From af42cb40e670a218512a7d8a5faef18ddc4e4b17 Mon Sep 17 00:00:00 2001
From: jenkins <jenkins@bstein.dev>
Date: Sun, 9 Aug 2026 16:12:00 -0300
Subject: [PATCH] fix(hermes): recover browser terminal sessions

---
 services/hermes/agent-deployment.yaml       |  2 +-
 services/hermes/scripts/patch_ttyd_index.py | 96 ++++++++++++++++++++-
 testing/tests/test_hermes_herdr.py          | 19 ++++
 3 files changed, 112 insertions(+), 5 deletions(-)

diff --git a/services/hermes/agent-deployment.yaml b/services/hermes/agent-deployment.yaml
index 633e1e5c7..13a4869bb 100644
--- a/services/hermes/agent-deployment.yaml
+++ b/services/hermes/agent-deployment.yaml
@@ -24,7 +24,7 @@ spec:
         ai.bstein.dev/execution: Herdr-supervised Codex and Claude Code
         ai.bstein.dev/model-policy: Jetson-assisted AUTO routing, low through xhigh, cross-provider fallback
         ai.bstein.dev/placement: rpi5 preferred; Jetson deferred until state storage is available
-        ai.bstein.dev/config-rev: "20260809-terminal-clipboard"
+        ai.bstein.dev/config-rev: "20260809-terminal-recovery"
         vault.hashicorp.com/agent-inject: "true"
         vault.hashicorp.com/role: hermes-agent
         vault.hashicorp.com/agent-inject-secret-anthropic-token: kv/data/atlas/hermes/agent-tokens
diff --git a/services/hermes/scripts/patch_ttyd_index.py b/services/hermes/scripts/patch_ttyd_index.py
index cb4a293c8..32a1d5eb6 100644
--- a/services/hermes/scripts/patch_ttyd_index.py
+++ b/services/hermes/scripts/patch_ttyd_index.py
@@ -1,5 +1,5 @@
 #!/usr/bin/env python3
-"""Add reliable browser clipboard handling to ttyd's pinned client page."""
+"""Add reliable clipboard and reconnect handling to ttyd's pinned client."""
 
 from __future__ import annotations
 
@@ -27,9 +27,25 @@ CLIPBOARD_ADAPTER = r"""
     top: 12px;
     z-index: 10000;
   }
+  #atlas-terminal-recovery {
+    background: #181825;
+    border: 1px solid #89b4fa;
+    border-radius: 8px;
+    color: #cdd6f4;
+    font: 16px/1.4 monospace;
+    left: 50%;
+    padding: 12px 18px;
+    position: fixed;
+    top: 50%;
+    transform: translate(-50%, -50%);
+    z-index: 10001;
+  }
 </style>
 <script id="atlas-ttyd-clipboard">
 (() => {
+  let unloading = false;
+  let recoveryStarted = false;
+
   const showStatus = (message, failed = false) => {
     let status = document.getElementById('atlas-clipboard-status');
     if (!status) {
@@ -44,6 +60,78 @@ CLIPBOARD_ADAPTER = r"""
     status.removeTimer = setTimeout(() => status.remove(), 1200);
   };
 
