Feat/run tg rework (#114)

Co-authored-by: Fedor Ignatov <[email protected]>
Co-authored-by: dilyararimovna <[email protected]>
Co-authored-by: Denis Kuznetosv <[email protected]>

Author: 4 people

MANIFEST.in

@@ -1,3 +1,4 @@
 include requirements.txt
 include deeppavlov_agent/http_api/templates/*.html
 include deeppavlov_agent/log_config.yml
+include deeppavlov_agent/channels/telegram/config/*.yml

deeppavlov_agent/channels/__init__.py

@@ -0,0 +1,173 @@
+import asyncio
+import logging
+from pathlib import Path
+
+from aiogram import Bot, Dispatcher, types
+from aiogram.contrib.fsm_storage.memory import MemoryStorage
+from aiogram.dispatcher import FSMContext
+from aiogram.dispatcher.filters.state import State, StatesGroup
+from aiogram.utils import executor
+
+from .utils import MessageResponder
+
+config_dir = Path(__file__).resolve().parent / 'config'
+
+logging.basicConfig(
+    format="%(asctime)s - %(name)s - %(levelname)s - %(message)s", level=logging.INFO
+)
+logger = logging.getLogger(__name__)
+
+
+class DialogState(StatesGroup):
+    active = State()
+    awaiting_rating = State()
+    inactive = State()
+
+
+def run_tg(token, proxy, agent):
+    loop = asyncio.get_event_loop()
+    bot = Bot(token=token, loop=loop, proxy=proxy)
+    storage = MemoryStorage()  # TODO change to actual storage maybe?
+    dp = Dispatcher(bot, storage=storage)
+    responder = MessageResponder(
+        config_path=config_dir / "telegram_config.yml",
+        messages_path=config_dir / "telegram_messages.yml",
+        keyboards_path=config_dir / "telegram_keyboards.yml",
+    )
+
+    @dp.message_handler(commands="start")
+    async def start_handler(message: types.Message):
+        text = responder.message("start")
+        reply_markup = responder.reply_keyboard("dialog_inactive")
+
+        await message.answer(text, reply_markup=reply_markup)
+
+    @dp.message_handler(commands="help", state="*")
+    async def help_handler(message: types.Message):
+        text = responder.message("help")
+
+        await message.answer(text)
+
+    @dp.message_handler(commands="complain", state="*")
+    async def complain_handler(message: types.Message, state: FSMContext):
+        # TODO Add actual complaint logic
+        if await state.get_state() == DialogState.active.state:
+            text = responder.message("complain_success")
+        else:
+            text = responder.message("complain_fail")
+
+        await message.answer(text)
+
+    @dp.message_handler(commands="begin", state="*")
+    async def begin_dialog(message: types.Message, state: FSMContext):
+        state = await state.get_state()
+        must_evaluate = (
+            state == DialogState.awaiting_rating.state
+            and responder.config.evaluation_options.user_must_evaluate
+        )
+        is_not_finished = state == DialogState.active
+
+        if must_evaluate or is_not_finished:
+            text = responder.message("begin_fail")
+            reply_markup = None
+
+        else:
+            await DialogState.active.set()
+
+            text = responder.message("begin_success")
+            reply_markup = responder.reply_keyboard("dialog_active")
+
+        await message.answer(text, reply_markup=reply_markup)
+
+    @dp.message_handler(commands="end", state="*")
+    async def end_dialog(message: types.Message, state: FSMContext):
+        if await state.get_state() != DialogState.active.state:
+            text = responder.message("end_fail")
+            reply_markup = responder.reply_keyboard("dialog_inactive")
+        else:
+            text = responder.message("end_success")
+            dialog_id = await agent.state_manager.drop_active_dialog(
+                str(message.from_user.id)
+            )
+            reply_markup = responder.dialog_rating_inline_keyboard(dialog_id)
+
+            await DialogState.awaiting_rating.set()
+
+        await message.answer(text, reply_markup=reply_markup)
+
+    @dp.callback_query_handler(
+        lambda c: c.data.startswith("utt"), state=DialogState.active
+    )
+    async def handle_utterance_rating(
+        callback_query: types.CallbackQuery, state: FSMContext
+    ):
+        _, utterance_id, rating = callback_query.data.split("-")
+        await agent.state_manager.set_rating_utterance(
+            str(callback_query.from_user.id), utterance_id, rating
+        )
+        await bot.answer_callback_query(callback_query.id, text=rating.capitalize())
+
+    @dp.callback_query_handler(lambda c: c.data.startswith("dialog"), state="*")
+    async def handle_dialog_rating(
+        callback_query: types.CallbackQuery, state: FSMContext
+    ):
+        if await state.get_state() != DialogState.active.state:
+            _, dialog_id, rating = callback_query.data.split("-")
+
+            await agent.state_manager.set_rating_dialog(
+                str(callback_query.from_user.id), dialog_id, rating
+            )
+
+            edited_inline_keyboard = responder.dialog_rating_inline_keyboard(
+                dialog_id, chosen_rating=rating
+            )
+
+            await bot.edit_message_reply_markup(
+                chat_id=callback_query.from_user.id,
+                message_id=callback_query.message.message_id,
+                reply_markup=edited_inline_keyboard,
+            )
+
+            if responder.config.dialog_options.reveal_dialog_id:
+                message_text = responder.message(
+                    "evaluate_dialog_success_reveal_id", dialog_id=dialog_id
+                )
+            else:
+                message_text = responder.message("evaluate_dialog_success")
+            callback_text = "Evaluation saved!"
+            reply_markup = responder.reply_keyboard("dialog_inactive")
+
+            await DialogState.inactive.set()
+
+        else:
+            callback_text = ""
+            message_text = responder.message("evaluate_dialog_success")
+            reply_markup = None
+
+        await bot.answer_callback_query(callback_query.id, text=callback_text)
+        await bot.send_message(
+            callback_query.from_user.id, message_text, reply_markup=reply_markup
+        )
+
+    @dp.message_handler(state="*")
+    async def handle_message(message: types.Message, state: FSMContext):
+        if await state.get_state() == DialogState.active.state:
+            response_data = await agent.register_msg(
+                utterance=message.text,
+                user_external_id=str(message.from_user.id),
+                user_device_type="telegram",
+                date_time=message.date,
+                location="",
+                channel_type="telegram",
+                require_response=True,
+            )
+            text = response_data["dialog"].utterances[-1].text
+            utterance_id = response_data["dialog"].utterances[-1].utt_id
+            reply_markup = responder.utterance_rating_inline_keyboard(utterance_id)
+        else:
+            text = responder.message("unexpected_message")
+            reply_markup = None
+
+        await message.answer(text, reply_markup=reply_markup)
+
+    executor.start_polling(dp, skip_updates=True)

