Don't summarize very short texts (#9)

Author: dmlls

src/jizt/config.py

@@ -44,3 +44,8 @@
     cast=Path,
     default=f"{ROOT_DIR}/language_detection/language_detection/models/lid.176.ftz"
 )
+
+# Summarization params
+# Minimum number of words a text has to have to be summarized. This prevents
+# trying to summarize very short texts, which will yield bad results.
+MIN_WORDS_SOURCE: int = 20

src/jizt/summaries/views.py

@@ -23,7 +23,7 @@
 import logging
 from fastapi import (APIRouter, HTTPException, BackgroundTasks, Depends,
                      Response, status)
-from jizt.config import LOG_LEVEL
+from jizt.config import LOG_LEVEL, MIN_WORDS_SOURCE
 from jizt.supported_languages import SupportedLanguage
 from jizt.language_detection.language_detection.language_detection import \
     LanguageDetectorSingleton
@@ -67,6 +67,12 @@ async def request_summary_view(
     if not request.source:
         raise HTTPException(status_code=status.HTTP_204_NO_CONTENT)
 
+    if len(request.source.split()) < MIN_WORDS_SOURCE:
+        raise HTTPException(
+            status_code=status.HTTP_422_UNPROCESSABLE_ENTITY,
+            detail=f"source too short (<{MIN_WORDS_SOURCE} words)"
+        )
+
     language = lang_detector.detect(request.source).language
     if not SupportedLanguage.is_supported(language):
         raise HTTPException(