tests, working Docker deploy. `docs/SDD.md` remains the design reference; the milestone map (§12) matches the git history (M0…M7 commits). The dedicated **PV/Solar** (`/solar`), **Oil/consumable** (`/consumables`) and **meter-detail** (`/meters/{id}`) views (SDD §8.4–§8.6) are implemented as read models in `Infrastructure/Dashboard` (`SolarService`, `ConsumableService`, `MeterDetailService`) — PV meters are found by `Mode == GenerationCounter` and grid/load meters by a `role` tag in `Meter.Meta` (`MeterRoles`/`MeterMeta`), so nothing is hardcoded by name. **Admin write-CRUD** (SDD §8.7) is implemented as MudBlazor inline-dialog pages: energy types, meters (+ recompute on mode/baseline change), a meter's ingest sources (meter-detail Sources tab), tariffs, cost categories + members, and connectors (`ingestion_endpoint`, secrets by env-var reference only). `/admin/settings` is a read-only effective-config view (settings are env-driven and reproducible, not DB-stored). **Home Assistant reading** is configured here: an HA connector (`BaseUrl` + `TokenEnv`) + an HA source (entity id) drives `HomeAssistantWorker`'s REST poll, or — with the connector's **WebSocket push** toggle (`HaEndpointConfig.UseWebSocket`) — `HomeAssistantWebSocketWorker` holds a persistent `state_changed` subscription and ingests in real time (the poll worker skips WS endpoints, so each is served once; `HaWebSocketProtocol` is the pure, unit-tested handshake/parse logic). `HaConnectionTester` powers the connector "Test connection" button. **Meter topology & flow**: `MeterLink` (a directed `from→to` edge; a downstream meter is a *subsection* of an upstream one, multi-parent allowed) drives a per-energy-type page `/energy/{id}` with a hand-rolled SVG **Sankey** (`SankeyChart.razor`, since ApexCharts has no Sankey type) computed by `FlowService` (link value = downstream consumption, split proportionally across multiple parents; unaccounted remainder → an "Other" node). Upstream meters are wired cycle-safely in the meter editor; the nav lists a link per energy type. **CSV mapping wizard** (`/import/wizard`): upload an arbitrary CSV, map columns → meters/roles, dry-run preview, then commit as a revertible `import_batch` (the `/import` page lists batches with one-click revert). The `instant_rate` mode is normalized (`InstantRateNormalizer`: rate integrated over time, trapezoidal). Remaining refinement: full de-DE UI-string localization (data parsing is already de-DE) — noted at its commit. Set `MeterVault__SeedReferenceData=true` (compose: `METERVAULT_SEED=true`) for a one-command populated demo.
 
 ## Source of truth
 

src/App/Components/Pages/Admin/Connectors.razor

@@ -71,6 +71,10 @@ else
         {
             <MudTextField @bind-Value="_working.BaseUrl" Label="Base URL (e.g. http://homeassistant.local:8123)" Class="mb-2" />
             <MudTextField @bind-Value="_working.TokenEnv" Label="Token env-var name (e.g. HA_TOKEN)" Class="mb-2" />
+            <MudSwitch T="bool" @bind-Value="_working.UseWebSocket" Label="Real-time WebSocket push" Color="Color.Primary" Class="mb-1" />
+            <MudText Typo="Typo.caption" Color="Color.Secondary" Class="mb-2 d-block">
+                On: subscribe to state changes and ingest in real time. Off: REST poll on each source's interval.
+            </MudText>
             <MudTextField @bind-Value="_working.TestEntityId" Label="Test entity id (optional, e.g. sensor.house_power)" Class="mb-2" />
             <MudButton Variant="Variant.Outlined" StartIcon="@Icons.Material.Filled.NetworkCheck" OnClick="TestHaAsync" Disabled="_testing" Class="mb-2">
                 Test connection
@@ -130,7 +134,7 @@ else
             _working = new EditModel
             {
                 Id = endpoint.Id, Type = endpoint.Type, Name = endpoint.Name, IsEnabled = endpoint.IsEnabled,
-                BaseUrl = ha.BaseUrl, TokenEnv = ha.TokenEnv,
+                BaseUrl = ha.BaseUrl, TokenEnv = ha.TokenEnv, UseWebSocket = ha.UseWebSocket,
             };
         }
         else
@@ -170,7 +174,7 @@ else
         }
 
         var config = _working.Type == EndpointType.HomeAssistant
-            ? new HaEndpointConfig { BaseUrl = Trim(_working.BaseUrl), TokenEnv = Trim(_working.TokenEnv) }.ToJson()
+            ? new HaEndpointConfig { BaseUrl = Trim(_working.BaseUrl), TokenEnv = Trim(_working.TokenEnv), UseWebSocket = _working.UseWebSocket }.ToJson()
             : new EndpointConfig
             {
                 Host = string.IsNullOrWhiteSpace(_working.Host) ? "localhost" : _working.Host.Trim(),
@@ -234,6 +238,7 @@ else
         // Home Assistant
         public string? BaseUrl { get; set; }
         public string? TokenEnv { get; set; }
+        public bool UseWebSocket { get; set; }
         public string? TestEntityId { get; set; }
 
