Better error handling (#25)

* Log errors
* return an error message when slack encounters an issue
* patch openai[embeddings] for now

app.py

@@ -1,15 +1,20 @@
+import logging
 import os
 
 from slack_bolt import App
 
 from buster.chatbot import Chatbot, ChatbotConfig
 
+logger = logging.getLogger(__name__)
+logging.basicConfig(level=logging.INFO)
+
+# Set Slack channel IDs
 MILA_CLUSTER_CHANNEL = "C04LR4H9KQA"
 ORION_CHANNEL = "C04LYHGUYB0"
 PYTORCH_CHANNEL = "C04MEK6N882"
 HF_TRANSFORMERS_CHANNEL = "C04NJNCJWHE"
 
-buster_cfg = ChatbotConfig(
+mila_doc_cfg = ChatbotConfig(
     documents_file="buster/data/document_embeddings.csv",
     unknown_prompt="This doesn't seem to be related to cluster usage.",
     embedding_model="text-embedding-ada-002",
@@ -43,7 +48,7 @@ buster_cfg = ChatbotConfig(
     Now answer the following question:
     """,
 )
-buster_chatbot = Chatbot(buster_cfg)
+mila_doc_chatbot = Chatbot(mila_doc_cfg)
 
 orion_cfg = ChatbotConfig(
     documents_file="buster/data/document_embeddings_orion.csv",
@@ -144,32 +149,38 @@ hf_transformers_cfg = ChatbotConfig(
 )
 hf_transformers_chatbot = Chatbot(hf_transformers_cfg)
 
+# TODO: eventually move this to a factory of sorts
+# Put all the bots in a dict by channel
+channel_id_to_bot = {
+    MILA_CLUSTER_CHANNEL: mila_doc_chatbot,
+    ORION_CHANNEL: orion_chatbot,
+    PYTORCH_CHANNEL: pytorch_chatbot,
+    HF_TRANSFORMERS_CHANNEL: hf_transformers_chatbot,
+}
+
 app = App(token=os.environ.get("SLACK_BOT_TOKEN"), signing_secret=os.environ.get("SLACK_SIGNING_SECRET"))
 
 
 @app.event("app_mention")
 def respond_to_question(event, say):
-    print(event)
+    logger.info(event)
 
     # user's text
     text = event["text"]
-    channel = event["channel"]
-
-    if channel == MILA_CLUSTER_CHANNEL:
-        print("*******using BUSTER********")
-        answer = buster_chatbot.process_input(text)
-    elif channel == ORION_CHANNEL:
-        print("*******using ORION********")
-        answer = orion_chatbot.process_input(text)
-    elif channel == PYTORCH_CHANNEL:
-        print("*******using PYTORCH********")
-        answer = pytorch_chatbot.process_input(text)
-    elif channel == HF_TRANSFORMERS_CHANNEL:
-        print("*******using HF TRANSFORMERS********")
-        answer = hf_transformers_chatbot.process_input(text)
-    else:
-        print(f"invalid channel: {channel}")
-        answer = "I was not yet implemented to support this channel."
+    channel_id = event["channel"]
+
+    try:
+        chatbot = channel_id_to_bot.get(channel_id)
+        if chatbot:
+            answer = chatbot.process_input(text)
+        else:
+            answer = "I was not yet implemented to support this channel."
+    except ValueError as e:
+        # log the error and return a generic response instead.
+        import traceback
+
+        logging.error(traceback.format_exc())
+        answer = "Oops, something went wrong. Try again later!"
 
     # responds to the message in the thread
     thread_ts = event["event_ts"]

buster/chatbot.py

@@ -112,8 +112,9 @@ class Chatbot:
             # log the error and return a generic response instead.
             import traceback
 
+            logger.error("Error connecting to OpenAI API")
             logging.error(traceback.format_exc())
-            response_text = "Oops, something went wrong. Try again later!"
+            response_text = "Hmm, we're having trouble connecting to OpenAI right now... Try again soon!"
             return response_text
 
     def add_sources(self, response: str, matched_documents: pd.DataFrame):

requirements.txt

@@ -2,9 +2,16 @@ bs4
 matplotlib
 numpy
 pandas
-plotly
-scikit-learn
 tabulate
 tenacity
 tiktoken
-openai[embeddings]
+openai
+
+# all openai[embeddings] deps, their list breaks our CI, see: https://github.com/openai/openai-python/issues/210 
+
+scikit-learn >= 1.0.2  # Needed for embedding utils, versions >= 1.1 require python 3.8
+tenacity >= 8.0.1
+matplotlib
+plotly
+pandas-stubs >= 1.1.0.11  # Needed for type hints for mypy
+openpyxl >= 3.0.7  # Needed for CLI fine-tuning data preparation tool xlsx format