Merge pull request #65 from TogetherCrew/feat/upgrade-llama-index-dep…

…endency

feat: llama-index code migration!

dags/hivemind_discord_etl.py

@@ -0,0 +1,67 @@
+import logging
+from datetime import datetime
+
+from airflow import DAG
+from airflow.decorators import task
+from dotenv import load_dotenv
+from hivemind_etl_helpers.discord_mongo_summary_etl import process_discord_summaries
+from hivemind_etl_helpers.discord_mongo_vector_store_etl import (
+    process_discord_guild_mongo,
+)
+from hivemind_etl_helpers.src.utils.mongo_discord_communities import (
+    get_all_discord_communities,
+)
+
+with DAG(
+    dag_id="discord_vector_store_update",
+    start_date=datetime(2024, 1, 1),
+    schedule_interval="0 2 * * *",
+    catchup=False,
+) as dag:
+
+    @task
+    def get_discord_communities() -> list[str]:
+        """
+        Getting all communities having discord from database
+        """
+        communities = get_all_discord_communities()
+        return communities
+
+    @task
+    def start_discord_vectorstore(community_id: str):
+        load_dotenv()
+        logging.info(f"Working on community, {community_id}")
+        process_discord_guild_mongo(community_id=community_id)
+        logging.info(f"Community {community_id} Job finished!")
+
+    communities = get_discord_communities()
+    # `start_discord_vectorstore` will be called multiple times
+    # with the length of the list
+    start_discord_vectorstore.expand(community_id=communities)
+
+
+with DAG(
+    dag_id="discord_summary_vector_store",
+    start_date=datetime(2024, 1, 1),
+    schedule_interval="0 2 * * *",
+) as dag:
+
+    @task
+    def get_mongo_discord_communities() -> list[str]:
+        """
+        Getting all communities having discord from database
+        this function is the same with `get_discord_communities`
+        we just changed the name for the pylint
+        """
+        communities = get_all_discord_communities()
+        return communities
+
+    @task
+    def start_discord_summary_vectorstore(community_id: str):
+        load_dotenv()
+        logging.info(f"Working on community, {community_id}")
+        process_discord_summaries(community_id=community_id, verbose=False)
+        logging.info(f"Community {community_id} Job finished!")
+
+    communities = get_mongo_discord_communities()
+    start_discord_summary_vectorstore.expand(community_id=communities)

dags/hivemind_discourse_etl.py

@@ -0,0 +1,57 @@
+import logging
+from datetime import datetime
+
+from airflow import DAG
+from airflow.decorators import task
+from hivemind_etl_helpers.discourse_summary_etl import process_discourse_summary
+from hivemind_etl_helpers.discourse_vectorstore_etl import process_discourse_vectorstore
+from hivemind_etl_helpers.src.utils.get_communities_data import (
+    get_discourse_communities,
+)
+
+with DAG(
+    dag_id="discourse_vector_store",
+    start_date=datetime(2024, 3, 1),
+    schedule_interval="0 2 * * *",
+) as dag:
+
+    @task
+    def process_discourse_community(community_information: dict[str, str | datetime]):
+        community_id = community_information["community_id"]
+        forum_endpoint = community_information["endpoint"]
+        from_date = community_information["from_date"]
+
+        logging.info(f"Starting Discourse ETL | community_id: {community_id}")
+        process_discourse_vectorstore(
+            community_id=community_id,
+            forum_endpoint=forum_endpoint,
+            from_starting_date=from_date,
+        )
+
+    communities_info = get_discourse_communities()
+    process_discourse_community.expand(community_information=communities_info)
+
+
+with DAG(
+    dag_id="discourse_summary_vector_store",
+    start_date=datetime(2024, 2, 21),
+    schedule_interval="0 2 * * *",
+) as dag:
+
+    @task
+    def process_discourse_community_summary(
+        community_information: dict[str, str | datetime]
+    ):
+        community_id = community_information["community_id"]
+        forum_endpoint = community_information["endpoint"]
+        from_date = community_information["from_date"]
+
+        logging.info(f"Starting Discourse ETL | community_id: {community_id}")
+        process_discourse_summary(
+            community_id=community_id,
+            forum_endpoint=forum_endpoint,
+            from_starting_date=from_date,
+        )
+
+    communities_info = get_discourse_communities()
+    process_discourse_community_summary.expand(community_information=communities_info)

dags/hivemind_etl_helpers/discord_mongo_summary_etl.py

@@ -10,7 +10,9 @@
 )
 from hivemind_etl_helpers.src.document_node_parser import configure_node_parser
 from hivemind_etl_helpers.src.utils.sort_summary_docs import sort_summaries_daily
-from llama_index.response_synthesizers import get_response_synthesizer
+from llama_index.core import Settings
+from llama_index.core.response_synthesizers import get_response_synthesizer
+from llama_index.llms.openai import OpenAI
 from tc_hivemind_backend.db.pg_db_utils import setup_db
 from tc_hivemind_backend.db.utils.model_hyperparams import load_model_hyperparams
 from tc_hivemind_backend.embeddings.cohere import CohereEmbedding
@@ -93,16 +95,16 @@ def process_discord_summaries(community_id: str, verbose: bool = False) -> None:
     node_parser = configure_node_parser(chunk_size=chunk_size)
     pg_vector = PGVectorAccess(table_name=table_name, dbname=dbname)
 
