Skip to content

Commit

Permalink
Add tone and voice to proofreading feature on home tab
Browse files Browse the repository at this point in the history
  • Loading branch information
@seratch
seratch committed Apr 16, 2024
1 parent 37b33cb commit 7fbc3e4
Show file tree
Hide file tree
Showing 2 changed files with 149 additions and 72 deletions.
202 changes: 136 additions & 66 deletions app/bolt_listeners.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -2,6 +2,7 @@
import logging
import re
import time
from typing import Optional

from openai.error import Timeout
from slack_bolt import App, Ack, BoltContext, BoltResponse
Expand Down Expand Up @@ -751,8 +752,26 @@ def prepare_and_share_thread_summary(
#


def build_proofreading_input_modal(prompt: str):
return {
def build_proofreading_input_modal(prompt: str, tone_and_voice: Optional[str]):
tone_and_voice_options = [
{"text": {"type": "plain_text", "text": persona}, "value": persona}
for persona in [
"Friendly and humble individual in Slack",
"Software developer discussing issues on GitHub",
"Engaging yet insightful social media poster",
"Customer service representative handling inquiries",
"Marketing manager creating a product launch script",
"Technical writer documenting software procedures",
"Product manager creating a roadmap",
"HR manager composing a job description",
"Public relations officer drafting statements",
"Scientific researcher publicizing findings",
"Travel blogger sharing experiences",
"Speechwriter crafting a persuasive speech",
]
]

modal = {
"type": "modal",
"callback_id": "proofread",
"title": {"type": "plain_text", "text": "Proofreading"},
Expand All @@ -774,14 +793,35 @@ def build_proofreading_input_modal(prompt: str):
"multiline": True,
},
},
{
"type": "input",
"block_id": "tone_and_voice",
"label": {"type": "plain_text", "text": "Tone and voice"},
"element": {
"type": "static_select",
"action_id": "input",
"options": tone_and_voice_options,
},
"optional": True,
},
],
}
if tone_and_voice is not None:
selected_option = {
"text": {"type": "plain_text", "text": tone_and_voice},
"value": tone_and_voice,
}
modal["blocks"][2]["element"]["initial_option"] = selected_option
else:
first_option = modal["blocks"][2]["element"]["options"][0]
modal["blocks"][2]["element"]["initial_option"] = first_option
return modal


def start_proofreading(client: WebClient, body: dict, payload: dict):
client.views_open(
trigger_id=body.get("trigger_id"),
view=build_proofreading_input_modal(payload.get("value")),
view=build_proofreading_input_modal(payload.get("value"), None),
)


Expand All @@ -792,33 +832,34 @@ def ack_proofreading_modal_submission(
):
original_text = extract_state_value(payload, "original_text").get("value")
text = "\n".join(map(lambda s: f">{s}", original_text.split("\n")))
ack(
response_action="update",
view={
"type": "modal",
"callback_id": "proofread",
"title": {"type": "plain_text", "text": "Proofreading"},
"close": {"type": "plain_text", "text": "Close"},
"private_metadata": payload["private_metadata"],
"blocks": [
{
"type": "context",
"elements": [
{
"type": "mrkdwn",
"text": f"Running OpenAI's *{context['OPENAI_MODEL']}* model:",
},
],
},
{
"type": "section",
"text": {
modal_view = {
"type": "modal",
"callback_id": "proofread",
"title": {"type": "plain_text", "text": "Proofreading"},
"close": {"type": "plain_text", "text": "Close"},
"private_metadata": payload["private_metadata"],
"blocks": [
{
"type": "context",
"elements": [
{
"type": "mrkdwn",
"text": f"{text}\n\nProofreading your input now ... :hourglass:",
"text": f"Running OpenAI's *{context['OPENAI_MODEL']}* model:",
},
],
},
{
"type": "section",
"text": {
"type": "mrkdwn",
"text": f"{text}\n\nProofreading your input now ... :hourglass:",
},
],
},
},
],
}
ack(
response_action="update",
view=modal_view,
)