+  const sleep = (milliseconds) => new Promise(
+    (resolve) => window.setTimeout(resolve, milliseconds),
+  );
+
+  const showRecoveryStatus = (message) => {
+    let status = document.getElementById('atlas-terminal-recovery');
+    if (!status) {
+      status = document.createElement('div');
+      status.id = 'atlas-terminal-recovery';
+      document.body.appendChild(status);
+    }
+    status.textContent = message;
+  };
+
+  const tokenUrl = new URL('token', window.location.href).toString();
+  const waitForFreshPage = async () => {
+    if (recoveryStarted || unloading) return;
+    recoveryStarted = true;
+    showRecoveryStatus('Reconnecting terminal automatically...');
+
+    let failures = 0;
+    while (!unloading) {
+      try {
+        const response = await fetch(tokenUrl, {
+          cache: 'no-store',
+          credentials: 'same-origin',
+          redirect: 'manual',
+        });
+        if (response.type === 'opaqueredirect'
+            || (response.status >= 300 && response.status < 400)) {
+          window.location.reload();
+          return;
+        }
+        if (response.ok) {
+          const payload = await response.json();
+          if (typeof payload.token === 'string') {
+            window.location.reload();
+            return;
+          }
+        }
+      } catch (error) {
+        console.debug('[atlas] terminal endpoint is not ready', error);
+      }
+
+      failures += 1;
+      if (failures === 5) {
+        showRecoveryStatus('Waiting for terminal service...');
+      }
+      await sleep(Math.min(500 + failures * 250, 3000));
+    }
+  };
+
+  const NativeWebSocket = window.WebSocket;
+  window.WebSocket = class AtlasRecoveringWebSocket extends NativeWebSocket {
+    constructor(...args) {
+      super(...args);
+      this.addEventListener('close', () => {
+        if (!unloading) void waitForFreshPage();
+      }, { once: true });
+    }
+  };
+
+  window.addEventListener('beforeunload', () => {
+    unloading = true;
+  });
+
+  window.addEventListener('keydown', (event) => {
+    if (!recoveryStarted || event.key !== 'Enter') return;
+    event.preventDefault();
+    event.stopImmediatePropagation();
+  }, true);
+
   const fallbackCopy = (text) => {
     const input = document.createElement('textarea');
     input.value = text;
@@ -111,13 +199,13 @@ CLIPBOARD_ADAPTER = r"""
 
 
 def patch_html(content: str) -> str:
-    """Insert the adapter once and disable ttyd's false-success copy call."""
-    if content.count(UPSTREAM_COPY) != 1 or content.count("</body>") != 1:
+    """Insert the adapter before ttyd and disable its false-success copy call."""
+    if content.count(UPSTREAM_COPY) != 1 or content.count("<body>") != 1:
         raise RuntimeError("ttyd index patch context changed")
     if MARKER in content:
         raise RuntimeError("ttyd index is already patched")
     content = content.replace(UPSTREAM_COPY, "void 0", 1)
-    return content.replace("</body>", f"{CLIPBOARD_ADAPTER}</body>", 1)
+    return content.replace("<body>", f"<body>{CLIPBOARD_ADAPTER}", 1)
 
 
 def fetch_embedded_index(ttyd: Path) -> str:
diff --git a/testing/tests/test_hermes_herdr.py b/testing/tests/test_hermes_herdr.py
index 69aea9717..1e42bfb8d 100644
--- a/testing/tests/test_hermes_herdr.py
+++ b/testing/tests/test_hermes_herdr.py
@@ -147,6 +147,25 @@ def test_ttyd_clipboard_patch_uses_system_clipboard_and_preserves_interrupt():
     assert "document.execCommand('copy')" in content
 
 
+def test_ttyd_client_patch_recovers_with_a_fresh_authenticated_page():
+    source = (
+        '<html><body><script>document.execCommand("copy")</script></body></html>'
+    )
+    content = ttyd_patch.patch_html(source)
+
+    assert "class AtlasRecoveringWebSocket" in content
+    assert "new URL('token', window.location.href)" in content
+    assert "credentials: 'same-origin'" in content
+    assert "response.type === 'opaqueredirect'" in content
+    assert "typeof payload.token === 'string'" in content
+    assert "window.location.reload()" in content
+    assert "Reconnecting terminal automatically..." in content
+    assert "if (!recoveryStarted || event.key !== 'Enter') return" in content
+    assert content.index('id="atlas-ttyd-clipboard"') < content.index(
+        '<script>void 0</script>'
+    )
+
+
 def test_ttyd_clipboard_patch_fails_closed_on_upstream_drift():
     with pytest.raises(RuntimeError, match="context changed"):
         ttyd_patch.patch_html("<html><body>changed</body></html>")