-    embed_model = CohereEmbedding()
+    Settings.node_parser = node_parser
+    Settings.embed_model = CohereEmbedding()
+    Settings.chunk_size = chunk_size
+    Settings.llm = OpenAI(model="gpt-3.5-turbo")
 
     pg_vector.save_documents_in_batches(
         community_id=community_id,
         documents=docs_daily_sorted,
         batch_size=100,
-        node_parser=node_parser,
         max_request_per_minute=None,
-        embed_model=embed_model,
-        embed_dim=embedding_dim,
         request_per_minute=10000,
         deletion_query=deletion_query,
     )

dags/hivemind_etl_helpers/discord_mongo_vector_store_etl.py

@@ -1,4 +1,3 @@
-import argparse
 import logging
 from datetime import timedelta
 
@@ -9,6 +8,8 @@
     find_guild_id_by_community_id,
 )
 from hivemind_etl_helpers.src.document_node_parser import configure_node_parser
+from llama_index.core import Settings
+from llama_index.llms.openai import OpenAI
 from tc_hivemind_backend.db.pg_db_utils import setup_db
 from tc_hivemind_backend.db.utils.model_hyperparams import load_model_hyperparams
 from tc_hivemind_backend.embeddings.cohere import CohereEmbedding
@@ -52,27 +53,16 @@ def process_discord_guild_mongo(community_id: str) -> None:
     node_parser = configure_node_parser(chunk_size=chunk_size)
     pg_vector = PGVectorAccess(table_name=table_name, dbname=dbname)
 
-    embed_model = CohereEmbedding()
+    Settings.node_parser = node_parser
+    Settings.embed_model = CohereEmbedding()
+    Settings.chunk_size = chunk_size
+    Settings.llm = OpenAI(model="gpt-3.5-turbo")
 
     pg_vector.save_documents_in_batches(
         community_id=community_id,
         documents=documents,
         batch_size=100,
         node_parser=node_parser,
         max_request_per_minute=None,
-        embed_model=embed_model,
-        embed_dim=embedding_dim,
         request_per_minute=10000,
-        # max_request_per_day=REQUEST_PER_DAY,
     )
-
-
-if __name__ == "__main__":
-    logging.basicConfig(level=logging.INFO)
-    parser = argparse.ArgumentParser()
-    parser.add_argument(
-        "community_id", type=str, help="the Community that the guild is related to"
-    )
-    args = parser.parse_args()
-
-    process_discord_guild_mongo(community_id=args.community_id)

dags/hivemind_etl_helpers/discourse_summary_etl.py

@@ -10,8 +10,9 @@
 from hivemind_etl_helpers.src.db.discourse.utils.get_forums import get_forum_uuid
 from hivemind_etl_helpers.src.document_node_parser import configure_node_parser
 from hivemind_etl_helpers.src.utils.sort_summary_docs import sort_summaries_daily
-from llama_index import Document
-from llama_index.response_synthesizers import get_response_synthesizer
+from llama_index.core import Document, Settings
+from llama_index.core.response_synthesizers import get_response_synthesizer
+from llama_index.llms.openai import OpenAI
 from neo4j._data import Record
 from tc_hivemind_backend.db.pg_db_utils import setup_db
 from tc_hivemind_backend.db.utils.model_hyperparams import load_model_hyperparams
@@ -133,8 +134,6 @@ def process_forum(
         node_parser = configure_node_parser(chunk_size=chunk_size)
         pg_vector = PGVectorAccess(table_name=table_name, dbname=dbname)
 
-        embed_model = CohereEmbedding()
-
         sorted_daily_docs = sort_summaries_daily(
             level1_docs=topic_summary_documents,
             level2_docs=category_summary_documenets,
@@ -145,13 +144,16 @@ def process_forum(
             f"{log_prefix} Saving discourse summaries (extracting the embedding and saving)"
         )
 
+        Settings.node_parser = node_parser
+        Settings.embed_model = CohereEmbedding()
+        Settings.chunk_size = chunk_size
+        Settings.llm = OpenAI(model="gpt-3.5-turbo")
+
         pg_vector.save_documents_in_batches(
             community_id=community_id,
             documents=sorted_daily_docs,
             batch_size=100,
-            node_parser=node_parser,
             max_request_per_minute=None,
-            embed_model=embed_model,
             embed_dim=embedding_dim,
             request_per_minute=10000,
             deletion_query=deletion_query,