Expand All @@ -832,55 +873,82 @@ def display_proofreading_result(
try:
openai_api_key = context.get("OPENAI_API_KEY")
original_text = extract_state_value(payload, "original_text").get("value")
tone_and_voice = extract_state_value(payload, "tone_and_voice")
tone_and_voice = (
tone_and_voice.get("selected_option").get("value")
if tone_and_voice.get("selected_option")
else None
)
text = "\n".join(map(lambda s: f">{s}", original_text.split("\n")))
result = generate_proofreading_result(
context=context,
logger=logger,
openai_api_key=openai_api_key,
original_text=original_text,
tone_and_voice=tone_and_voice,
timeout_seconds=OPENAI_TIMEOUT_SECONDS,
)
client.views_update(
view_id=payload["id"],
view={
"type": "modal",
"callback_id": "proofread-result",
"title": {"type": "plain_text", "text": "Proofreading"},
"submit": {"type": "plain_text", "text": "Try Another"},
"close": {"type": "plain_text", "text": "Close"},
"private_metadata": payload["private_metadata"],
"blocks": [
{
"type": "context",
"elements": [
{
"type": "mrkdwn",
"text": f"Provided using OpenAI's *{context['OPENAI_MODEL']}* model:",
},
],
},
{
"type": "section",
"text": {
private_metadata = payload["private_metadata"]
if tone_and_voice is not None:
pm = json.loads(payload["private_metadata"])
pm["tone_and_voice"] = tone_and_voice
private_metadata = json.dumps(pm)

modal_view = {
"type": "modal",
"callback_id": "proofread-result",
"title": {"type": "plain_text", "text": "Proofreading"},
"submit": {"type": "plain_text", "text": "Try Another"},
"close": {"type": "plain_text", "text": "Close"},
"private_metadata": private_metadata,
"blocks": [
{
"type": "context",
"elements": [
{
"type": "mrkdwn",
"text": f"{text}\n\n{result}",
"text": f"Provided using OpenAI's *{context['OPENAI_MODEL']}* model:",
},
],
},
{
"type": "section",
"text": {
"type": "mrkdwn",
"text": f"{text}\n\n{result}",
},
{
"type": "section",
"text": {"type": "mrkdwn", "text": " "},
"accessory": {
"type": "button",
"text": {
"type": "plain_text",
"text": "Send this result in DM",
},
"value": "clicked",
"action_id": "send-proofread-result-in-dm",
},
},
],
}
if tone_and_voice is not None:
modal_view["blocks"].append(
{
"type": "context",
"elements": [
{"type": "mrkdwn", "text": f"Tone and voice: {tone_and_voice}"}
],
}
)

modal_view["blocks"].append(
{
"type": "section",
"text": {"type": "mrkdwn", "text": " "},
"accessory": {
"type": "button",
"text": {
"type": "plain_text",
"text": "Send this result in DM",
},
],
},
"value": "clicked",
"action_id": "send-proofread-result-in-dm",
},
}
)

client.views_update(
view_id=payload["id"],
view=modal_view,
)
except Timeout:
client.views_update(
Expand All @@ -904,7 +972,7 @@ def display_proofreading_result(
},
)
except Exception as e:
logger.error(f"Failed to share a thread summary: {e}")
logger.exception(f"Failed to share a proofreading result: {e}")
client.views_update(
view_id=payload["id"],
view={
Expand Down Expand Up @@ -932,7 +1000,9 @@ def display_proofreading_modal_again(ack: Ack, payload):
private_metadata = json.loads(payload["private_metadata"])
ack(
response_action="update",
view=build_proofreading_input_modal(private_metadata["prompt"]),
view=build_proofreading_input_modal(
private_metadata["prompt"], private_metadata.get("tone_and_voice")
),
)


Expand Down
19 changes: 13 additions & 6 deletions app/openai_ops.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -544,17 +544,24 @@ def generate_proofreading_result(
logger: logging.Logger,
openai_api_key: str,
original_text: str,
tone_and_voice: Optional[str] = None,
timeout_seconds: int,
) -> str:
system_content = (
"You're an assistant tasked with helping Slack users by proofreading a given text. "
"Your task is to enhance the quality of the sentences provided "
"without altering their original meaning as far as possible. "
)
if tone_and_voice is not None:
system_content += (
f"The generated output must be a suitable one for {tone_and_voice}. "
)
system_content += "Lastly, generating results swiftly should be prioritized over achieving perfection."

messages = [
{
"role": "system",
"content": (
"You're an assistant tasked with helping Slack users by proofreading a given text. "
"Your task is to enhance the quality of the sentences provided "
"without altering their original meaning as far as possible. "
"Lastly, generating results swiftly should be prioritized over achieving perfection."
),
"content": system_content,
},
{
"role": "user",
Expand Down

0 comments on commit 7fbc3e4

Please sign in to comment.