feat(storage): add auto language detection for semantic summary generation

openviking/prompts/templates/semantic/code_ast_summary.yaml

@@ -17,7 +17,14 @@ variables:
     description: "AST-extracted code skeleton (classes, functions, imports)"
     required: true
 
+  - name: "output_language"
+    type: "string"
+    description: "Language code for output (e.g., 'en', 'zh-CN', 'ja', 'ko', 'ru', 'ar')"
+    required: true
+
 template: |
+  Output Language: {{ output_language }}
+
   You are a code analysis expert. Based on the structural skeleton below,
   generate a concise summary (80-200 words) focusing on purpose, key components,
   and relationships. The skeleton was extracted via AST parsing.

openviking/prompts/templates/semantic/code_summary.yaml

@@ -17,7 +17,14 @@ variables:
     description: "File content (code)"
     required: true
 
+  - name: "output_language"
+    type: "string"
+    description: "Language code for output (e.g., 'en', 'zh-CN', 'ja', 'ko', 'ru', 'ar')"
+    required: true
+
 template: |
+  Output Language: {{ output_language }}
+
   You are a code analysis expert. Generate a concise yet informative summary for the following code file.
 
   【File Name】
@@ -43,4 +50,4 @@ template: |
   4. Role in the larger codebase context
 
 llm_config:
-  temperature: 0.0
+  temperature: 0.0

openviking/prompts/templates/semantic/document_summary.yaml

@@ -17,7 +17,14 @@ variables:
     description: "File content (documentation text)"
     required: true
 
+  - name: "output_language"
+    type: "string"
+    description: "Language code for output (e.g., 'en', 'zh-CN', 'ja', 'ko', 'ru', 'ar')"
+    required: true
+
 template: |
+  Output Language: {{ output_language }}
+
   You are a documentation analysis expert. Generate a concise yet informative summary for the following documentation file.
 
   【File Name】
@@ -49,4 +56,4 @@ template: |
   - For reference docs: focus on completeness and organization
 
 llm_config:
-  temperature: 0.0
+  temperature: 0.0

openviking/prompts/templates/semantic/file_summary.yaml

@@ -17,7 +17,14 @@ variables:
     description: "File content"
     required: true
 
+  - name: "output_language"
+    type: "string"
+    description: "Language code for output (e.g., 'en', 'zh-CN', 'ja', 'ko', 'ru', 'ar')"
+    required: true
+
 template: |
+  Output Language: {{ output_language }}
+
   Please generate a summary for the following file:
 
   【File Name】

openviking/prompts/templates/semantic/overview_generation.yaml

@@ -22,7 +22,14 @@ variables:
     description: "List of subdirectories and their abstracts, format: '- dirname/: abstract'"
     default: ""
 
+  - name: "output_language"
+    type: "string"
+    description: "Language code for output (e.g., 'en', 'zh-CN', 'ja', 'ko', 'ru', 'ar')"
+    required: true
+
 template: |
+  Output Language: {{ output_language }}
+
   Generate an overview document based on the following directory content:
 
   [Directory Name]

openviking/storage/queuefs/semantic_processor.py

@@ -729,6 +729,11 @@ async def _generate_text_summary(
             logger.warning("VLM not available, using empty summary")
             return {"name": file_name, "summary": ""}
 
+        from openviking.session.memory.utils.language import _detect_language_from_text
+
+        fallback_language = (get_openviking_config().language_fallback or "en").strip() or "en"
+        output_language = _detect_language_from_text(content, fallback_language)
+
         # Detect file type and select appropriate prompt
         file_type = self._detect_file_type(file_name)
 
