feat: adds hot poll for in-flight CI and workflow status (#30)

* feat(api): adds nodeId to PullRequest and hot poll query functions

* feat(config): adds hotPollInterval setting with 10-120s range

* feat(poll): adds hot poll coordinator with hot item set management

* feat(dashboard): integrates hot poll coordinator with store updates

* test(hot-poll): adds hot poll and config validation tests

* fix(hot-poll): addresses review findings and adds coverage

* fix(poll): moves hot set cap check after qualification filter

* fix(poll): guards eviction against stale generation

* fix(hot-poll): logs batch and cycle errors instead of swallowing

* fix(hot-poll): adds failure backoff and missing test coverage

* chore(hot-poll): documents backoff design tradeoff

* fix(hot-poll): wires backoff to actual errors, clears hot sets on unmount

* fix(hot-poll): wires hadErrors through fetchHotPRStatus to backoff

* fix(dashboard): copies nodeId in phaseOne enrichment merge

* fix(hot-poll): addresses all PR review findings

* fix(hot-poll): addresses domain review findings from quality gate

* fix(hot-poll): surfaces hadErrors via pushError for user visibility

* fix(hot-poll): auto-dismisses error notification on recovery

Author: wgordon17

src/app/components/dashboard/DashboardPage.tsx

@@ -10,12 +10,23 @@ import { config, setConfig } from "../../stores/config";
 import { viewState, updateViewState } from "../../stores/view";
 import type { Issue, PullRequest, WorkflowRun } from "../../services/api";
 import { fetchOrgs } from "../../services/api";
-import { createPollCoordinator, fetchAllData, type DashboardData } from "../../services/poll";
+import {
+  createPollCoordinator,
+  createHotPollCoordinator,
+  rebuildHotSets,
+  clearHotSets,
+  getHotPollGeneration,
+  fetchAllData,
+  type DashboardData,
+} from "../../services/poll";
 import { clearAuth, user, onAuthCleared, DASHBOARD_STORAGE_KEY } from "../../stores/auth";
 import { getClient, getGraphqlRateLimit } from "../../services/github";
 
 // ── Shared dashboard store (module-level to survive navigation) ─────────────
 
+// Bump only for breaking schema changes (renames, type changes). Additive optional
+// fields (e.g., nodeId?: string) don't require a bump — missing fields deserialize
+// as undefined, which consuming code handles gracefully.
 const CACHE_VERSION = 2;
 
 interface DashboardStore {
@@ -72,6 +83,11 @@ onAuthCleared(() => {
     coord.destroy();
     _setCoordinator(null);
   }
+  const hotCoord = _hotCoordinator();
+  if (hotCoord) {
+    hotCoord.destroy();
+    _setHotCoordinator(null);
+  }
 });
 
 async function pollFetch(): Promise<DashboardData> {
@@ -140,6 +156,7 @@ async function pollFetch(): Promise<DashboardData> {
               pr.reviewThreads = e.reviewThreads;
               pr.totalReviewCount = e.totalReviewCount;
               pr.enriched = e.enriched;
+              pr.nodeId = e.nodeId;
             }
           } else {
             state.pullRequests = data.pullRequests;
@@ -157,6 +174,7 @@ async function pollFetch(): Promise<DashboardData> {
           lastRefreshedAt: now,
         });
       }
+      rebuildHotSets(data);
       // Persist for stale-while-revalidate on full page reload.
       // Errors are transient and not persisted. Deferred to avoid blocking paint.
       const cachePayload = {
@@ -198,6 +216,7 @@ async function pollFetch(): Promise<DashboardData> {
 }
 
 const [_coordinator, _setCoordinator] = createSignal<ReturnType<typeof createPollCoordinator> | null>(null);
+const [_hotCoordinator, _setHotCoordinator] = createSignal<{ destroy: () => void } | null>(null);
 
 export default function DashboardPage() {
 
@@ -220,6 +239,38 @@ export default function DashboardPage() {
       _setCoordinator(createPollCoordinator(() => config.refreshInterval, pollFetch));
     }
 