         // MQTT broker

src/Infrastructure/DependencyInjection.cs

@@ -57,6 +57,7 @@ public static class DependencyInjection
         services.AddHttpClient();
         services.AddHostedService<MqttIngestionWorker>();
         services.AddHostedService<HomeAssistantWorker>();
+        services.AddHostedService<HomeAssistantWebSocketWorker>();
         return services;
     }
 }

src/Infrastructure/Ingestion/HaEndpointConfig.cs

@@ -22,6 +22,13 @@ public sealed record HaEndpointConfig
     /// <summary>Name of the environment variable holding the long-lived access token.</summary>
     public string? TokenEnv { get; init; }
 
+    /// <summary>
+    /// When true, a persistent WebSocket subscription pushes state changes in real time
+    /// (<see cref="HomeAssistantWebSocketWorker"/>); when false (default) the REST poll worker
+    /// samples on each source's interval. An endpoint is handled by exactly one of the two.
+    /// </summary>
+    public bool UseWebSocket { get; init; }
+
     public static HaEndpointConfig Parse(string? json)
     {
         if (string.IsNullOrWhiteSpace(json))

src/Infrastructure/Ingestion/HaStateClient.cs

@@ -31,10 +31,16 @@ public sealed class HaStateClient(HttpClient httpClient)
         return ParseState(await JsonDocument.ParseAsync(stream, cancellationToken: cancellationToken).ConfigureAwait(false), attribute);
     }
 
-    internal static HaState? ParseState(JsonDocument document, string? attribute)
+    internal static HaState? ParseState(JsonDocument document, string? attribute) =>
-    {
+        ParseStateElement(document.RootElement, attribute);
-        var root = document.RootElement;
 
+    /// <summary>
+    /// Extracts a numeric value + last-updated time from a Home Assistant state object (the REST
+    /// entity payload, or a <c>new_state</c> from a WebSocket <c>state_changed</c> event — they share
+    /// the same shape). Returns null for <c>unavailable</c>/<c>unknown</c>/non-numeric states.
+    /// </summary>
+    internal static HaState? ParseStateElement(JsonElement root, string? attribute)
+    {
         double value;
         if (string.IsNullOrWhiteSpace(attribute))
         {

src/Infrastructure/Ingestion/HaWebSocketProtocol.cs

@@ -0,0 +1,70 @@
+using System.Text.Json;
+
+namespace MeterVault.Infrastructure.Ingestion;
+
+/// <summary>
+/// Pure helpers for the Home Assistant WebSocket API (<c>/api/websocket</c>): the auth handshake
+/// (<c>auth_required</c> → <c>auth</c> → <c>auth_ok</c>), subscribing to <c>state_changed</c> events,
+/// and reading the entity id + <c>new_state</c> out of an event frame. No I/O — the transport lives
+/// in <see cref="HomeAssistantWebSocketWorker"/>, which keeps this unit-testable.
+/// </summary>
+internal static class HaWebSocketProtocol
+{
+    /// <summary>Derives the WebSocket endpoint from an HTTP base URL (http→ws, https→wss, path <c>/api/websocket</c>).</summary>
+    public static Uri WebSocketUri(string baseUrl)
+    {
+        var http = new Uri(baseUrl.TrimEnd('/') + "/api/websocket", UriKind.Absolute);
+        var scheme = http.Scheme.Equals("https", StringComparison.OrdinalIgnoreCase) ? "wss" : "ws";
+        return new UriBuilder(http) { Scheme = scheme }.Uri;
+    }
+
+    public static string AuthMessage(string token) =>
+        JsonSerializer.Serialize(new { type = "auth", access_token = token });
+
+    public static string SubscribeStateChanged(int id) =>
+        JsonSerializer.Serialize(new { id, type = "subscribe_events", event_type = "state_changed" });
+
+    public static string? MessageType(JsonElement root) =>
+        root.TryGetProperty("type", out var type) && type.ValueKind == JsonValueKind.String ? type.GetString() : null;
+
+    public static bool IsAuthRequired(JsonElement root) => MessageType(root) == "auth_required";
+
+    public static bool IsAuthOk(JsonElement root) => MessageType(root) == "auth_ok";
+
+    public static bool IsAuthInvalid(JsonElement root) => MessageType(root) == "auth_invalid";
+
+    /// <summary>
+    /// If <paramref name="root"/> is a <c>state_changed</c> event with a numeric-capable
+    /// <c>new_state</c> object, yields the entity id and that state object. The caller extracts the
+    /// numeric value (state or a named attribute) per source via
+    /// <see cref="HaStateClient.ParseStateElement"/>.
+    /// </summary>
+    public static bool TryReadStateChanged(JsonElement root, out string entityId, out JsonElement newState)
+    {
+        entityId = string.Empty;
+        newState = default;
+
+        if (MessageType(root) != "event"
+            || !root.TryGetProperty("event", out var evt)
+            || !evt.TryGetProperty("event_type", out var evtType)
+            || evtType.GetString() != "state_changed"
+            || !evt.TryGetProperty("data", out var data))
+        {
+            return false;
+        }
+
+        if (!data.TryGetProperty("entity_id", out var id) || id.ValueKind != JsonValueKind.String)
+        {
+            return false;
+        }
+
+        if (!data.TryGetProperty("new_state", out var state) || state.ValueKind != JsonValueKind.Object)
+        {
+            return false; // entity removed (new_state null) — nothing to ingest.
+        }
+
+        entityId = id.GetString()!;
+        newState = state;
+        return true;
+    }
+}

