Skip to content

Self-evolving Context Database for AI Agents. Unify Agent Memory, Knowledge RAG and Skills.

fix(language): ignore bare import paths during detection (#4543)

chenyu7b0482576cc1
diff --git a/openviking/session/memory/utils/language.py b/openviking/session/memory/utils/language.py
index 7b57c668a..edc325bcc 100644
--- a/openviking/session/memory/utils/language.py
+++ b/openviking/session/memory/utils/language.py
@@ -21,7 +21,12 @@ _JAPANESE_KANA_MIN_CHARS = 3
_STRONG_DOMINANT_MIN_CHARS = 10
_STRONG_DOMINANT_RATIO = 0.95
_PRIMARY_LANGUAGES = {"zh-CN", "en"}
-_URI_LANGUAGE_NOISE_RE = re.compile(r"\b(?:viking|https?)://[^\s<>\]\)\"']+")
+# Bare import paths are machine tokens too: repeated .com domains otherwise
+# count as the Portuguese stopword "com" in code skeletons.
+_URI_LANGUAGE_NOISE_RE = re.compile(
+ r"\b(?:(?:viking|https?)://|(?:[a-zA-Z0-9-]+\.)+[a-zA-Z]{2,}/)"
+ r"[^\s<>\]\)\"'`]+"
+)
_LATIN_STOPWORDS = {
"en": set(
@@ -373,7 +378,7 @@ def resolve_output_language_from_text(
def strip_language_detection_noise(text: str) -> str:
- """Remove URI-like machine tokens that should not affect output language."""
+ """Remove URIs and bare domain paths that should not affect output language."""
return _URI_LANGUAGE_NOISE_RE.sub(" ", text or "")
diff --git a/tests/storage/test_semantic_processor_language.py b/tests/storage/test_semantic_processor_language.py
index 7861f48a0..2dd284cf4 100644
--- a/tests/storage/test_semantic_processor_language.py
+++ b/tests/storage/test_semantic_processor_language.py
@@ -185,6 +185,47 @@ class TestLanguageFlow:
class TestOverviewGenerationFlow:
"""目录概述生成流程测试。"""
+ @pytest.mark.asyncio
+ @pytest.mark.parametrize(
+ "description,override,expected_language",
+ [
+ ("", "", "en"),
+ ("这是用于查询任务状态和统计运行时间的客户端代码。", "", "zh-CN"),
+ (
+ "Este documento descreve as preferências do usuário e o projeto para completar.",
+ "",
+ "pt",
+ ),
+ ("", "zh-CN", "zh-CN"),
+ ],
+ )
+ async def test_import_paths_do_not_set_overview_language(
+ self, description, override, expected_language
+ ):
+ from openviking.storage.queuefs.semantic_processor import SemanticProcessor
+
+ imports = [f'"github.com/example/module{index}"' for index in range(6)]
+ skeleton = "package client\nimport (\n" + "\n".join(imports) + "\n)\nfunc NewClient()"
+ config = MagicMock()
+ config.output_language_override = override
+ config.semantic.max_overview_prompt_chars = 60_000
+ config.semantic.overview_batch_size = 50
+ config.vlm.get_completion_async = AsyncMock(return_value="# client\nClient overview.")
+
+ with patch(
+ "openviking.storage.queuefs.semantic_processor.get_openviking_config",
+ return_value=config,
+ ):
+ await SemanticProcessor()._generate_overview(
+ "viking://resources/example/client",
+ [{"name": "client.go", "summary": description + "\n" + skeleton}],
+ [],
+ )
+
+ prompt = config.vlm.get_completion_async.call_args.args[0]
+ assert f"Output Language: {expected_language}" in prompt
+ assert all(path in prompt for path in imports)
+
@pytest.mark.parametrize(
"lang,file_summaries",
[
@@ -265,6 +306,7 @@ class TestOverviewGenerationFlow:
assert "**Quick Navigation** (H2)" not in prompt
assert "**Detailed Description** (H2)" not in prompt
+
class LanguageAwareMockVLM:
"""语言感知的 MockVLM,根据 prompt 中的 Output Language 返回对应语言的响应。"""