deeppavlov_agent/channels/telegram/__init__.py

@@ -0,0 +1,7 @@
+dialog_options:
+  reveal_dialog_id: true
+
+evaluation_options:
+  user_must_evaluate: true
+  min_score: 1
+  max_score: 5

deeppavlov_agent/channels/telegram/bot.py

@@ -0,0 +1,7 @@
+dialog_inactive:
+  - '/begin'
+  - '/help'
+
+dialog_active:
+  - '/end'
+  - '/complain'

deeppavlov_agent/channels/telegram/config/telegram_config.yml

@@ -0,0 +1,61 @@
+start: >
+  This chatbot is developed by Neural Networks and Deep Learning Lab at MIPT.
+  Please have a chat with it and evaluate its performance.
+  To begin a conversation enter /begin. To end a conversation: enter /end.
+  When the dialogue is finished, you will be asked to evaluate it.
+  You will have to rate the conversation from 1 (bad) to 5 (excellent).
+  Your conversations will be recorded for further use.
+  By starting a chat you give permission for your anonymized conversation data
+  to be released publicly under Apache License Version 2.0
+  https://www.apache.org / licenses / LICENSE – 2.0.
+
+help: >
+  This chatbot is developed by Neural Networks and Deep Learning Lab at MIPT.
+  Please have a chat with it and evaluate its performance.
+  To begin a conversation enter /begin. To end a conversation: enter /end.
+  When the dialogue is finished, you will be asked to evaluate it.
+  You will have to rate the conversation from 1 (bad) to 5 (excellent).
+  Your conversations will be recorded for further use.
+  By starting a chat you give permission for your anonymized conversation data
+  to be released publicly under Apache License Version 2.0
+  https://www.apache.org / licenses / LICENSE – 2.0.
+
+complain_success: >
+  Your complaint has been recorded and will be examined by the system administrator.
+  Note that your conversation is still active.
+  You can always use /end command to end it.
+
+complain_fail: >
+  Could not save your complaint. Did you /begin the dialog?
+  You cannot complain when there are no messages in the dialog.
+
+begin_success: >
+  Starting a dialog. To finish the dialog enter /end.
+
+begin_fail: >
+  Cannot start a new conversation.
+  Please finish and evaluate your current dialog first.
+  Use /help command for usage instructions.
+
+end_success: >
+  Dialog is finished.
+  Please evaluate the whole dialog using one of the buttons below:
+
+end_fail: >
+  You're not in a dialog now.
+
+evaluate_dialog_success: >
+  Thank you for participation!
+
+evaluate_dialog_success_reveal_id: >
+  Thank you for participation!
+  Evaluated dialog secret id: ${dialog_id}
+
+evaluate_dialog_fail: >
+  Evaluation is not allowed at the moment.
+  Use /help command for usage instructions.
+
+unexpected_message: >
+  Unexpected message.
+  You are not in a dialog yet or the dialog has already been finished.
+  Use /help command for usage instructions.