Fix #2051: [Feature]: LocalAgentRunner Agent Loop 应增加上下文保护与截断机制

src/langbot/pkg/provider/runners/localagent.py

@@ -10,6 +10,9 @@
 import langbot_plugin.api.entities.builtin.rag.context as rag_context
 
 
+MAX_TOOL_ITERATIONS = 16
+MAX_TOOL_RESULT_LENGTH = 8000
+
 rag_combined_prompt_template = """
 The following are relevant context entries retrieved from the knowledge base. 
 Please use them to answer the user's message. 
@@ -290,7 +293,9 @@ async def run(
 
         # Once a model succeeds, commit to it for the tool call loop
         # (no fallback mid-conversation — different models may interpret tool results differently)
-        while pending_tool_calls:
+        iteration_count = 0
+        while pending_tool_calls and iteration_count < MAX_TOOL_ITERATIONS:
+            iteration_count += 1
             for tool_call in pending_tool_calls:
                 try:
                     func = tool_call.function
@@ -312,6 +317,8 @@ async def run(
                         tool_content = func_ret
                     else:
                         tool_content = json.dumps(func_ret, ensure_ascii=False)
+                        if len(tool_content) > MAX_TOOL_RESULT_LENGTH:
+                            tool_content = tool_content[:MAX_TOOL_RESULT_LENGTH] + '\n...[truncated]'
 
                     if is_stream:
                         msg = provider_message.MessageChunk(

tests/unit_tests/provider/test_localagent.py

@@ -0,0 +1,67 @@
+"""
+Unit tests for LocalAgentRunner protection mechanisms:
+- Maximum tool-call iteration limit
+- Tool result length truncation
+
+These tests verify the logic independently of the full langbot import chain.
+"""
+
+import json
+import pytest
+
+
+# Mirror the module-level constants from localagent.py so tests remain valid
+# even when the full import chain is unavailable.
+MAX_TOOL_ITERATIONS = 16
+MAX_TOOL_RESULT_LENGTH = 8000
+
+
+def test_max_tool_iterations_constant():
+    """MAX_TOOL_ITERATIONS should be a positive integer."""
+    assert isinstance(MAX_TOOL_ITERATIONS, int)
+    assert MAX_TOOL_ITERATIONS > 0
+
+
+def test_max_tool_result_length_constant():
+    """MAX_TOOL_RESULT_LENGTH should be a positive integer."""
+    assert isinstance(MAX_TOOL_RESULT_LENGTH, int)
+    assert MAX_TOOL_RESULT_LENGTH > 0
+
+
+def test_tool_result_truncation_logic():
+    """Simulate the truncation logic applied in LocalAgentRunner."""
+    large_result = {"data": "x" * (MAX_TOOL_RESULT_LENGTH + 1000)}
+    tool_content = json.dumps(large_result, ensure_ascii=False)
+
+    if len(tool_content) > MAX_TOOL_RESULT_LENGTH:
+        tool_content = tool_content[:MAX_TOOL_RESULT_LENGTH] + '\n...[truncated]'
+
+    assert len(tool_content) == MAX_TOOL_RESULT_LENGTH + len('\n...[truncated]')
+    assert tool_content.endswith('\n...[truncated]')
+
+
+def test_tool_result_no_truncation_when_within_limit():
+    """Short tool results should not be truncated."""
+    small_result = {"data": "hello"}
+    tool_content = json.dumps(small_result, ensure_ascii=False)
+
+    original = tool_content
+    if len(tool_content) > MAX_TOOL_RESULT_LENGTH:
+        tool_content = tool_content[:MAX_TOOL_RESULT_LENGTH] + '\n...[truncated]'
+
+    assert tool_content == original
+
+
+def test_iteration_limit_logic():
+    """Simulate the iteration counter guard used in the agent loop."""
+    pending = [object()]  # always has a pending call
+    iteration_count = 0
+    executed = 0
+
+    while pending and iteration_count < MAX_TOOL_ITERATIONS:
+        iteration_count += 1
+        executed += 1
+        # pending never clears — simulates infinite tool-call loop
+
+    assert executed == MAX_TOOL_ITERATIONS
+    assert iteration_count == MAX_TOOL_ITERATIONS