src/Infrastructure/Ingestion/HomeAssistantWebSocketWorker.cs

@@ -0,0 +1,292 @@
+using System.Collections.Concurrent;
+using System.Net.WebSockets;
+using System.Text;
+using System.Text.Json;
+using MeterVault.Core.Domain;
+using MeterVault.Infrastructure.Persistence;
+using Microsoft.EntityFrameworkCore;
+using Microsoft.Extensions.DependencyInjection;
+using Microsoft.Extensions.Hosting;
+using Microsoft.Extensions.Logging;
+
+namespace MeterVault.Infrastructure.Ingestion;
+
+/// <summary>
+/// Real-time Home Assistant ingestion via the WebSocket API (SDD §6.2 push path). For each enabled
+/// HA endpoint whose config sets <c>UseWebSocket</c>, holds a persistent connection that authenticates,
+/// subscribes to <c>state_changed</c> events, and ingests changes for the endpoint's configured
+/// entities as they happen. Reconnects with capped backoff. Endpoints without <c>UseWebSocket</c> stay
+/// on the REST poll worker (<see cref="HomeAssistantWorker"/>) — each endpoint is served by exactly one.
+/// Tokens are resolved from environment variables named in the endpoint config, never stored plaintext.
+/// </summary>
+public sealed class HomeAssistantWebSocketWorker(
+    IServiceScopeFactory scopeFactory,
+    ILogger<HomeAssistantWebSocketWorker> logger) : BackgroundService
+{
+    private static readonly TimeSpan SuperviseInterval = TimeSpan.FromSeconds(15);
+    private static readonly TimeSpan InitialBackoff = TimeSpan.FromSeconds(2);
+    private static readonly TimeSpan MaxBackoff = TimeSpan.FromSeconds(60);
+    private static readonly TimeSpan EntityMapTtl = TimeSpan.FromSeconds(30);
+
+    private readonly IServiceScopeFactory _scopeFactory = scopeFactory;
+    private readonly ILogger<HomeAssistantWebSocketWorker> _logger = logger;
+    private readonly ConcurrentDictionary<int, Task> _connections = new();
+
+    protected override async Task ExecuteAsync(CancellationToken stoppingToken)
+    {
+        using var timer = new PeriodicTimer(SuperviseInterval);
+        do
+        {
+            try
+            {
+                await SuperviseAsync(stoppingToken).ConfigureAwait(false);
+            }
+            catch (OperationCanceledException) when (stoppingToken.IsCancellationRequested)
+            {
+                break;
+            }
+            catch (Exception ex)
+            {
+                _logger.LogWarning(ex, "Home Assistant WebSocket supervisor tick failed; will retry");
+            }
+        }
+        while (await timer.WaitForNextTickAsync(stoppingToken).ConfigureAwait(false));
+    }
+
+    /// <summary>Starts a connection loop for each WebSocket-enabled endpoint that isn't already running.</summary>
+    private async Task SuperviseAsync(CancellationToken stoppingToken)
+    {
+        List<int> wsEndpointIds;
+        await using (var scope = _scopeFactory.CreateAsyncScope())
+        {
+            var db = scope.ServiceProvider.GetRequiredService<MeterVaultDbContext>();
+            var endpoints = await db.IngestionEndpoints
+                .Where(e => e.IsEnabled && e.Type == EndpointType.HomeAssistant)
+                .ToListAsync(stoppingToken).ConfigureAwait(false);
+            wsEndpointIds = endpoints
+                .Where(e => HaEndpointConfig.Parse(e.Config).UseWebSocket)
+                .Select(e => e.Id)
+                .ToList();
+        }
+
+        // Reap finished connection loops (endpoint disabled/removed, or terminal failure).
+        foreach (var id in _connections.Where(kv => kv.Value.IsCompleted).Select(kv => kv.Key).ToList())
+        {
+            _connections.TryRemove(id, out _);
+        }
+
+        foreach (var endpointId in wsEndpointIds)
+        {
+            // Fire-and-forget: the loop stores its own Task in _connections and self-terminates when
+            // the endpoint is disabled/removed; the supervisor reaps completed entries above.
+            _ = _connections.GetOrAdd(endpointId, id => Task.Run(() => RunConnectionAsync(id, stoppingToken), stoppingToken));
+        }
+    }
+
+    /// <summary>Connect → listen → reconnect loop for one endpoint, until it is disabled or the app stops.</summary>
+    private async Task RunConnectionAsync(int endpointId, CancellationToken stoppingToken)
+    {
+        var backoff = InitialBackoff;
+        while (!stoppingToken.IsCancellationRequested)
+        {
+            HaEndpointConfig config;
+            string? token;
