New Features
Agents & Assistants
A better Agents & Assistants support in Bolt is now available!
While you already can implement your agents using app.event(...)
listeners for assistant_thread_started
, assistant_thread_context_changed
, and message
events, Bolt offers a simpler approach. You just need to create an Assistant
instance, attach the needed event handlers to it, and then add the assistant to your App
instance.
assistant = Assistant()
# This listener is invoked when a human user opened an assistant thread
@assistant.thread_started
def start_assistant_thread(say: Say, set_suggested_prompts: SetSuggestedPrompts):
# Send the first reply to the human who started chat with your app's assistant bot
say(":wave: Hi, how can I help you today?")
# Setting suggested prompts is optional
set_suggested_prompts(
prompts=[
# If the suggested prompt is long, you can use {"title": "short one to display", "message": "full prompt"} instead
"What does SLACK stand for?",
"When Slack was released?",
],
)
# This listener is invoked when the human user sends a reply in the assistant thread
@assistant.user_message
def respond_in_assistant_thread(
payload: dict,
logger: logging.Logger,
context: BoltContext,
set_status: SetStatus,
say: Say,
client: WebClient,
):
try:
# Tell the human user the assistant bot acknowledges the request and is working on it
set_status("is typing...")
# Collect the conversation history with this user
replies_in_thread = client.conversations_replies(
channel=context.channel_id,
ts=context.thread_ts,
oldest=context.thread_ts,
limit=10,
)
messages_in_thread: List[Dict[str, str]] = []
for message in replies_in_thread["messages"]:
role = "user" if message.get("bot_id") is None else "assistant"
messages_in_thread.append({"role": role, "content": message["text"]})
# Pass the latest prompt and chat history to the LLM (call_llm is your own code)
returned_message = call_llm(messages_in_thread)
# Post the result in the assistant thread
say(text=returned_message)
except Exception as e:
logger.exception(f"Failed to respond to an inquiry: {e}")
# Don't forget sending a message telling the error
# Without this, the status 'is typing...' won't be cleared, therefore the end-user is unable to continue the chat
say(f":warning: Sorry, something went wrong during processing your request (error: {e})")
# Enable this assistant middleware in your Bolt app
app.use(assistant)
Please refer to https://tools.slack.dev/bolt-python/concepts/assistant/ and https://github.com/slack-samples/bolt-python-assistant-template for more details.
Changes
- #1162 Add Agents & Assistants support - Thanks @seratch
- #1142 Add listener_runner to context object to enable developers to leverage lazy listeners in middleware - Thanks @seratch
- #1170 Fix double quoted img alt text in the default OAuth page rendering - @toofishes
- #1173 Expose
auto_acknowledge
option for custom function handlers - Thanks @WilliamBergamin - #1143 Remove
Optional
typing ofcontext.client
- Thanks @WilliamBergamin - #1164 Simplify Python code snippets in authorization.md - Thanks @arkid15r
References
- Release Milestone: https://github.com/slackapi/bolt-python/milestone/86?closed=1
- All Diff: v1.20.1...v1.21.0