@@ -749,7 +754,11 @@ async def _generate_text_summary(
                     else:  # ast_llm
                         prompt = render_prompt(
                             "semantic.code_ast_summary",
-                            {"file_name": file_name, "skeleton": skeleton_text},
+                            {
+                                "file_name": file_name,
+                                "skeleton": skeleton_text,
+                                "output_language": output_language,
+                            },
                         )
                         async with llm_sem:
                             summary = await vlm.get_completion_async(prompt)
@@ -762,7 +771,7 @@ async def _generate_text_summary(
             # "llm" mode or fallback when skeleton is None/empty
             prompt = render_prompt(
                 "semantic.code_summary",
-                {"file_name": file_name, "content": content},
+                {"file_name": file_name, "content": content, "output_language": output_language},
             )
             async with llm_sem:
                 summary = await vlm.get_completion_async(prompt)
@@ -775,7 +784,7 @@ async def _generate_text_summary(
 
         prompt = render_prompt(
             prompt_id,
-            {"file_name": file_name, "content": content},
+            {"file_name": file_name, "content": content, "output_language": output_language},
         )
 
         async with llm_sem:
@@ -924,6 +933,10 @@ async def _generate_overview(
             logger.warning("VLM not available, using default overview")
             return f"# {dir_uri.split('/')[-1]}\n\nDirectory overview"
 
+        from openviking.session.memory.utils.language import _detect_language_from_text
+
+        fallback_language = (config.language_fallback or "en").strip() or "en"
+
         # Build file index mapping and summary string
         file_index_map = {}
         file_summaries_lines = []
@@ -932,6 +945,8 @@ async def _generate_overview(
             file_summaries_lines.append(f"[{idx}] {item['name']}: {item['summary']}")
         file_summaries_str = "\n".join(file_summaries_lines) if file_summaries_lines else "None"
 
+        output_language = _detect_language_from_text(file_summaries_str, fallback_language)
+
         # Build subdirectory summary string
         children_abstracts_str = (
             "\n".join(f"- {item['name']}/: {item['abstract']}" for item in children_abstracts)
@@ -957,6 +972,7 @@ async def _generate_overview(
                 children_abstracts,
                 file_index_map,
                 llm_sem=llm_sem,
+                output_language=output_language,
             )
         elif over_budget:
             # Few files but long summaries → truncate summaries to fit budget
@@ -978,13 +994,15 @@ async def _generate_overview(
                 file_summaries_str,
                 children_abstracts_str,
                 file_index_map,
+                output_language=output_language,
             )
         else:
             overview = await self._single_generate_overview(
                 dir_uri,
                 file_summaries_str,
                 children_abstracts_str,
                 file_index_map,
+                output_language=output_language,
             )
 
         return overview
@@ -995,6 +1013,7 @@ async def _single_generate_overview(
         file_summaries_str: str,
         children_abstracts_str: str,
         file_index_map: Dict[int, str],
+        output_language: str = "en",
     ) -> str:
         """Generate overview from a single prompt (small directories)."""
         import re
@@ -1008,6 +1027,7 @@ async def _single_generate_overview(
                     "dir_name": dir_uri.split("/")[-1],
                     "file_summaries": file_summaries_str,
                     "children_abstracts": children_abstracts_str,
+                    "output_language": output_language,
                 },
             )
 
@@ -1036,6 +1056,7 @@ async def _batched_generate_overview(
         children_abstracts: List[Dict[str, str]],
         file_index_map: Dict[int, str],
         llm_sem: Optional[asyncio.Semaphore] = None,
+        output_language: str = "en",
     ) -> str:
         """Generate overview by batching file summaries and merging.
 
@@ -1089,6 +1110,7 @@ async def _batched_generate_overview(
                     "dir_name": dir_name,
                     "file_summaries": batch_str,
                     "children_abstracts": children_str,
+                    "output_language": output_language,
                 },
             )
             batch_prompts.append((batch_idx, prompt, batch_index_map))
@@ -1131,6 +1153,7 @@ async def _run_batch(batch_idx: int, prompt: str, batch_index_map: Dict[int, str
                     "dir_name": dir_name,
                     "file_summaries": combined,
                     "children_abstracts": children_abstracts_str,
+                    "output_language": output_language,
                 },
             )
             overview = await vlm.get_completion_async(prompt)

openviking_cli/utils/config/open_viking_config.py

@@ -139,8 +139,8 @@ class OpenVikingConfig(BaseModel):
     language_fallback: str = Field(
         default="en",
         description=(
-            "Fallback language used by memory extraction when dominant user language "
-            "cannot be confidently detected"
+            "Fallback language used by memory extraction and semantic processing when dominant "
+            "user language cannot be confidently detected"
         ),
     )