Chatbot improvements (#188)

Author: miguelgrinberg

example-apps/chatbot-rag-app/api/chat.py

@@ -38,17 +38,21 @@ def ask_question(question, session_id):
         condense_question_prompt = render_template(
             'condense_question_prompt.txt', question=question,
             chat_history=chat_history.messages)
-        question = get_llm().invoke(condense_question_prompt).content
+        condensed_question = get_llm().invoke(condense_question_prompt).content
+    else:
+        condensed_question = question
 
+    current_app.logger.debug('Condensed question: %s', condensed_question)
     current_app.logger.debug('Question: %s', question)
 
-    docs = store.as_retriever().invoke(question)
+    docs = store.as_retriever().invoke(condensed_question)
     for doc in docs:
         doc_source = {**doc.metadata, 'page_content': doc.page_content}
         current_app.logger.debug('Retrieved document passage from: %s', doc.metadata['name'])
         yield f'data: {SOURCE_TAG} {json.dumps(doc_source)}\n\n'
 
-    qa_prompt = render_template('rag_prompt.txt', question=question, docs=docs)
+    qa_prompt = render_template('rag_prompt.txt', question=question, docs=docs,
+                                chat_history=chat_history.messages)
 
     answer = ''
     for chunk in get_llm().stream(qa_prompt):

example-apps/chatbot-rag-app/api/templates/rag_prompt.txt

@@ -1,4 +1,4 @@
-Use the following passages to answer the user's question. 
+Use the following passages and chat history to answer the user's question. 
 Each passage has a NAME which is the title of the document. After your answer, leave a blank line and then give the source name of the passages you answered from. Put them in a comma separated list, prefixed with SOURCES:.
 
 Example:
@@ -22,5 +22,10 @@ PASSAGE:
 
 {% endfor -%}
 ----
+Chat history:
+{% for dialogue_turn in chat_history -%}
+{% if dialogue_turn.type == 'human' %}Question: {{ dialogue_turn.content }}{% elif dialogue_turn.type == 'ai' %}Response: {{ dialogue_turn.content }}{% endif %}
+{% endfor -%}
+
 Question: {{ question }}
 Response:

example-apps/chatbot-rag-app/data/index_data.py

@@ -91,6 +91,9 @@ def main():
         es_connection=elasticsearch_client,
         index_name=INDEX,
         strategy=ElasticsearchStore.SparseVectorRetrievalStrategy(model_id=ELSER_MODEL),
+        bulk_kwargs={
+            'request_timeout': 60,
+        },
     )
 
 

example-apps/chatbot-rag-app/frontend/src/store/provider.tsx

@@ -238,7 +238,7 @@ export const thunkActions = {
               dispatch(
                 actions.updateMessage({
                   id: conversationId,
-                  content: message.replace(/SOURCES: (.+)+/, ''),
+                  content: message.replace(/SOURCES:(.+)*/, ''),
                 })
               )
             }
@@ -300,8 +300,8 @@ const parseSources = (
   message: string
 ) => {
   message = message.replaceAll("\"", "");
-  const match = message.match(/SOURCES: (.+)+/)
-  if (match) {
+  const match = message.match(/SOURCES:(.+)*/)
+  if (match && match[1]) {
     return match[1].split(',').map(element => {
       return element.trim();
     });