Skip to content

Commit

Permalink
refactor/performance (#19)
Browse files Browse the repository at this point in the history 
* requirements/latest_stable

* refactor/performance

reuse solver object, avoid reinit of translate plugin

switch to spoken answers api, more natural and faster

skip translation if request lang is english

session support

* refactor/performance

reuse solver object, avoid reinit of translate plugin

switch to spoken answers api, more natural and faster

skip translation if request lang is english

session support
  • Loading branch information
@JarbasAl
JarbasAl authored Jan 12, 2024
1 parent 83d54f6 commit 2a13610
Show file tree
Hide file tree
Showing 3 changed files with 55 additions and 179 deletions.
144 changes: 52 additions & 92 deletions __init__.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -17,11 +17,11 @@
import requests
from ovos_backend_client.api import WolframAlphaApi as _WA
from ovos_bus_client import Message
from ovos_bus_client.session import SessionManager
from ovos_config import Configuration
from ovos_plugin_manager.templates.solvers import QuestionSolver
from ovos_utils import classproperty
from ovos_utils.gui import can_use_gui
from ovos_utils.intents import IntentBuilder
from ovos_utils.process_utils import RuntimeRequirements
from ovos_workshop.decorators import intent_handler
from ovos_workshop.skills.common_query_skill import CommonQuerySkill, CQSMatchLevel
Expand Down Expand Up @@ -211,45 +211,11 @@ def get_expanded_answer(self, query, context=None):
class WolframAlphaSkill(CommonQuerySkill):
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
# continuous dialog, "tell me more"
self.idx = 0
self.last_query = None
self.results = []

# answer processing options
self.skip_images = True # some wolfram results are pictures with no speech
# if a gui is available the title is read and image displayed

# These results are usually unwanted as spoken responses
# they are either spammy or cant be handled by TTS properly
self.skips = [
# quantities, eg speed of light
'Comparison', # spammy
'Corresponding quantities', # spammy
'Basic unit dimensions', # TTS will fail hard 99% of time
# when asking about word definitions
'American pronunciation', # can not pronounce IPA phonemes
'Translations', # TTS wont handle other langs or charsets
'Hyphenation', # spammy
'Anagrams', # spammy
'Lexically close words', # spammy
'Overall typical frequency', # spammy
'Crossword puzzle clues', # spammy
'Scrabble score', # spammy
'Other notable uses' # spammy
]

@property
def wolfie(self):
# property to allow api key changes in config
try:
return WolframAlphaSolver({
"units": self.config_core['system_unit'],
"appid": self.settings.get("api_key")
})
except Exception as err:
self.log.error("WolframAlphaSkill failed to initialize: %s", err)
return None
self.session_results = {} # session_id: {}
self.wolfie = WolframAlphaSolver({
"units": self.config_core['system_unit'],
"appid": self.settings.get("api_key")
})

@classproperty
def runtime_requirements(self):
Expand All @@ -267,28 +233,33 @@ def runtime_requirements(self):
@intent_handler("search_wolfie.intent")
def handle_search(self, message: Message):
query = message.data["query"]
sess = SessionManager.get(message)
self.session_results[sess.session_id] = {"phrase": query,
"image": None,
"spoken_answer": ""}
response = self.ask_the_wolf(query)
if response:
self.speak_result()
self.session_results[sess.session_id]["spoken_answer"] = response
self.speak(response)
else:
self.speak_dialog("no_answer")

@intent_handler(IntentBuilder("WolfieMore").require("More").
require("WolfieKnows"))
def handle_tell_more(self, message: Message):
""" Follow up query handler, "tell me more"."""
self.speak_result()

# common query integration
def CQS_match_query_phrase(self, phrase: str):
self.log.debug("WolframAlpha query: " + phrase)
if self.wolfie is None:
self.log.error("WolframAlphaSkill not initialized, no response")
return

sess = SessionManager.get()
self.session_results[sess.session_id] = {"phrase": phrase,
"image": None,
"spoken_answer": None}

response = self.ask_the_wolf(phrase)
if response:
self.idx += 1 # spoken by common query framework
self.log.debug("WolframAlpha response: %s", response)
self.session_results[sess.session_id]["spoken_answer"] = response
self.log.debug(f"WolframAlpha response: {response}")
return (phrase, CQSMatchLevel.GENERAL, response,
{'query': phrase, 'answer': response})