+            await using (var scope = _scopeFactory.CreateAsyncScope())
+            {
+                var db = scope.ServiceProvider.GetRequiredService<MeterVaultDbContext>();
+                var endpoint = await db.IngestionEndpoints
+                    .FirstOrDefaultAsync(e => e.Id == endpointId, stoppingToken).ConfigureAwait(false);
+                if (endpoint is null || !endpoint.IsEnabled)
+                {
+                    return; // gone/disabled — stop; the supervisor will restart it if it comes back.
+                }
+
+                config = HaEndpointConfig.Parse(endpoint.Config);
+                if (!config.UseWebSocket)
+                {
+                    return; // switched to poll mode.
+                }
+
+                token = config.ResolveToken();
+            }
+
+            if (string.IsNullOrWhiteSpace(config.BaseUrl) || string.IsNullOrWhiteSpace(token))
+            {
+                _logger.LogWarning("HA WebSocket endpoint {EndpointId} missing base URL or token; retrying", endpointId);
+            }
+            else
+            {
+                try
+                {
+                    await ListenAsync(endpointId, config.BaseUrl!, token!, stoppingToken).ConfigureAwait(false);
+                    backoff = InitialBackoff; // clean close → reset backoff.
+                }
+                catch (OperationCanceledException) when (stoppingToken.IsCancellationRequested)
+                {
+                    return;
+                }
+                catch (Exception ex)
+                {
+                    _logger.LogWarning(ex, "HA WebSocket connection to endpoint {EndpointId} dropped; reconnecting", endpointId);
+                    await UpdateEndpointStatusAsync(endpointId, "disconnected", CancellationToken.None).ConfigureAwait(false);
+                }
+            }
+
+            try
+            {
+                await Task.Delay(backoff, stoppingToken).ConfigureAwait(false);
+            }
+            catch (OperationCanceledException) when (stoppingToken.IsCancellationRequested)
+            {
+                return;
+            }
+
+            backoff = TimeSpan.FromSeconds(Math.Min(MaxBackoff.TotalSeconds, backoff.TotalSeconds * 2));
+        }
+    }
+
+    private async Task ListenAsync(int endpointId, string baseUrl, string token, CancellationToken stoppingToken)
+    {
+        using var ws = new ClientWebSocket();
+        await ws.ConnectAsync(HaWebSocketProtocol.WebSocketUri(baseUrl), stoppingToken).ConfigureAwait(false);
+
+        // Handshake: auth_required → auth → auth_ok.
+        using (var required = await ReceiveJsonAsync(ws, stoppingToken).ConfigureAwait(false))
+        {
+            if (!HaWebSocketProtocol.IsAuthRequired(required.RootElement))
+            {
+                // Some setups may not send auth_required first; proceed to auth regardless.
+                _logger.LogDebug("HA WebSocket did not send auth_required first (type={Type})", HaWebSocketProtocol.MessageType(required.RootElement));
+            }
+        }
+
+        await SendAsync(ws, HaWebSocketProtocol.AuthMessage(token), stoppingToken).ConfigureAwait(false);
+        using (var authResult = await ReceiveJsonAsync(ws, stoppingToken).ConfigureAwait(false))
+        {
+            if (!HaWebSocketProtocol.IsAuthOk(authResult.RootElement))
+            {
+                throw new InvalidOperationException(
+                    $"Home Assistant WebSocket auth failed (type={HaWebSocketProtocol.MessageType(authResult.RootElement)}).");
+            }
+        }
+
+        await SendAsync(ws, HaWebSocketProtocol.SubscribeStateChanged(1), stoppingToken).ConfigureAwait(false);
+        await UpdateEndpointStatusAsync(endpointId, "connected (ws)", stoppingToken).ConfigureAwait(false);
+        _logger.LogInformation("HA WebSocket connected for endpoint {EndpointId}", endpointId);
+
+        var entityMap = await LoadEntityMapAsync(endpointId, stoppingToken).ConfigureAwait(false);
+        var mapLoadedAt = DateTimeOffset.UtcNow;
+
+        while (!stoppingToken.IsCancellationRequested && ws.State == WebSocketState.Open)
+        {
+            using var doc = await ReceiveJsonAsync(ws, stoppingToken).ConfigureAwait(false);
+            if (!HaWebSocketProtocol.TryReadStateChanged(doc.RootElement, out var entityId, out var newState))
+            {
+                continue;
+            }
+
