feat(gateway): hot-reload memory_char_limit and user_char_limit

gateway/run.py

@@ -16083,6 +16083,11 @@ async def _run_process_watcher(self, watcher: dict) -> None:
         ("compression", "protect_last_n"),
         ("agent", "disabled_toolsets"),
         ("memory", "provider"),
+        # memory_char_limit / user_char_limit are passed to MemoryStore at
+        # agent construction; changing them mid-session must rebuild the
+        # cached agent so the new limits take effect on the next message.
+        ("memory", "memory_char_limit"),
+        ("memory", "user_char_limit"),
     )
 
     _HONCHO_CACHE_BUSTING_KEYS = (

tests/gateway/test_agent_cache.py

@@ -237,6 +237,50 @@ def test_reads_compression_subkeys(self):
         assert out["compression.target_ratio"] == 0.3
         assert out["compression.protect_last_n"] == 25
 
+    def test_reads_memory_char_limit_subkeys(self):
+        """memory_char_limit / user_char_limit must be picked up so editing
+        the cap in config.yaml invalidates the cached agent on the next
+        message (no gateway restart)."""
+        from gateway.run import GatewayRunner
+
+        out = GatewayRunner._extract_cache_busting_config(
+            {
+                "memory": {
+                    "memory_char_limit": 5000,
+                    "user_char_limit": 2500,
+                    "provider": "mem0",  # unrelated — should not affect these
+                }
+            }
+        )
+        assert out["memory.memory_char_limit"] == 5000
+        assert out["memory.user_char_limit"] == 2500
+
+    def test_memory_char_limit_change_busts_signature(self):
+        """Changing memory_char_limit between messages must produce a
+        different agent-config signature, so the gateway rebuilds the
+        cached AIAgent and the new cap takes effect immediately."""
+        from gateway.run import GatewayRunner
+
+        sig_default = GatewayRunner._agent_config_signature(
+            model="minimax/MiniMax-M2.7",
+            runtime={"api_key": "k", "base_url": "u", "provider": "p", "api_mode": ""},
+            enabled_toolsets=["memory"],
+            ephemeral_prompt="",
+            cache_keys=GatewayRunner._extract_cache_busting_config(
+                {"memory": {"memory_char_limit": 2200, "user_char_limit": 1375}}
+            ),
+        )
+        sig_bumped = GatewayRunner._agent_config_signature(
+            model="minimax/MiniMax-M2.7",
+            runtime={"api_key": "k", "base_url": "u", "provider": "p", "api_mode": ""},
+            enabled_toolsets=["memory"],
+            ephemeral_prompt="",
+            cache_keys=GatewayRunner._extract_cache_busting_config(
+                {"memory": {"memory_char_limit": 5000, "user_char_limit": 2500}}
+            ),
+        )
+        assert sig_default != sig_bumped
+
     def test_missing_keys_yield_none(self):
         """Absent config keys must produce None values (still contribute to signature)."""
         from gateway.run import GatewayRunner

website/docs/user-guide/configuration.md

@@ -651,7 +651,7 @@ Older configs with `compression.summary_model`, `compression.summary_provider`,
 `protect_first_n` controls how many **non-system** head messages are pinned across every compaction. Default `3` — the opening user/assistant exchange survives every summarizer pass so the original goal stays visible. On long-running rolling-compaction sessions where the opening turn is no longer relevant, set `protect_first_n: 0` to pin nothing but the system prompt + summary + tail. The system prompt itself is always preserved regardless of this setting.
 
 :::tip Gateway hot-reload of compression and context length
-As of recent releases, editing `model.context_length` or any `compression.*` key in `config.yaml` on a running gateway takes effect on the next message — no gateway restart, no `/reset`, no session rotation required. The cached-agent signature includes these keys, so the gateway transparently rebuilds the agent when it sees a change. API keys and tool/skill config still require the usual reload paths.
+As of recent releases, editing `model.context_length`, any `compression.*` key, or `memory.memory_char_limit` / `memory.user_char_limit` in `config.yaml` on a running gateway takes effect on the next message — no gateway restart, no `/reset`, no session rotation required. The cached-agent signature includes these keys, so the gateway transparently rebuilds the agent when it sees a change. API keys and tool/skill config still require the usual reload paths.
 :::
 
 ### Common setups