+    if (!_hotCoordinator()) {
+      _setHotCoordinator(createHotPollCoordinator(
+        () => config.hotPollInterval,
+        (prUpdates, runUpdates, fetchGeneration) => {
+          // Guard against stale hot poll results overlapping with a full refresh.
+          // fetchGeneration was captured BEFORE fetchHotData() started its async work.
+          // If a full refresh completed during the fetch, _hotPollGeneration will have
+          // been incremented by rebuildHotSets(), and fetchGeneration will be stale.
+          if (fetchGeneration !== getHotPollGeneration()) return; // stale, discard
+          setDashboardData(produce((state) => {
+            // Apply PR status updates
+            for (const pr of state.pullRequests) {
+              const update = prUpdates.get(pr.id);
+              if (!update) continue;
+              pr.state = update.state; // detect closed/merged quickly
+              pr.checkStatus = update.checkStatus;
+              pr.reviewDecision = update.reviewDecision;
+            }
+            // Apply workflow run updates
+            for (const run of state.workflowRuns) {
+              const update = runUpdates.get(run.id);
+              if (!update) continue;
+              run.status = update.status;
+              run.conclusion = update.conclusion;
+              run.updatedAt = update.updatedAt;
+              run.completedAt = update.completedAt;
+            }
+          }));
+        }
+      ));
+    }
+
     // Auto-sync orgs on dashboard load — picks up newly accessible orgs
     // after re-auth, scope changes, or org policy updates.
     // Only adds orgs to the filter list — repos are user-selected via Settings.
@@ -242,6 +293,9 @@ export default function DashboardPage() {
     onCleanup(() => {
       _coordinator()?.destroy();
       _setCoordinator(null);
+      _hotCoordinator()?.destroy();
+      _setHotCoordinator(null);
+      clearHotSets();
     });
   });
 

src/app/components/settings/SettingsPage.tsx

@@ -135,6 +135,7 @@ export default function SettingsPage() {
         selectedOrgs: config.selectedOrgs,
         selectedRepos: config.selectedRepos,
         refreshInterval: config.refreshInterval,
+        hotPollInterval: config.hotPollInterval,
         maxWorkflowsPerRepo: config.maxWorkflowsPerRepo,
         maxRunsPerWorkflow: config.maxRunsPerWorkflow,
         notifications: config.notifications,
@@ -332,6 +333,24 @@ export default function SettingsPage() {
               ))}
             </select>
           </SettingRow>
+          <SettingRow
+            label="CI status refresh"
+            description="How often to re-check in-flight CI checks and workflow runs (10-120s)"
+          >
+            <input
+              type="number"
+              min={10}
+              max={120}
+              value={config.hotPollInterval}
+              onInput={(e) => {
+                const val = parseInt(e.currentTarget.value, 10);
+                if (!isNaN(val) && val >= 10 && val <= 120) {
+                  saveWithFeedback({ hotPollInterval: val });
+                }
+              }}
+              class="input input-sm w-20"
+            />
+          </SettingRow>
         </Section>
 
         {/* Section 3: GitHub Actions */}

src/app/services/api.ts

@@ -1,4 +1,4 @@
-import { getClient, cachedRequest, updateGraphqlRateLimit } from "./github";
+import { getClient, cachedRequest, updateGraphqlRateLimit, updateRateLimitFromHeaders } from "./github";
 import { pushNotification } from "../lib/errors";
 
 // ── Types ────────────────────────────────────────────────────────────────────
@@ -67,6 +67,8 @@ export interface PullRequest {
   totalReviewCount: number;
   /** False when only light fields are loaded (phase 1); true/undefined when fully enriched */
   enriched?: boolean;
+  /** GraphQL global node ID — used for hot-poll status updates */
+  nodeId?: string;
 }
 
 export interface WorkflowRun {
@@ -212,7 +214,7 @@ type GitHubOctokit = NonNullable<ReturnType<typeof getClient>>;
  * Unlike chunked Promise.allSettled, tasks start immediately as slots free up
  * rather than waiting for an entire chunk to finish.
  */
-async function pooledAllSettled<T>(
+export async function pooledAllSettled<T>(
   tasks: (() => Promise<T>)[],
   concurrency: number
 ): Promise<PromiseSettledResult<T>[]> {
@@ -543,6 +545,41 @@ const HEAVY_PR_BACKFILL_QUERY = `
   }
 `;
 
