Audit trail

src/filler.py

@@ -1,7 +1,7 @@
 from pdfrw import PdfReader, PdfWriter
 from src.llm import LLM
 from datetime import datetime
-
+import os
 
 class Filler:
     def __init__(self):
@@ -23,7 +23,14 @@ def fill_form(self, pdf_form: str, llm: LLM):
         t2j = llm.main_loop()
         textbox_answers = t2j.get_data()  # This is a dictionary
 
-        answers_list = list(textbox_answers.values())
+        answers_list = []
+        for data in textbox_answers.values():
+            if isinstance(data, dict):
+                answers_list.append(data.get("value", ""))
+            elif isinstance(data, list) and len(data) > 0:
+                answers_list.append(data[0].get("value", ""))
+            else:
+                answers_list.append(str(data))
 
         # Read PDF
         pdf = PdfReader(pdf_form)
@@ -45,8 +52,37 @@ def fill_form(self, pdf_form: str, llm: LLM):
                         else:
                             # Stop if we run out of answers
                             break
-
+        
         PdfWriter().write(output_pdf, pdf)
 
+        # --- ZERO-DEPENDENCY AUDIT TRAIL ---
+        # Create a text file with the same name as the PDF
+        audit_txt_path = output_pdf.replace(".pdf", "_audit.txt")
+
+        with open(audit_txt_path, "w", encoding="utf-8") as f:
+            f.write("="*60 + "\n")
+            f.write("FIREFORM AI DATA EXTRACTION AUDIT TRAIL\n")
+            f.write(f"Generated: {datetime.now().strftime('%Y-%m-%d %H:%M:%S')}\n")
+            f.write("="*60 + "\n\n")
+
+            for field, data in textbox_answers.items():
+                # Parse out the value and quote safely
+                if isinstance(data, dict):
+                    val = data.get("value", "N/A")
+                    quote = data.get("quote", "N/A")
+                elif isinstance(data, list) and len(data) > 0:
+                    val = data[0].get("value", "N/A")
+                    quote = data[0].get("quote", "N/A")
+                else:
+                    val, quote = str(data), "N/A"
+
+                f.write(f"FIELD : {field}\n")
+                f.write(f"VALUE : {val}\n")
+                f.write(f"SOURCE: \"{quote}\"\n")
+                f.write("-" * 60 + "\n")
+
+        print(f"\t[LOG] Audit trail saved to: {audit_txt_path}")
+        # -----------------------------------
+
         # Your main.py expects this function to return the path
-        return output_pdf
+        return output_pdf

src/llm.py

@@ -30,11 +30,10 @@ def build_prompt(self, current_field):
         """
         prompt = f""" 
             SYSTEM PROMPT:
-            You are an AI assistant designed to help fillout json files with information extracted from transcribed voice recordings. 
-            You will receive the transcription, and the name of the JSON field whose value you have to identify in the context. Return 
-            only a single string containing the identified value for the JSON field. 
-            If the field name is plural, and you identify more than one possible value in the text, return both separated by a ";".
-            If you don't identify the value in the provided text, return "-1".
+            You are an AI assistant designed to extract information from transcribed voice recordings. 
+            You must return your answer STRICTLY as a valid JSON object with two keys: "value" and "quote".
+            - "value": The identified value for the field. If not found, use "-1". If plural, separate with ";".
+            - "quote": The exact sentence or phrase from the text that justifies your value. If not found, use "N/A".
             ---
             DATA:
             Target JSON field to find in text: {current_field}
@@ -46,7 +45,7 @@ def build_prompt(self, current_field):
 
     def main_loop(self):
         # self.type_check_all()
-        for field in self._target_fields.keys():
+        for field in self._target_fields:
             prompt = self.build_prompt(field)
             # print(prompt)
             # ollama_url = "http://localhost:11434/api/generate"
@@ -56,7 +55,8 @@ def main_loop(self):
             payload = {
                 "model": "mistral",
                 "prompt": prompt,
-                "stream": False,  # don't really know why --> look into this later.
+                "stream": False, # don't really know why --> look into this later.
+                "format": "json"
             }
 
             try:
@@ -72,9 +72,15 @@ def main_loop(self):
 
             # parse response
             json_data = response.json()
-            parsed_response = json_data["response"]
-            # print(parsed_response)
-            self.add_response_to_json(field, parsed_response)
+            try:
+                parsed_obj = json.loads(json_data["response"])
+                val = parsed_obj.get("value", "-1")
+                quote = parsed_obj.get("quote", "N/A")
+            except json.JSONDecodeError:
+                val = "-1"
+                quote = "JSON Parse Error"
+
+            self.add_response_to_json(field, val, quote)
 
         print("----------------------------------")
         print("\t[LOG] Resulting JSON created from the input text:")
@@ -83,11 +89,19 @@ def main_loop(self):
 
         return self
 
-    def add_response_to_json(self, field, value):
+    def add_response_to_json(self, field, value, quote="N/A"):
         """
         this method adds the following value under the specified field,
         or under a new field if the field doesn't exist, to the json dict
         """
+        # SAFETY CHECK: If the LLM returns a list, join it with semicolons
+        if isinstance(value, list):
+            value = ";".join([str(v) for v in value])
+        # If it returns a number or boolean, cast it to a string
+        elif not isinstance(value, str):
+            value = str(value)
+
+        # Now it's guaranteed to be a string
         value = value.strip().replace('"', "")
         parsed_value = None
 
@@ -96,11 +110,17 @@ def add_response_to_json(self, field, value):
 
         if ";" in value:
             parsed_value = self.handle_plural_values(value)
+
+        entry = {"value": parsed_value, "quote": quote}
 
         if field in self._json.keys():
-            self._json[field].append(parsed_value)
+            # If it's already a list, append. Otherwise convert to list.
+            if isinstance(self._json[field], list):
+                self._json[field].append(entry)
+            else:
+                self._json[field] = [self._json[field], entry]
         else:
-            self._json[field] = parsed_value
+            self._json[field] = entry
 
         return
 

src/main.py

@@ -3,6 +3,7 @@
 from commonforms import prepare_form 
 from pypdf import PdfReader
 from controller import Controller
+from typing import Union
 
 def input_fields(num_fields: int):
     fields = []
@@ -68,7 +69,7 @@ def run_pdf_fill_process(user_input: str, definitions: list, pdf_form_path: Unio
 if __name__ == "__main__":
     file = "./src/inputs/file.pdf"
     user_input = "Hi. The employee's name is John Doe. His job title is managing director. His department supervisor is Jane Doe. His phone number is 123456. His email is jdoe@ucsc.edu. The signature is <Mamañema>, and the date is 01/02/2005"
-    fields = ["Employee's name", "Employee's job title", "Employee's department supervisor", "Employee's phone number", "Employee's email", "Signature", "Date"]
+    descriptive_fields = ["Employee's name", "Employee's job title", "Employee's department supervisor", "Employee's phone number", "Employee's email", "Signature", "Date"]
     prepared_pdf = "temp_outfile.pdf"
     prepare_form(file, prepared_pdf)
 
@@ -80,4 +81,4 @@ def run_pdf_fill_process(user_input: str, definitions: list, pdf_form_path: Unio
         num_fields = 0
 
     controller = Controller()
-    controller.fill_form(user_input, fields, file)
+    controller.fill_form(user_input, descriptive_fields, file)