+            if (DateTimeOffset.UtcNow - mapLoadedAt > EntityMapTtl)
+            {
+                entityMap = await LoadEntityMapAsync(endpointId, stoppingToken).ConfigureAwait(false);
+                mapLoadedAt = DateTimeOffset.UtcNow;
+            }
+
+            if (!entityMap.TryGetValue(entityId, out var sources))
+            {
+                continue;
+            }
+
+            foreach (var (sourceId, attribute) in sources)
+            {
+                if (HaStateClient.ParseStateElement(newState, attribute) is { } state)
+                {
+                    await IngestAsync(sourceId, state, stoppingToken).ConfigureAwait(false);
+                }
+            }
+        }
+    }
+
+    /// <summary>Maps each configured entity id to the sources (and optional attribute) that read it.</summary>
+    private async Task<Dictionary<string, List<(int SourceId, string? Attribute)>>> LoadEntityMapAsync(
+        int endpointId, CancellationToken cancellationToken)
+    {
+        await using var scope = _scopeFactory.CreateAsyncScope();
+        var db = scope.ServiceProvider.GetRequiredService<MeterVaultDbContext>();
+        var sources = await db.MeterSources.AsNoTracking()
+            .Where(s => s.IsEnabled && s.SourceType == SourceType.HomeAssistant && s.EndpointId == endpointId)
+            .ToListAsync(cancellationToken).ConfigureAwait(false);
+
+        var map = new Dictionary<string, List<(int, string?)>>(StringComparer.OrdinalIgnoreCase);
+        foreach (var source in sources)
+        {
+            var config = SourceConfig.Parse(source.Config);
+            if (string.IsNullOrWhiteSpace(config.EntityId))
+            {
+                continue;
+            }
+
+            if (!map.TryGetValue(config.EntityId, out var list))
+            {
+                map[config.EntityId] = list = [];
+            }
+
+            list.Add((source.Id, config.Attribute));
+        }
+
+        return map;
+    }
+
+    private async Task IngestAsync(int sourceId, HaState state, CancellationToken cancellationToken)
+    {
+        await using var scope = _scopeFactory.CreateAsyncScope();
+        var ingestion = scope.ServiceProvider.GetRequiredService<IngestionService>();
+        await ingestion.IngestAsync(sourceId, state.Time, state.Value, cancellationToken).ConfigureAwait(false);
+    }
+
+    private async Task UpdateEndpointStatusAsync(int endpointId, string status, CancellationToken cancellationToken)
+    {
+        try
+        {
+            await using var scope = _scopeFactory.CreateAsyncScope();
+            var db = scope.ServiceProvider.GetRequiredService<MeterVaultDbContext>();
+            var endpoint = await db.IngestionEndpoints
+                .FirstOrDefaultAsync(e => e.Id == endpointId, cancellationToken).ConfigureAwait(false);
+            if (endpoint is not null)
+            {
+                endpoint.LastStatus = status;
+                endpoint.LastSeenAt = DateTimeOffset.UtcNow;
+                await db.SaveChangesAsync(cancellationToken).ConfigureAwait(false);
+            }
+        }
+        catch (Exception ex)
+        {
+            _logger.LogDebug(ex, "Failed to update HA endpoint {EndpointId} status", endpointId);
+        }
+    }
+
+    private static Task SendAsync(ClientWebSocket ws, string json, CancellationToken cancellationToken) =>
+        ws.SendAsync(Encoding.UTF8.GetBytes(json), WebSocketMessageType.Text, endOfMessage: true, cancellationToken);
+
+    /// <summary>Reads one (possibly fragmented) text message and parses it as JSON.</summary>
+    private static async Task<JsonDocument> ReceiveJsonAsync(ClientWebSocket ws, CancellationToken cancellationToken)
+    {
+        using var buffer = new MemoryStream();
+        var chunk = new byte[8192];
+        WebSocketReceiveResult result;
+        do
+        {
+            result = await ws.ReceiveAsync(chunk, cancellationToken).ConfigureAwait(false);
+            if (result.MessageType == WebSocketMessageType.Close)
+            {
+                throw new InvalidOperationException(
+                    $"Home Assistant closed the WebSocket ({result.CloseStatus}: {result.CloseStatusDescription}).");
+            }
+
+            buffer.Write(chunk, 0, result.Count);
+        }
+        while (!result.EndOfMessage);
+
+        buffer.Position = 0;
+        return await JsonDocument.ParseAsync(buffer, cancellationToken: cancellationToken).ConfigureAwait(false);
+    }
+}

