- Significantly changed the PDF Processor to use Mistral's OCR model

- ensure very long chunks get split into smaller chunks
- ensure TrackedMistralAIEmbedding is batched if needed to ensure correct execution
- upgraded some of the packages to a higher version

common/models/entitlements.py

@@ -25,6 +25,7 @@ class BusinessEventLog(db.Model):
     llm_metrics_prompt_tokens = db.Column(db.Integer)
     llm_metrics_completion_tokens = db.Column(db.Integer)
     llm_metrics_total_time = db.Column(db.Float)
+    llm_metrics_nr_of_pages = db.Column(db.Integer)
     llm_metrics_call_count = db.Column(db.Integer)
     llm_interaction_type = db.Column(db.String(20))
     message = db.Column(db.Text)