+/** Hot-poll query: fetches current status fields for a batch of PR node IDs. */
+const HOT_PR_STATUS_QUERY = `
+  query($ids: [ID!]!) {
+    nodes(ids: $ids) {
+      ... on PullRequest {
+        databaseId
+        state
+        mergeStateStatus
+        reviewDecision
+        commits(last: 1) {
+          nodes {
+            commit {
+              statusCheckRollup { state }
+            }
+          }
+        }
+      }
+    }
+    rateLimit { remaining resetAt }
+  }
+`;
+
+interface HotPRStatusNode {
+  databaseId: number;
+  state: string;
+  mergeStateStatus: string;
+  reviewDecision: string | null;
+  commits: { nodes: { commit: { statusCheckRollup: { state: string } | null } }[] };
+}
+
+interface HotPRStatusResponse {
+  nodes: (HotPRStatusNode | null)[];
+  rateLimit?: { remaining: number; resetAt: string };
+}
+
 interface GraphQLLightPRNode {
   id: string; // GraphQL global node ID
   databaseId: number;
@@ -870,6 +907,7 @@ function processLightPRNode(
     reviewDecision: mapReviewDecision(node.reviewDecision),
     totalReviewCount: 0,
     enriched: false,
+    nodeId: node.id,
   });
   return true;
 }
@@ -1715,3 +1753,94 @@ export async function fetchWorkflowRuns(
 
   return { workflowRuns: allRuns, errors: allErrors };
 }
+
+// ── Hot poll: targeted status updates ────────────────────────────────────────
+
+export interface HotPRStatusUpdate {
+  state: string;
+  checkStatus: CheckStatus["status"];
+  mergeStateStatus: string;
+  reviewDecision: PullRequest["reviewDecision"];
+}
+
+/**
+ * Fetches current status fields (check status, review decision, state) for a
+ * batch of PR node IDs using the nodes() GraphQL query. Returns a map keyed
+ * by databaseId. Uses Promise.allSettled per batch for error resilience.
+ */
+export async function fetchHotPRStatus(
+  octokit: GitHubOctokit,
+  nodeIds: string[]
+): Promise<{ results: Map<number, HotPRStatusUpdate>; hadErrors: boolean }> {
+  const results = new Map<number, HotPRStatusUpdate>();
+  if (nodeIds.length === 0) return { results, hadErrors: false };
+
+  const batches = chunkArray(nodeIds, NODES_BATCH_SIZE);
+  let hadErrors = false;
+  const settled = await Promise.allSettled(batches.map(async (batch) => {
+    const response = await octokit.graphql<HotPRStatusResponse>(HOT_PR_STATUS_QUERY, { ids: batch });
+    if (response.rateLimit) updateGraphqlRateLimit(response.rateLimit);
+
+    for (const node of response.nodes) {
+      if (!node || node.databaseId == null) continue;
+
+      let checkStatus = mapCheckStatus(node.commits.nodes[0]?.commit?.statusCheckRollup?.state ?? null);
+      const mss = node.mergeStateStatus;
+      if (mss === "DIRTY" || mss === "BEHIND") {
+        checkStatus = "conflict";
+      } else if (mss === "UNSTABLE") {
+        checkStatus = "failure";
+      }
+
+      results.set(node.databaseId, {
+        state: node.state,
+        checkStatus,
+        mergeStateStatus: node.mergeStateStatus,
+        reviewDecision: mapReviewDecision(node.reviewDecision),
+      });
+    }
+  }));
+
+  for (const s of settled) {
+    if (s.status === "rejected") {
+      hadErrors = true;
+      console.warn("[hot-poll] PR status batch failed:", s.reason);
+    }
+  }
+
+  return { results, hadErrors };
+}
+
+export interface HotWorkflowRunUpdate {
+  id: number;
+  status: string;
+  conclusion: string | null;
+  updatedAt: string;
+  completedAt: string | null;
+}
+
+/**
+ * Fetches current status for a single workflow run by ID.
+ * Used by hot-poll to refresh in-progress runs without a full re-fetch.
+ */
+export async function fetchWorkflowRunById(
+  octokit: GitHubOctokit,
+  descriptor: { id: number; owner: string; repo: string }
+): Promise<HotWorkflowRunUpdate> {
+  const { id, owner, repo } = descriptor;
+  const response = await octokit.request("GET /repos/{owner}/{repo}/actions/runs/{run_id}", {
+    owner,
+    repo,
+    run_id: id,
+  });
+  updateRateLimitFromHeaders(response.headers as Record<string, string>);
+  // Octokit's generated type for this endpoint omits completed_at; cast to our full raw shape
+  const run = response.data as unknown as RawWorkflowRun;
+  return {
+    id: run.id,
+    status: run.status ?? "",
+    conclusion: run.conclusion ?? null,
+    updatedAt: run.updated_at,
+    completedAt: run.completed_at ?? null,
+  };
+}