src/Infrastructure/Ingestion/HomeAssistantWorker.cs

@@ -54,9 +54,14 @@ public sealed class HomeAssistantWorker(
         var ingestion = scope.ServiceProvider.GetRequiredService<IngestionService>();
         var client = new HaStateClient(_httpClientFactory.CreateClient());
 
-        var endpoints = await db.IngestionEndpoints
+        // Endpoints using the WebSocket push path are served by HomeAssistantWebSocketWorker; skip
+        // them here so a source is never both polled and pushed.
+        var enabled = await db.IngestionEndpoints
             .Where(e => e.IsEnabled && e.Type == EndpointType.HomeAssistant)
-            .ToDictionaryAsync(e => e.Id, cancellationToken).ConfigureAwait(false);
+            .ToListAsync(cancellationToken).ConfigureAwait(false);
+        var endpoints = enabled
+            .Where(e => !HaEndpointConfig.Parse(e.Config).UseWebSocket)
+            .ToDictionary(e => e.Id);
         if (endpoints.Count == 0)
         {
             return;

tests/Integration.Tests/Ingestion/HaWebSocketProtocolTests.cs

@@ -0,0 +1,119 @@
+using System.Text.Json;
+using MeterVault.Infrastructure.Ingestion;
+
+namespace MeterVault.Integration.Tests.Ingestion;
+
+/// <summary>
+/// The pure Home Assistant WebSocket protocol helpers: URL derivation, the auth/subscribe frames,
+/// and reading the entity id + numeric value out of a <c>state_changed</c> event (the transport is
+/// exercised separately). No network — these are the parts that must be provably correct.
+/// </summary>
+public sealed class HaWebSocketProtocolTests
+{
+    [Theory]
+    [InlineData("http://ha.local:8123", "ws", "ha.local", 8123)]
+    [InlineData("http://ha.local:8123/", "ws", "ha.local", 8123)]
+    [InlineData("https://ha.example.com", "wss", "ha.example.com", 443)]
+    public void WebSocketUri_maps_scheme_and_appends_api_path(string baseUrl, string scheme, string host, int port)
+    {
+        var uri = HaWebSocketProtocol.WebSocketUri(baseUrl);
+
+        Assert.Equal(scheme, uri.Scheme);
+        Assert.Equal(host, uri.Host);
+        Assert.Equal(port, uri.Port);
+        Assert.Equal("/api/websocket", uri.AbsolutePath);
+    }
+
+    [Fact]
+    public void AuthMessage_carries_type_and_token()
+    {
+        using var doc = JsonDocument.Parse(HaWebSocketProtocol.AuthMessage("secret-token"));
+
+        Assert.Equal("auth", doc.RootElement.GetProperty("type").GetString());
+        Assert.Equal("secret-token", doc.RootElement.GetProperty("access_token").GetString());
+    }
+
+    [Fact]
+    public void SubscribeStateChanged_requests_state_changed_events()
+    {
+        using var doc = JsonDocument.Parse(HaWebSocketProtocol.SubscribeStateChanged(7));
+
+        Assert.Equal(7, doc.RootElement.GetProperty("id").GetInt32());
+        Assert.Equal("subscribe_events", doc.RootElement.GetProperty("type").GetString());
+        Assert.Equal("state_changed", doc.RootElement.GetProperty("event_type").GetString());
+    }
+
+    [Theory]
+    [InlineData("{\"type\":\"auth_required\",\"ha_version\":\"2024.6\"}", true, false, false)]
+    [InlineData("{\"type\":\"auth_ok\"}", false, true, false)]
+    [InlineData("{\"type\":\"auth_invalid\",\"message\":\"bad token\"}", false, false, true)]
+    public void Auth_message_types_are_recognized(string json, bool required, bool ok, bool invalid)
+    {
+        using var doc = JsonDocument.Parse(json);
+        var root = doc.RootElement;
+
+        Assert.Equal(required, HaWebSocketProtocol.IsAuthRequired(root));
+        Assert.Equal(ok, HaWebSocketProtocol.IsAuthOk(root));
+        Assert.Equal(invalid, HaWebSocketProtocol.IsAuthInvalid(root));
+    }
+
+    [Fact]
+    public void TryReadStateChanged_extracts_entity_and_numeric_state()
+    {
+        using var doc = JsonDocument.Parse(StateChangedEvent(state: "1234.5"));
+
+        Assert.True(HaWebSocketProtocol.TryReadStateChanged(doc.RootElement, out var entityId, out var newState));
+        Assert.Equal("sensor.house_power", entityId);
+
+        var parsed = HaStateClient.ParseStateElement(newState, attribute: null);
+        Assert.NotNull(parsed);
+        Assert.Equal(1234.5, parsed!.Value.Value, 3);
+        Assert.Equal(new DateTimeOffset(2024, 6, 15, 10, 0, 0, TimeSpan.Zero), parsed.Value.Time);
+    }
+
+    [Fact]
+    public void TryReadStateChanged_reads_a_named_attribute()
+    {
+        using var doc = JsonDocument.Parse(StateChangedEvent(state: "on"));
+
+        Assert.True(HaWebSocketProtocol.TryReadStateChanged(doc.RootElement, out _, out var newState));
+        // state "on" is non-numeric, but the 'current' attribute is a number.
+        Assert.Null(HaStateClient.ParseStateElement(newState, attribute: null));
+        var byAttribute = HaStateClient.ParseStateElement(newState, attribute: "current");
+        Assert.Equal(42, byAttribute!.Value.Value, 3);
+    }
+
+    [Fact]
+    public void TryReadStateChanged_ignores_removed_entities_and_other_events()
+    {
+        using var removed = JsonDocument.Parse(
+            "{\"type\":\"event\",\"event\":{\"event_type\":\"state_changed\",\"data\":{\"entity_id\":\"sensor.x\",\"new_state\":null}}}");
+        Assert.False(HaWebSocketProtocol.TryReadStateChanged(removed.RootElement, out _, out _));
+
+        using var other = JsonDocument.Parse(
+            "{\"type\":\"event\",\"event\":{\"event_type\":\"call_service\",\"data\":{}}}");
+        Assert.False(HaWebSocketProtocol.TryReadStateChanged(other.RootElement, out _, out _));
+
+        using var result = JsonDocument.Parse("{\"id\":1,\"type\":\"result\",\"success\":true}");
+        Assert.False(HaWebSocketProtocol.TryReadStateChanged(result.RootElement, out _, out _));
+    }
+
+    private static string StateChangedEvent(string state) => $$"""
+        {
+          "id": 1,
+          "type": "event",
+          "event": {
+            "event_type": "state_changed",
+            "data": {
+              "entity_id": "sensor.house_power",
+              "new_state": {
+                "entity_id": "sensor.house_power",
+                "state": "{{state}}",
+                "attributes": { "unit_of_measurement": "W", "current": 42 },
+                "last_updated": "2024-06-15T10:00:00+00:00"
+              }
+            }
+          }
+        }
+        """;
+}

