Preserve native Codex history and add configurable cross-model compression

Separate encrypted reasoning from actual compaction, stop migration after native return, and retain full history and recent exchanges. Add typed settings, Gemini fallback, verified provider windows, bounded summaries and durable concurrent checkpoints.

Validation: 245 bridge tests plus focused concurrency tests; 71 settings/connection tests; 66 supplementary tests; TypeScript checks; renderer and Electron bundles; real primary/native/fallback summary probes and 700K input acceptance.

apps/desktop/assets/codex-bridge/test_pencere_phase.py

@@ -43,11 +43,15 @@ def message(text, ident="msg_1"):
 
 
 class Windows(unittest.TestCase):
-    def test_provider_data_local_table_and_cap(self):
+    def test_current_provider_aliases_keep_their_own_capacity(self):
+        self.assertEqual(pencere.coz('agy-gemini-3.8-flash-high', 'model', 'Gemini 3.8 Flash (High)')['pencere'], 1000000)
+        self.assertEqual(pencere.yol_bilgisi('codex/gpt-6-astra', {})['pencere'], 272000)
+
+    def test_provider_data_and_local_table_preserve_large_windows(self):
         coz = pencere.coz
         # The provider that carries the request decides; nothing is declared above the summarizer's window.
         self.assertEqual(coz("9router-nano-z-ai-glm-5.3-flash-uncensored", "model", "nano/z-ai/glm-5.3-flash-uncensored",
-                             nano=NANO)["pencere"], pencere.TAVAN)
+                             nano=NANO)["pencere"], 1048576)
         small = coz("9router-nano-mlabonne-NeuralDaredevil-8B-abliterated", "model",
                     "nano/mlabonne/NeuralDaredevil-8B-abliterated", nano=NANO)
         self.assertEqual((small["pencere"], small["arac"], small["gorsel"]), (8192, False, False))
@@ -73,6 +77,7 @@ class Windows(unittest.TestCase):
         self.assertTrue(pencere.tasma_mi(text))
         self.assertEqual(pencere.sinir_oku(text), 32768)
         self.assertEqual(pencere.sinir_oku("prompt is too long: 3040251 tokens > 1000000 maximum"), 1000000)
+        self.assertEqual(pencere.sinir_oku("Input exceeds maximum input tokens: estimated 1125036 input tokens, max input 1048576."), 1048576)
         gateway = ("Your prompt exceeds the model's context length. Estimated input tokens: 171016. "
                    "Max context tokens: 65536.")
         self.assertTrue(pencere.tasma_mi(gateway))