clarify public disclosures to prevent 'sorry I can't do that' response

automated_updates/modules/process/parse_house_clean_llm.py

@@ -7,17 +7,18 @@
 import re
 
 def assets_from_house_clean_image_to_csv(input_image_path):
-
 	base64_image = encode_image(input_image_path)
-	response = send_to_api("is there an assets table in the image with a column called 'Asset'? answer Y or N only.", base64_image)
+	if False:
+		# this idea was to skip pages that don't list assets. it is a waste of tokens and doesn't seem to hurt accuracy to just shove all pages
+		response = send_to_api("is there an assets table in the image with a column called 'Asset'? answer Y or N only.", base64_image)
 
-	if response.lower() == 'n':
-		# not present, skip the page
-		print('skipped')
-		return []
+		if response.lower() == 'n':
+			# not present, skip the page
+			print('skipped')
+			return []
 
 	# use '|' separator to avoid characters in asset names
-	response = send_to_api("get the asset names in the 'Asset' column of the Assets table. return only a | separated list. no other commentary.", base64_image)
+	response = send_to_api("get the asset names in the 'Asset' column of the Assets table in the public disclosure form. return only a | separated list. no other commentary.", base64_image)
 	asset_list = [response.strip() for response in response.split("|")]
 
 	return asset_list

automated_updates/modules/process/parse_house_messy_llm.py

@@ -8,16 +8,18 @@
 
 def assets_from_house_messy_image_to_csv(input_image_path):
 	base64_image = encode_image(input_image_path)
-	response = send_to_api(message="is this the schedule A part of the form that lists assets? answer Y or N only.", 
-						   base64_image=base64_image)
+	if False:
+		# this idea was to skip pages that don't list assets. it is a waste of tokens and doesn't seem to hurt accuracy to just shove all pages
+		response = send_to_api(message="is this the schedule A part of the form that lists assets? answer Y or N only.", 
+							base64_image=base64_image)
 
-	if response.lower() == 'n':
-		# not schedule A, skip the page
-		print('skipped')
-		return []
+		if response.lower() == 'n':
+			# not schedule A, skip the page
+			print('skipped')
+			return []
 
 	# use '|' separator to avoid characters in asset names
-	response = send_to_api(message="get the asset names in the image, return them in a | separated list only, no other commentary.", 
+	response = send_to_api(message="get the asset names in the public disclosure form, return them in a | separated list only, no other commentary.", 
 						   base64_image=base64_image, 
 						   model="gpt-4o")
 	asset_list = [response.strip() for response in response.split("|")]

automated_updates/modules/process/parse_senate_llm.py

@@ -8,16 +8,18 @@
 
 def assets_from_senate_image_to_csv(input_image_path):
 	base64_image = encode_image(input_image_path)
-	response = send_to_api(message="does this part of the form list asset disclosures? answer Y or N only.", 
-						   base64_image=base64_image)
+	if False:
+		# this idea was to skip pages that don't list assets. it is a waste of tokens and doesn't seem to hurt accuracy to just shove all pages
+		response = send_to_api(message="does this part of the form list asset disclosures? answer Y or N only.", 
+							base64_image=base64_image)
 
-	if response.lower() == 'n':
-		print('skipped')
-		# not schedule A, skip the page
-		return []
+		if response.lower() == 'n':
+			print('skipped')
+			# not schedule A, skip the page
+			return []
 
 	# use '|' separator to avoid characters in asset names
-	response = send_to_api(message="get the asset names in the image. return them in a | separated list only. no other commentary.", 
+	response = send_to_api(message="get the asset names in the public disclosure form. return them in a | separated list only. no other commentary.", 
 						   base64_image=base64_image,
 						   model='gpt-4o')
 	asset_list = [response.strip() for response in response.split("|")]