bump gpt-index, better async query support

gpt3discord.py

@@ -32,7 +32,7 @@ from services.environment_service import EnvService
 from models.openai_model import Model
 
 
-__version__ = "10.9.15"
+__version__ = "10.9.16"
 
 
 PID_FILE = Path("bot.pid")

models/index_model.py

@@ -20,6 +20,7 @@ from discord.ext import pages
 from langchain.llms import OpenAIChat
 from llama_index.langchain_helpers.chatgpt import ChatGPTLLMPredictor
 from langchain import OpenAI
+from llama_index.optimization.optimizer import SentenceEmbeddingOptimizer
 from llama_index.prompts.chat_prompts import CHAT_REFINE_PROMPT
 
 from llama_index.readers import YoutubeTranscriptReader
@@ -55,7 +56,7 @@ RemoteReader = download_loader("RemoteReader")
 RemoteDepthReader = download_loader("RemoteDepthReader")
 
 
-def get_and_query(
+async def get_and_query(
     user_id,
     index_storage,
     query,
@@ -69,23 +70,23 @@ def get_and_query(
         user_id
     ].get_index_or_throw()
     if isinstance(index, GPTTreeIndex):
-        response = index.query(
+        response = await index.aquery(
             query,
             child_branch_factor=child_branch_factor,
             llm_predictor=llm_predictor,
             refine_template=CHAT_REFINE_PROMPT,
             embed_model=embed_model,
-            use_async=True,
+            #optimizer=SentenceEmbeddingOptimizer(threshold_cutoff=0.7)
         )
     else:
-        response = index.query(
+        response = await index.aquery(
             query,
             response_mode=response_mode,
             llm_predictor=llm_predictor,
             embed_model=embed_model,
             similarity_top_k=nodes,
             refine_template=CHAT_REFINE_PROMPT,
-            use_async=True,
+            #optimizer=SentenceEmbeddingOptimizer(threshold_cutoff=0.7)
         )
     return response
 
@@ -921,20 +922,7 @@ class Index_handler:
         try:
             embedding_model = OpenAIEmbedding()
             embedding_model.last_token_usage = 0
-            response = await self.loop.run_in_executor(
-                None,
-                partial(
-                    get_and_query,
-                    ctx.user.id,
-                    self.index_storage,
-                    query,
-                    response_mode,
-                    nodes,
-                    self.llm_predictor,
-                    embedding_model,
-                    child_branch_factor,
-                ),
-            )
+            response = await get_and_query(ctx.user.id, self.index_storage, query, response_mode, nodes, self.llm_predictor, embedding_model, child_branch_factor)
             print("The last token usage was ", self.llm_predictor.last_token_usage)
             await self.usage_service.update_usage(
                 self.llm_predictor.last_token_usage, chatgpt=True

models/search_model.py

@@ -455,36 +455,9 @@ class Search:
         embedding_model.last_token_usage = 0
 
         if not deep:
-            response = await self.loop.run_in_executor(
-                None,
-                partial(
-                    index.query,
-                    query,
-                    embed_model=embedding_model,
-                    llm_predictor=llm_predictor,
-                    refine_template=CHAT_REFINE_PROMPT,
-                    similarity_top_k=nodes or DEFAULT_SEARCH_NODES,
-                    text_qa_template=self.qaprompt,
-                    use_async=True,
-                    response_mode=response_mode,
-                ),
-            )
+            response = await index.aquery(query, embed_model=embedding_model, llm_predictor=llm_predictor, refine_template=CHAT_REFINE_PROMPT, similarity_top_k=nodes or DEFAULT_SEARCH_NODES, text_qa_template=self.qaprompt,  response_mode=response_mode)
         else:
-            response = await self.loop.run_in_executor(
-                None,
-                partial(
-                    index.query,
-                    query,
-                    embedding_mode="hybrid",
-                    llm_predictor=llm_predictor,
-                    refine_template=CHAT_REFINE_PROMPT,
-                    include_text=True,
-                    embed_model=embedding_model,
-                    use_async=True,
-                    similarity_top_k=nodes or DEFAULT_SEARCH_NODES,
-                    response_mode=response_mode,
-                ),
-            )
+            response = await index.aquery(query, embed_model=embedding_model, llm_predictor=llm_predictor, refine_template=CHAT_REFINE_PROMPT, similarity_top_k=nodes or DEFAULT_SEARCH_NODES, text_qa_template=self.qaprompt, response_mode=response_mode)
 
         await self.usage_service.update_usage(
             llm_predictor.last_token_usage, chatgpt=True

pyproject.toml

@@ -32,7 +32,7 @@ dependencies = [
 "sqlitedict==2.1.0",
 "backoff==2.2.1",
 "flask==2.2.3",
-"llama-index==0.4.23",
+"llama-index==0.4.26",
 "PyPDF2==3.0.1",
 "youtube_transcript_api==0.5.0",
 "sentencepiece==0.1.97",

requirements.txt

@@ -12,7 +12,7 @@ pinecone-client==2.1.0
 sqlitedict==2.1.0
 backoff==2.2.1
 flask==2.2.3
-llama-index==0.4.23
+llama-index==0.4.26
 PyPDF2==3.0.1
 youtube_transcript_api==0.5.0
 sentencepiece==0.1.97

requirements_base.txt

@@ -12,7 +12,7 @@ pinecone-client==2.1.0
 sqlitedict==2.1.0
 backoff==2.2.1
 flask==2.2.3
-llama-index==0.4.23
+llama-index==0.4.26
 PyPDF2==3.0.1
 youtube_transcript_api==0.5.0
 sentencepiece==0.1.97