Sync updates from source repository

query.py

@@ -1,56 +1,6 @@
 import requests
 import json
-import re
-from urllib.parse import quote
-
-def extract_between_tags(text, start_tag, end_tag):
-    start_index = text.find(start_tag)
-    end_index = text.find(end_tag, start_index)
-    return text[start_index+len(start_tag):end_index-len(end_tag)]
-
-class CitationNormalizer():
-
-    def __init__(self, responses, docs):
-        self.docs = docs
-        self.responses = responses
-        self.refs = []
-
-    def normalize_citations(self, summary):
-        start_tag = "%START_SNIPPET%"
-        end_tag = "%END_SNIPPET%"
-
-        # find all references in the summary
-        pattern = r'\[\d{1,2}\]'
-        matches = [match.span() for match in re.finditer(pattern, summary)]
-
-        # figure out unique list of references
-        for match in matches:
-            start, end = match
-            response_num = int(summary[start+1:end-1])
-            doc_num = self.responses[response_num-1]['documentIndex']
-            metadata = {item['name']: item['value'] for item in self.docs[doc_num]['metadata']}
-            text = extract_between_tags(self.responses[response_num-1]['text'], start_tag, end_tag)
-            if 'url' in metadata.keys():
-                url = f"{metadata['url']}#:~:text={quote(text)}"
-                if url not in self.refs:
-                    self.refs.append(url)
-
-        # replace references with markdown links
-        refs_dict = {url:(inx+1) for inx,url in enumerate(self.refs)}
-        for match in reversed(matches):
-            start, end = match
-            response_num = int(summary[start+1:end-1])
-            doc_num = self.responses[response_num-1]['documentIndex']
-            metadata = {item['name']: item['value'] for item in self.docs[doc_num]['metadata']}
-            text = extract_between_tags(self.responses[response_num-1]['text'], start_tag, end_tag)
-            if 'url' in metadata.keys():
-                url = f"{metadata['url']}#:~:text={quote(text)}"
-                citation_inx = refs_dict[url]
-                summary = summary[:start] + f'[\[{citation_inx}\]]({url})' + summary[end:]
-            else:
-                summary = summary[:start] + summary[end:]
 
-        return summary
 
 class VectaraQuery():
     def __init__(self, api_key: str, customer_id: str, corpus_ids: list[str], prompt_name: str = None):
@@ -82,9 +32,6 @@ class VectaraQuery():
                     'rerankingConfig':
                     {
                         'rerankerId': 272725719,
-                        'mmrConfig': {
-                            'diversityBias': 0.3
-                        }
                     },
                     'summary': [
                         {
@@ -95,11 +42,15 @@ class VectaraQuery():
                                 'store': True,
                                 'conversationId': self.conv_id
                             },
+                            'citationParams': {
+                                "style": "NONE",
+                            }
                         }
                     ]
                 } 
             ]
         }
+    
 
     def get_headers(self):
         return {
@@ -122,10 +73,10 @@ class VectaraQuery():
 
         res = response.json()
 
-        top_k = 10
+        #top_k = 10
         summary = res['responseSet'][0]['summary'][0]['text']
-        responses = res['responseSet'][0]['response'][:top_k]
-        docs = res['responseSet'][0]['document']
+        #responses = res['responseSet'][0]['response'][:top_k]
+        #docs = res['responseSet'][0]['document']
         chat = res['responseSet'][0]['summary'][0].get('chat', None)
 
         if chat and chat['status'] is not None:
@@ -137,12 +88,11 @@ class VectaraQuery():
             return 'Sorry, something went wrong in my brain. Please try again later.'
         
         self.conv_id = chat['conversationId'] if chat else None
-        summary = CitationNormalizer(responses, docs).normalize_citations(summary)        
         return summary
 
     def submit_query_streaming(self, query_str: str):
 
-        endpoint = f"https://api.vectara.io/v1/stream-query"
+        endpoint = "https://api.vectara.io/v1/stream-query"
         body = self.get_body(query_str)
 
         response = requests.post(endpoint, data=json.dumps(body), verify=True, headers=self.get_headers(), stream=True) 
@@ -151,8 +101,6 @@ class VectaraQuery():
             return "Sorry, something went wrong in my brain. Please try again later."
 
         chunks = []
-        accumulated_text = ""  # Initialize text accumulation
-        pattern_max_length = 50  # Example heuristic
         for line in response.iter_lines():
             if line:  # filter out keep-alive new lines
                 data = json.loads(line.decode('utf-8'))
@@ -177,22 +125,10 @@ class VectaraQuery():
                             self.conv_id = conv_id
                         
                     chunk = summary['text']
-                    accumulated_text += chunk  # Append current chunk to accumulation
-                    if len(accumulated_text) > pattern_max_length:
-                        accumulated_text = re.sub(r"\[\d+\]", "", accumulated_text)
-                        accumulated_text = re.sub(r"\s+\.", ".", accumulated_text)
-                        out_chunk = accumulated_text[:-pattern_max_length]
-                        chunks.append(out_chunk)
-                        yield out_chunk
-                        accumulated_text = accumulated_text[-pattern_max_length:]
+                    chunks.append(chunk)
+                    yield chunk
 
                     if summary['done']:
                         break
-
-        # yield the last piece
-        if len(accumulated_text) > 0:
-            accumulated_text = re.sub(r" \[\d+\]\.", ".", accumulated_text)
-            chunks.append(accumulated_text)
-            yield accumulated_text        
         
         return ''.join(chunks)