fix: Improve extraction prompt with JSON structure and fix fact parsing

- Updated prompt to specify expected JSON structure for facts and entities
- Added 'text' field support in fact parsing (alongside 'full_text')
- Listed explicit type values for facts and entities

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

zopk_knowledge_service.py

@@ -94,12 +94,25 @@ RELATION_TYPES = [
 
 # Ultra-simplified prompt to avoid Gemini safety filter issues
 # Note: Complex JSON schemas with pipe characters were triggering filters
-EXTRACTION_USER_PROMPT = """Przeanalizuj artykuł i zwróć JSON.
+# Note: max_tokens parameter also triggers filters - don't use it!
+EXTRACTION_USER_PROMPT = """Przeanalizuj artykuł i wyodrębnij informacje w formacie JSON.
 
 ARTYKUŁ:
 {chunk_text}
 
-Zwróć JSON z facts, entities, summary."""
+Zwróć JSON z następującą strukturą:
+{{
+  "facts": [
+    {{"text": "pełny fakt", "type": "investment"}}
+  ],
+  "entities": [
+    {{"name": "Nazwa", "type": "company"}}
+  ],
+  "summary": "krótkie podsumowanie"
+}}
+
+Typy faktów: investment, decision, event, statistic, partnership, milestone
+Typy encji: company, person, place, organization, project"""
 
 # System prompt is now empty - the user prompt contains all necessary instructions
 EXTRACTION_SYSTEM_PROMPT = ""
@@ -429,6 +442,9 @@ class ZOPKKnowledgeService:
                     except (ValueError, TypeError):
                         date_value = None
 
+                # Support both old format (full_text) and new simplified format (text)
+                fact_text = fact.get('text') or fact.get('full_text', '')
+
                 db_fact = ZOPKKnowledgeFact(
                     source_chunk_id=chunk.id,
                     source_news_id=news.id,
@@ -436,7 +452,7 @@ class ZOPKKnowledgeService:
                     subject=fact.get('subject'),
                     predicate=fact.get('predicate'),
                     object=fact.get('object'),
-                    full_text=fact.get('full_text', ''),
+                    full_text=fact_text,
                     numeric_value=numeric_value,
                     numeric_unit=fact.get('numeric_unit'),
                     date_value=date_value,