Expand All @@ -297,60 +268,49 @@ def CQS_action(self, phrase: str, data: dict):
self.display_wolfie()

# wolfram integration
def ask_the_wolf(self, query: str):
# context for follow up questions
self.set_context("WolfieKnows", query)
results = self.wolfie.long_answer(query,
context={"lang": self.lang})
self.idx = 0
self.last_query = query
self.results = [s for s in results if s.get("title") not in self.skips]
if len(self.results):
return self.results[0]["summary"]
self.log.debug("WolframAlpha had no answers for %s", query)
def ask_the_wolf(self, query: str, lang: str = None):
lang = lang or self.lang
if lang.startswith("en"):
self.log.debug(f"skipping auto translation for wolfram alpha, "
f"{lang} is supported")
WolframAlphaSolver.enable_tx = False
else:
self.log.info(f"enabling auto translation for wolfram alpha, "
f"{lang} is not supported internally")
WolframAlphaSolver.enable_tx = True
return self.wolfie.spoken_answer(query, context={"lang": lang})

def display_wolfie(self):
if not can_use_gui(self.bus):
return
image = None
# issues can happen if skill reloads
# eg. "tell me more" -> invalid self.idx
if self.idx < len(self.results):
image = self.results[self.idx].get("img")
if self.last_query:
image = image or self.wolfie.visual_answer(self.last_query,
context={"lang": self.lang})

# generate image for the last query this session made
# only after skill was selected for speed
sess = SessionManager.get()
res = self.session_results.get(sess.session_id)
if not res or not res["spoken_response"]:
return

image = res.get("image") or self.wolfie.visual_answer(res["phrase"],
context={"lang": self.lang})
if image:
self.gui["wolfram_image"] = image
# scrollable full result page
self.gui.show_page(join(self.root_dir, "ui", "wolf.qml"), override_idle=45)
self.gui.show_page(join(self.root_dir, "ui", "wolf"), override_idle=45)

def speak_result(self):
if self.idx + 1 > len(self.results):
self.speak_dialog("thats all")
self.remove_context("WolfieKnows")
self.idx = 0
else:
if not self.results[self.idx].get("summary"):
if not self.skip_images and can_use_gui(self.bus):
self.speak(self.results[self.idx]["title"])
self.display_wolfie()
else:
# skip image only result
self.idx += 1
self.speak_result()
return
else:
self.display_wolfie()
# make it more speech friendly
ans = self.results[self.idx]["summary"]
ans = ans.replace(" | ", "; ")
self.speak(ans)
self.idx += 1
def stop_session(self, sess):
if sess.session_id in self.session_results:
self.session_results.pop(sess.session_id)


if __name__ == "__main__":
d = WolframAlphaSolver()
from ovos_utils.fakebus import FakeBus

d = WolframAlphaSkill(bus=FakeBus(), skill_id="fake.wolf")

print(d.ask_the_wolf("what is the speed of light"))

exit()

query = "who is Isaac Newton"

Expand Down
6 changes: 3 additions & 3 deletions requirements.txt
View file
Original file line number Diff line number Diff line change
@@ -1,5 +1,5 @@
ovos-translate-server-plugin
ovos-config>=0.0.11a10
ovos-utils~=0.0, >=0.0.28a4
ovos_workshop~=0.0, >=0.0.11a4
ovos-config>=0.0.12
ovos-utils~=0.0, >=0.0.38
ovos_workshop~=0.0, >=0.0.15
ovos-plugin-manager
84 changes: 0 additions & 84 deletions test/unittests/test_continuous_dialog.py
View file

This file was deleted.

0 comments on commit 2a13610

Please sign in to comment.