tests/Integration.Tests/Ingestion/HomeAssistantWebSocketWorkerTests.cs

@@ -0,0 +1,182 @@
+using System.Net.WebSockets;
+using System.Text;
+using System.Text.Json;
+using MeterVault.Core.Domain;
+using MeterVault.Infrastructure.Ingestion;
+using MeterVault.Infrastructure.Persistence;
+using Microsoft.AspNetCore.Builder;
+using Microsoft.AspNetCore.Hosting;
+using Microsoft.AspNetCore.Hosting.Server;
+using Microsoft.AspNetCore.Hosting.Server.Features;
+using Microsoft.AspNetCore.Http;
+using Microsoft.EntityFrameworkCore;
+using Microsoft.Extensions.DependencyInjection;
+using Microsoft.Extensions.Logging;
+using Microsoft.Extensions.Logging.Abstractions;
+
+namespace MeterVault.Integration.Tests.Ingestion;
+
+/// <summary>
+/// End-to-end proof of the HA WebSocket push path against an in-process fake Home Assistant server:
+/// the worker performs the auth handshake, subscribes to <c>state_changed</c>, and a pushed change
+/// for a configured entity lands as a <c>reading</c>. This exercises the real ClientWebSocket
+/// transport and handshake sequencing that the pure-protocol unit tests can't.
+/// </summary>
+[Collection("Timescale")]
+public sealed class HomeAssistantWebSocketWorkerTests(TimescaleFixture fx)
+{
+    private const string TokenEnvVar = "MV_TEST_HA_WS_TOKEN";
+
+    [Fact]
+    public async Task Pushed_state_change_becomes_a_reading()
+    {
+        await using var db = fx.CreateContext();
+        Environment.SetEnvironmentVariable(TokenEnvVar, "test-token");
+        await using var fake = await FakeHaServer.StartAsync(entityId: "sensor.house_power", state: "4711");
+
+        try
+        {
+            var type = new EnergyType { Key = "ha_ws_test", DisplayName = "HA WS", BaseUnit = "kWh", DefaultMode = MeterMode.CumulativeCounter };
+            db.EnergyTypes.Add(type);
+            await db.SaveChangesAsync();
+
+            var meter = new Meter { Name = "HA WS Meter", EnergyTypeId = type.Id, Mode = MeterMode.CumulativeCounter, Unit = "kWh" };
+            db.Meters.Add(meter);
+            await db.SaveChangesAsync();
+
+            var endpoint = new IngestionEndpoint
+            {
+                Type = EndpointType.HomeAssistant,
+                Name = "Fake HA",
+                IsEnabled = true,
+                Config = new HaEndpointConfig { BaseUrl = fake.BaseUrl, TokenEnv = TokenEnvVar, UseWebSocket = true }.ToJson(),
+            };
+            db.IngestionEndpoints.Add(endpoint);
+            await db.SaveChangesAsync();
+
+            db.MeterSources.Add(new MeterSource
+            {
+                MeterId = meter.Id,
+                SourceType = SourceType.HomeAssistant,
+                EndpointId = endpoint.Id,
+                IsEnabled = true,
+                Config = JsonSerializer.Serialize(new { entityId = "sensor.house_power" }),
+            });
+            await db.SaveChangesAsync();
+
+            await using var provider = BuildProvider(fx.ConnectionString);
+            var worker = new HomeAssistantWebSocketWorker(
+                provider.GetRequiredService<IServiceScopeFactory>(), NullLogger<HomeAssistantWebSocketWorker>.Instance);
+
+            await worker.StartAsync(CancellationToken.None);
+            try
+            {
+                Reading? reading = null;
+                for (var i = 0; i < 60 && reading is null; i++)
+                {
+                    await Task.Delay(200);
+                    reading = await db.Readings.AsNoTracking().FirstOrDefaultAsync(r => r.MeterId == meter.Id);
+                }
+
+                Assert.NotNull(reading);
+                Assert.Equal(4711, reading!.Value, 3);
+                Assert.Equal(ReadingQuality.Measured, reading.Quality);
+            }
+            finally
+            {
+                await worker.StopAsync(CancellationToken.None);
+            }
+        }
+        finally
+        {
