diff --git a/README.md b/README.md
index 251181b..d788e5e 100644
--- a/README.md
+++ b/README.md
@@ -420,7 +420,7 @@ Uses **WebSocket** long connection — no public IP required.
 **1. Create a Feishu bot**
 - Visit [Feishu Open Platform](https://open.feishu.cn/app)
 - Create a new app → Enable **Bot** capability
-- **Permissions**: Add `im:message` (send messages)
+- **Permissions**: Add `im:message` (send messages) and `im:message.p2p_msg:readonly` (receive messages)
 - **Events**: Add `im.message.receive_v1` (receive messages)
   - Select **Long Connection** mode (requires running nanobot first to establish connection)
 - Get **App ID** and **App Secret** from "Credentials & Basic Info"
diff --git a/nanobot/agent/loop.py b/nanobot/agent/loop.py
index e30ed23..b93c477 100644
--- a/nanobot/agent/loop.py
+++ b/nanobot/agent/loop.py
@@ -227,6 +227,12 @@ class AgentLoop:
                     )
             else:
                 clean = self._strip_think(response.content)
+                # Don't persist error responses to session history — they can
+                # poison the context and cause permanent 400 loops (#1303).
+                if response.finish_reason == "error":
+                    logger.error("LLM returned error: {}", (clean or "")[:200])
+                    final_content = clean or "Sorry, I encountered an error calling the AI model."
+                    break
                 messages = self.context.add_assistant_message(
                     messages, clean, reasoning_content=response.reasoning_content,
                 )
@@ -448,6 +454,8 @@ class AgentLoop:
         for m in messages[skip:]:
             entry = {k: v for k, v in m.items() if k != "reasoning_content"}
             role, content = entry.get("role"), entry.get("content")
+            if role == "assistant" and not content and not entry.get("tool_calls"):
+                continue  # skip empty assistant messages — they poison session context
             if role == "tool" and isinstance(content, str) and len(content) > self._TOOL_RESULT_MAX_CHARS:
                 entry["content"] = content[:self._TOOL_RESULT_MAX_CHARS] + "\n... (truncated)"
             elif role == "user":
diff --git a/nanobot/channels/feishu.py b/nanobot/channels/feishu.py
index 4a6312e..6703f21 100644
--- a/nanobot/channels/feishu.py
+++ b/nanobot/channels/feishu.py
@@ -89,8 +89,9 @@ def _extract_interactive_content(content: dict) -> list[str]:
         elif isinstance(title, str):
             parts.append(f"title: {title}")
 
-    for element in content.get("elements", []) if isinstance(content.get("elements"), list) else []:
-        parts.extend(_extract_element_content(element))
+    for elements in content.get("elements", []) if isinstance(content.get("elements"), list) else []:
+        for element in elements:
+            parts.extend(_extract_element_content(element))
 
     card = content.get("card", {})
     if card:
diff --git a/nanobot/channels/qq.py b/nanobot/channels/qq.py
index 5352a30..50dbbde 100644
--- a/nanobot/channels/qq.py
+++ b/nanobot/channels/qq.py
@@ -100,10 +100,12 @@ class QQChannel(BaseChannel):
             logger.warning("QQ client not initialized")
             return
         try:
+            msg_id = msg.metadata.get("message_id")
             await self._client.api.post_c2c_message(
                 openid=msg.chat_id,
                 msg_type=0,
                 content=msg.content,
+                msg_id=msg_id,
             )
         except Exception as e:
             logger.error("Error sending QQ message: {}", e)
diff --git a/nanobot/providers/litellm_provider.py b/nanobot/providers/litellm_provider.py
index 0918954..5427d97 100644
--- a/nanobot/providers/litellm_provider.py
+++ b/nanobot/providers/litellm_provider.py
@@ -3,6 +3,8 @@
 import json
 import json_repair
 import os
+import secrets
+import string
 from typing import Any
 
 import litellm
@@ -15,6 +17,11 @@ from nanobot.providers.registry import find_by_model, find_gateway
 # Standard OpenAI chat-completion message keys plus reasoning_content for
 # thinking-enabled models (Kimi k2.5, DeepSeek-R1, etc.).
 _ALLOWED_MSG_KEYS = frozenset({"role", "content", "tool_calls", "tool_call_id", "name", "reasoning_content"})
+_ALNUM = string.ascii_letters + string.digits
+
+def _short_tool_id() -> str:
+    """Generate a 9-char alphanumeric ID compatible with all providers (incl. Mistral)."""
+    return "".join(secrets.choice(_ALNUM) for _ in range(9))
 
 
 class LiteLLMProvider(LLMProvider):
@@ -245,7 +252,7 @@ class LiteLLMProvider(LLMProvider):
                     args = json_repair.loads(args)
                 
                 tool_calls.append(ToolCallRequest(
-                    id=tc.id,
+                    id=_short_tool_id(),
                     name=tc.function.name,
                     arguments=args,
                 ))
diff --git a/nanobot/providers/registry.py b/nanobot/providers/registry.py
index 2766929..df915b7 100644
--- a/nanobot/providers/registry.py
+++ b/nanobot/providers/registry.py
@@ -201,7 +201,7 @@ PROVIDERS: tuple[ProviderSpec, ...] = (
     # OpenAI Codex: uses OAuth, not API key.
     ProviderSpec(
         name="openai_codex",
-        keywords=("openai-codex", "codex"),
+        keywords=("openai-codex",),
         env_key="",                         # OAuth-based, no API key
         display_name="OpenAI Codex",
         litellm_prefix="",                  # Not routed through LiteLLM