Search: use ChatGPT 3.5 for summaries

src/semantic_search/semantic_search/external_services/openai.py

@@ -22,7 +22,7 @@ def create_embeddings(texts: List[str]) -> List[Any]:
 
 
 @retry(delay=3, backoff=2, tries=8)
-def gpt_query_json(query: str) -> str:
+def query_chat_gpt_forcing_json(query: str) -> str:
     summary = openai.ChatCompletion.create(
         model="gpt-4-1106-preview",
         messages=[{"role": "user", "content": query}],
@@ -32,7 +32,7 @@ def gpt_query_json(query: str) -> str:
 
 
 @retry(delay=3, backoff=2, tries=8)
-def gpt_query(query: str) -> str:
+def query_chat_gpt(query: str) -> str:
     summary = openai.ChatCompletion.create(
         model="gpt-4-1106-preview",
         messages=[{"role": "user", "content": query}],
@@ -41,9 +41,25 @@ def gpt_query(query: str) -> str:
 
 
 @retry(delay=3, backoff=2, tries=8)
-def gpt_summarize_thread(thread_messages: List[str]) -> str:
+def query_chat_gpt_3_5(query: str) -> str:
+    summary = openai.ChatCompletion.create(
+        model="gpt-3.5-turbo",
+        messages=[{"role": "user", "content": query}],
+    )
+    return summary.choices[0].message.content.strip()
+
+
+def summarize_thread_with_chat_gpt(thread_messages: List[str]) -> str:
+    text = "\n".join(thread_messages)
+    return query_chat_gpt(
+        "Summarize the following conversation and include channel information and usernames and actual names of the "
+        "author of the message: " + text
+    )
+
+
+def summarize_thread_with_chat_gpt_3_5(thread_messages: List[str]) -> str:
     text = "\n".join(thread_messages)
-    return gpt_query(
+    return query_chat_gpt_3_5(
         "Summarize the following conversation and include channel information and usernames and actual names of the "
         "author of the message: " + text
     )

src/semantic_search/semantic_search/load_messages.py

@@ -4,7 +4,7 @@
 
 from .config import CONTEXT_LENGTH
 from .external_services.pinecone import get_pinecone_index
-from .external_services.openai import create_embeddings, gpt_summarize_thread
+from .external_services.openai import create_embeddings, summarize_thread_with_chat_gpt_3_5
 import datetime
 from .external_services.slack_api import fetch_thread_messages, fetch_channel_messages, is_thread, \
     is_actual_message, \
@@ -152,7 +152,7 @@ def index_messages(channel_id, messages, start_from, pinecone_index, pinecone_na
 
             try:
                 logging.info(f"  - Summarizing thread {message['thread_ts']}")
-                summary = gpt_summarize_thread(raw_messages_for_summary)
+                summary = summarize_thread_with_chat_gpt_3_5(raw_messages_for_summary)
                 messages_for_embedding.append(thread_header.convert_to_summary(summary))
             except:
                 logging.info(f"  - Failed to summarize - {message['thread_ts']}")

src/semantic_search/semantic_search/query.py

@@ -4,7 +4,7 @@
 import uuid
 from datetime import date
 from .external_services.pinecone import get_pinecone_index
-from .external_services.openai import create_embedding, gpt_query_json
+from .external_services.openai import create_embedding, query_chat_gpt_forcing_json
 
 
 def build_slack_message_link(workspace_name, channel_id, message_timestamp, thread_timestamp=None):
@@ -91,7 +91,7 @@ def smart_query(namespace, query, username: str):
 
     gpt_response = None
     try:
-        gpt_response = gpt_query_json(prompt)
+        gpt_response = query_chat_gpt_forcing_json(prompt)
         result = json.loads(gpt_response)
         gpt_request_time = time.perf_counter() - gpt_request_start_time
         total_time = time.perf_counter() - smart_query_start_time