Update email database for July 2025

app.py

@@ -21,12 +21,16 @@ from graph import BuildGraph
 load_dotenv(dotenv_path=".env", override=True)
 
 # Download model snapshots from Hugging Face Hub
-print(f"Downloading checkpoints for {model_id}...")
-ckpt_dir = snapshot_download(model_id, local_dir_use_symlinks=False)
-print(f"Using checkpoints from {ckpt_dir}")
-print(f"Downloading checkpoints for {embedding_model_id}...")
-embedding_ckpt_dir = snapshot_download(embedding_model_id, local_dir_use_symlinks=False)
-print(f"Using embedding checkpoints from {embedding_ckpt_dir}")
+if torch.cuda.is_available():
+    print(f"Downloading checkpoints for {model_id}...")
+    ckpt_dir = snapshot_download(model_id, local_dir_use_symlinks=False)
+    print(f"Using checkpoints from {ckpt_dir}")
+    print(f"Downloading checkpoints for {embedding_model_id}...")
+    embedding_ckpt_dir = snapshot_download(embedding_model_id, local_dir_use_symlinks=False)
+    print(f"Using embedding checkpoints from {embedding_ckpt_dir}")
+else:
+    ckpt_dir = None
+    embedding_ckpt_dir = None
 
 # Download and extract data if data directory is not present
 if not os.path.isdir(db_dir):

data.py

@@ -45,7 +45,7 @@ def download_data():
 
     if not os.path.exists("db.zip"):
         # For S3 (need AWS_ACCESS_KEY_ID and AWS_ACCESS_KEY_SECRET)
-        download_file_from_bucket("r-help-chat", "db.zip", "db.zip")
+        download_file_from_bucket("r-help-chat", "db_20250801.zip", "db.zip")
         ## For Dropbox (shared file - key is in URL)
         # shared_link = "https://www.dropbox.com/scl/fi/jx90g5lorpgkkyyzeurtc/db.zip?rlkey=wvqa3p9hdy4rmod1r8yf2am09&st=l9tsam56&dl=0"
         # output_filename = "db.zip"

mods/langchain_chroma.py

@@ -470,9 +470,6 @@ class Chroma(VectorStore):
 
         See more: https://docs.trychroma.com/reference/py-collection#query
         """
-        # Possible fix for ValueError('Could not connect to tenant default_tenant. Are you sure it exists?')
-        # https://github.com/langchain-ai/langchain/issues/26884
-        chromadb.api.client.SharedSystemClient.clear_system_cache()
         return self._collection.query(
             query_texts=query_texts,
             query_embeddings=query_embeddings,  # type: ignore[arg-type]
@@ -481,6 +478,9 @@ class Chroma(VectorStore):
             where_document=where_document,  # type: ignore[arg-type]
             **kwargs,
         )
+        # Possible fix for ValueError('Could not connect to tenant default_tenant. Are you sure it exists?')
+        # https://github.com/langchain-ai/langchain/issues/26884
+        chromadb.api.client.SharedSystemClient.clear_system_cache()
 
     @staticmethod
     def encode_image(uri: str) -> str: