From 47155fa3a0f2533f29bb86eb35ddcc2cc6b45d82 Mon Sep 17 00:00:00 2001
From: Diego Rodrigues de Sa e Souza
 <8016841+diegosouzapw@users.noreply.github.com>
Date: Thu, 11 Jun 2026 12:55:02 -0300
Subject: [PATCH] fix(api): flag provider topology error state by current
 status, not stale history (#3619) (#3666)

Closes #3619
---
 CHANGELOG.md                              |  1 +
 src/app/api/provider-metrics/route.ts     |  7 ++++-
 tests/unit/provider-metrics-route.test.ts | 37 +++++++++++++++++++++++
 3 files changed, 44 insertions(+), 1 deletion(-)

diff --git a/CHANGELOG.md b/CHANGELOG.md
index 06e96885fe..0d0e22a679 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -34,6 +34,7 @@
 
 ### 🔧 Bug Fixes
 
+- **Provider Topology**: stop flagging healthy providers as errored based on stale historical failures; use current request status (#3619)
 - **OpenCode Free**: fetch the live model catalog from the provider's `modelsUrl` for the no-auth model picker instead of serving a stale hardcoded list (#3611)
 - **Hermes Agent**: honour the `HERMES_HOME` env var when writing/reading the agent config instead of always using `~/.hermes` (#3628). Introduced `getHermesHome()` / `getHermesConfigPath()` helpers (read at call-time) and routed all four hardcoded callsites through them so OmniRoute's config lands in the same directory that the Hermes PowerShell installer configures on Windows.
 - **MITM/cert**: remove the duplicated "Command failed:" prefix in system-command error messages ([#3641](https://github.com/diegosouzapw/OmniRoute/issues/3641)): `execFileText` was prepending its own `"Command failed: "` prefix on top of Node's `execFile` error message, which already begins with `"Command failed: <cmd>"` for non-zero exits. The error message now surfaces Node's message directly (no double prefix), with stderr appended only when non-empty.
diff --git a/src/app/api/provider-metrics/route.ts b/src/app/api/provider-metrics/route.ts
index 38a16da010..d849cb9a9e 100644
--- a/src/app/api/provider-metrics/route.ts
+++ b/src/app/api/provider-metrics/route.ts
@@ -71,7 +71,12 @@ export async function GET() {
         lastProviderTs = requestTs;
       }
 
-      const errorTs = lastErrorAt ? Date.parse(lastErrorAt) : 0;
+      // Only flag as errorProvider if the provider's MOST RECENT request was itself
+      // a failure. A provider with a historical lastErrorAt but a recent success
+      // (lastStatus 2xx/3xx) must not be shown as currently errored (#3619).
+      const isCurrentlyInError =
+        lastStatus !== null && (lastStatus < 200 || lastStatus >= 400);
+      const errorTs = isCurrentlyInError && lastErrorAt ? Date.parse(lastErrorAt) : 0;
       if (Number.isFinite(errorTs) && errorTs > errorProviderTs) {
         errorProvider = provider;
         errorProviderTs = errorTs;
diff --git a/tests/unit/provider-metrics-route.test.ts b/tests/unit/provider-metrics-route.test.ts
index e56a71afea..ca5def5bb5 100644
--- a/tests/unit/provider-metrics-route.test.ts
+++ b/tests/unit/provider-metrics-route.test.ts
@@ -85,6 +85,43 @@ test("GET /api/provider-metrics includes provider recency and error topology", a
   assert.equal(body.topology.errorProvider, "openai");
 });
 
+test("GET /api/provider-metrics errorProvider must NOT flag a provider whose most recent request succeeded", async () => {
+  // Arrange: providerA had an error long ago but recovered (last request = 200).
+  //          providerB never errored.
+  // Bug (pre-fix): errorProvider = "providerA" because lastErrorAt > 0.
+  // Fix (post-fix): errorProvider = "" because lastStatus for providerA is 200.
+  const db = core.getDbInstance();
+  db.prepare(
+    `INSERT INTO call_logs (id, timestamp, provider, status, duration, error_summary)
+     VALUES (?, ?, ?, ?, ?, ?)`
+  ).run("a-old-error", "2026-01-01T00:00:00.000Z", "providerA", 500, 100, "old error");
+  db.prepare(
+    `INSERT INTO call_logs (id, timestamp, provider, status, duration, error_summary)
+     VALUES (?, ?, ?, ?, ?, ?)`
+  ).run("a-recent-ok", "2026-06-01T00:00:00.000Z", "providerA", 200, 80, null);
+  db.prepare(
+    `INSERT INTO call_logs (id, timestamp, provider, status, duration, error_summary)
+     VALUES (?, ?, ?, ?, ?, ?)`
+  ).run("b-ok", "2026-05-01T00:00:00.000Z", "providerB", 200, 60, null);
+
+  const response = await providerMetricsRoute.GET();
+  const body = (await response.json()) as ProviderMetricsResponse;
+
+  assert.equal(response.status, 200);
+  // providerA's last request succeeded — must NOT appear as errorProvider
+  assert.notEqual(
+    body.topology.errorProvider,
+    "providerA",
+    "providerA recovered (lastStatus=200) and must not be marked as errorProvider"
+  );
+  // No provider is currently in error — errorProvider should be empty
+  assert.equal(
+    body.topology.errorProvider,
+    "",
+    "errorProvider must be empty when no provider's most recent request was a failure"
+  );
+});
+
 test("GET /api/provider-metrics returns sanitized 500 when metrics cannot be loaded", async () => {
   const db = core.getDbInstance();
   db.close();