+            Environment.SetEnvironmentVariable(TokenEnvVar, null);
+            await db.Readings.ExecuteDeleteAsync();
+            await db.MeterSources.ExecuteDeleteAsync();
+            await db.IngestionEndpoints.ExecuteDeleteAsync();
+            await db.Meters.ExecuteDeleteAsync();
+            await db.EnergyTypes.Where(t => t.Key == "ha_ws_test").ExecuteDeleteAsync();
+        }
+    }
+
+    private static ServiceProvider BuildProvider(string connectionString)
+    {
+        var services = new ServiceCollection();
+        services.AddLogging();
+        services.AddDbContextFactory<MeterVaultDbContext>(o => o
+            .UseNpgsql(connectionString, n => n.MigrationsAssembly(typeof(MeterVaultDbContext).Assembly.FullName))
+            .UseSnakeCaseNamingConvention());
+        services.AddScoped<MeterVaultDbContext>(sp => sp.GetRequiredService<IDbContextFactory<MeterVaultDbContext>>().CreateDbContext());
+        services.AddScoped<IngestionService>();
+        return services.BuildServiceProvider();
+    }
+
+    /// <summary>A minimal Home Assistant WebSocket server: handshake, then push one state_changed event.</summary>
+    private sealed class FakeHaServer : IAsyncDisposable
+    {
+        private readonly WebApplication _app;
+
+        private FakeHaServer(WebApplication app, string baseUrl)
+        {
+            _app = app;
+            BaseUrl = baseUrl;
+        }
+
+        public string BaseUrl { get; }
+
+        public static async Task<FakeHaServer> StartAsync(string entityId, string state)
+        {
+            var builder = WebApplication.CreateBuilder();
+            builder.Logging.ClearProviders();
+            builder.WebHost.UseUrls("http://127.0.0.1:0");
+            var app = builder.Build();
+            app.UseWebSockets();
+
+            var eventJson =
+                "{\"id\":1,\"type\":\"event\",\"event\":{\"event_type\":\"state_changed\",\"data\":{\"entity_id\":\""
+                + entityId + "\",\"new_state\":{\"entity_id\":\"" + entityId + "\",\"state\":\"" + state
+                + "\",\"attributes\":{},\"last_updated\":\"2024-06-15T10:00:00+00:00\"}}}}";
+
+            app.Map("/api/websocket", async context =>
+            {
+                if (!context.WebSockets.IsWebSocketRequest)
+                {
+                    context.Response.StatusCode = 400;
+                    return;
+                }
+
+                using var ws = await context.WebSockets.AcceptWebSocketAsync();
+                await SendAsync(ws, """{"type":"auth_required","ha_version":"2024.6"}""");
+                await ReceiveAsync(ws); // client "auth"
+                await SendAsync(ws, """{"type":"auth_ok"}""");
+                await ReceiveAsync(ws); // client "subscribe_events"
+                await SendAsync(ws, """{"id":1,"type":"result","success":true}""");
+                await SendAsync(ws, eventJson);
+
+                try
+                {
+                    await Task.Delay(Timeout.Infinite, context.RequestAborted);
+                }
+                catch (OperationCanceledException)
+                {
+                    // client/test closed — expected.
+                }
+            });
+
+            await app.StartAsync();
+            var address = app.Services.GetRequiredService<IServer>().Features
+                .Get<IServerAddressesFeature>()!.Addresses.First();
+            return new FakeHaServer(app, address);
+        }
+
+        public async ValueTask DisposeAsync() => await _app.DisposeAsync();
+
+        private static Task SendAsync(WebSocket ws, string json) =>
+            ws.SendAsync(Encoding.UTF8.GetBytes(json), WebSocketMessageType.Text, endOfMessage: true, CancellationToken.None);
+
+        private static async Task ReceiveAsync(WebSocket ws)
+        {
+            var buffer = new byte[8192];
+            await ws.ReceiveAsync(buffer, CancellationToken.None);
+        }
+    }
+}