[update] added validation for article data in NewsList.vue, removed unused Cypress config, expanded LLM models in example.env, adjusted context size and max article length in backend configuration, and updated workspace naming in yarn.lock

backend/app/config.py

@@ -8,11 +8,11 @@ MIN_CRON_HOURS = float(os.getenv("MIN_CRON_HOURS", 0.5))
 DEFAULT_CRON_HOURS = float(os.getenv("CRON_HOURS", MIN_CRON_HOURS))
 CRON_HOURS = max(MIN_CRON_HOURS, DEFAULT_CRON_HOURS)
 SYNC_COOLDOWN_MINUTES = int(os.getenv("SYNC_COOLDOWN_MINUTES", 30))
-LLM_MODEL = os.getenv("LLM_MODEL", "mistral-nemo:12b")
+LLM_MODEL = os.getenv("LLM_MODEL", "phi3:3.8b-mini-128k-instruct-q4_0")
 LLM_TIMEOUT_SECONDS = int(os.getenv("LLM_TIMEOUT_SECONDS", 180))
 OLLAMA_API_TIMEOUT_SECONDS = int(os.getenv("OLLAMA_API_TIMEOUT_SECONDS", 10))
 ARTICLE_FETCH_TIMEOUT = int(os.getenv("ARTICLE_FETCH_TIMEOUT", 30))
-MAX_ARTICLE_LENGTH = int(os.getenv("MAX_ARTICLE_LENGTH", 10_000))
+MAX_ARTICLE_LENGTH = int(os.getenv("MAX_ARTICLE_LENGTH", 40_000))
 
 frontend_path = os.path.join(
     os.path.dirname(os.path.dirname(os.path.dirname(__file__))),

backend/app/services.py

@@ -209,7 +209,7 @@ class NewsFetcher:
             "format": "json",
             "options": {
                 "num_gpu": 1,  # Force GPU usage
-                "num_ctx": 128_000,  # Context size
+                "num_ctx": 64_000,  # Context size
             }
         }
 

backend/example.env

@@ -11,10 +11,11 @@ MIN_CRON_HOURS=0.5
 SYNC_COOLDOWN_MINUTES=30
 
 # LLM model to use for summarization
-LLM_MODEL=qwen2:7b-instruct-q4_K_M
-LLM_MODEL=phi3:3.8b-mini-128k-instruct-q4_0
-LLM_MODEL=mistral-nemo:12b
-LLM_MODEL=cnjack/mistral-samll-3.1:24b-it-q4_K_S
+LLM_MODEL=qwen2:7b-instruct-q4_K_M # ca 7-9GB (typisch 8GB)
+LLM_MODEL=phi3:3.8b-mini-128k-instruct-q4_0 # ca 6-8GB (langer kontext)
+LLM_MODEL=mistral-nemo:12b # ca 16-24+GB
+LLM_MODEL=cnjack/mistral-samll-3.1:24b-it-q4_K_S # ca 22GB
+LLM_MODEL=yarn-mistral:7b-64k-q4_K_M # ca 11GB
 
 # Timeout in seconds for LLM requests
 LLM_TIMEOUT_SECONDS=180