added google speech-to-text

main.py

@@ -8,7 +8,7 @@ from bot.bot import Bot
 from playground import build_pdf_extractor, build_googletranslator, build_unstable_diffusion, build_sketcher, \
     build_dalle, \
     build_whisperx, build_libretranslator, build_external_dvm, build_media_converter, build_inactive_follows_finder, \
-    build_image_converter
+    build_image_converter, build_googletranscribe
 from utils.definitions import EventDefinitions
 from utils.dvmconfig import DVMConfig
 from utils.nostr_utils import check_and_set_private_key
@@ -70,6 +70,10 @@ def run_nostr_dvm_with_local_config():
         bot_config.SUPPORTED_DVMS.append(whisperer)  # We also add Sketcher to the bot
         whisperer.run()
 
+    transcriptor = build_googletranscribe("Transcriptor", "speech_recognition")
+    bot_config.SUPPORTED_DVMS.append(transcriptor)  # We also add Sketcher to the bot
+    transcriptor.run()
+
     # Spawn DVM6, this one requires an OPENAI API Key and balance with OpenAI, you will move the task to them and pay
     # per call. Make sure you have enough balance and the DVM's cost is set higher than what you pay yourself, except, you know,
     # you're being generous.

playground.py

@@ -9,6 +9,7 @@ from tasks.discovery_inactive_follows import DiscoverInactiveFollows
 from tasks.imagegeneration_openai_dalle import ImageGenerationDALLE
 from tasks.imagegeneration_sdxl import ImageGenerationSDXL
 from tasks.imagegeneration_sdxlimg2img import ImageGenerationSDXLIMG2IMG
+from tasks.textextraction_google import SpeechToTextGoogle
 from tasks.textextraction_whisperx import SpeechToTextWhisperX
 from tasks.textextraction_pdf import TextExtractionPDF
 from tasks.translation_google import TranslationGoogle
@@ -203,7 +204,7 @@ def build_whisperx(name, identifier):
     nip89info = {
         "name": name,
         "image": "https://image.nostr.build/c33ca6fc4cc038ca4adb46fdfdfda34951656f87ee364ef59095bae1495ce669.jpg",
-        "about": "I am a test dvm to extract text from media files (very beta)",
+        "about": "I extract text from media files with WhisperX",
         "nip90Params": nip90params
     }
     nip89config = NIP89Config()
@@ -214,6 +215,34 @@ def build_whisperx(name, identifier):
                                 admin_config=admin_config, options=options)
 
 
+def build_googletranscribe(name, identifier):
+    dvm_config = DVMConfig()
+    dvm_config.PRIVATE_KEY = check_and_set_private_key(identifier)
+    dvm_config.LNBITS_INVOICE_KEY = os.getenv("LNBITS_INVOICE_KEY")
+    dvm_config.LNBITS_URL = os.getenv("LNBITS_HOST")
+    options = {'api_key': None}
+    # A module might have options it can be initialized with, here we set a default model, and the nova-server
+    # address it should use. These parameters can be freely defined in the task component
+
+    nip90params = {
+        "language": {
+            "required": False,
+            "values": ["en-US"]
+        }
+    }
+    nip89info = {
+        "name": name,
+        "image": "https://image.nostr.build/c33ca6fc4cc038ca4adb46fdfdfda34951656f87ee364ef59095bae1495ce669.jpg",
+        "about": "I extract text from media files with the Google API. I understand English by default",
+        "nip90Params": nip90params
+    }
+    nip89config = NIP89Config()
+    nip89config.DTAG = nip89config.DTAG = check_and_set_d_tag(identifier, name, dvm_config.PRIVATE_KEY,
+                                                              nip89info["image"])
+    nip89config.CONTENT = json.dumps(nip89info)
+    return SpeechToTextGoogle(name=name, dvm_config=dvm_config, nip89config=nip89config,
+                                admin_config=admin_config, options=options)
+
 def build_sketcher(name, identifier):
     dvm_config = DVMConfig()
     dvm_config.PRIVATE_KEY = check_and_set_private_key(identifier)

tasks/README.md

@@ -7,9 +7,10 @@ Reusable backend functions can be defined in backends (e.g. API calls)
 Current List of Tasks:
 
 | Module                  | Kind | Description                                    | Backend     |
-|--------------------------|------|------------------------------------------------|-------------| 
+|-------------------------|------|------------------------------------------------|-------------| 
 | TextExtractionPDF       | 5000 | Extracts Text from a PDF file                  | local       |
 | SpeechToTextWhisperX    | 5000 | Extracts Speech from Media files               | nova-server |
+| SpeechToTextGoogle      | 5000 | Extracts Speech from Media files via Google    | googleAPI   |
 | TranslationGoogle       | 5002 | Translates Inputs to another language          | googleAPI   |
 | TranslationLibre        | 5002 | Translates Inputs to another language          | libreAPI    |
 | ImageGenerationSDXL     | 5100 | Generates an Image with StableDiffusionXL      | nova-server |

tasks/textextraction_google.py

@@ -0,0 +1,125 @@
+import json
+import os
+import time
+from multiprocessing.pool import ThreadPool
+from pathlib import Path
+
+from backends.nova_server import check_nova_server_status, send_request_to_nova_server, send_file_to_nova_server
+from interfaces.dvmtaskinterface import DVMTaskInterface
+from utils.admin_utils import AdminConfig
+from utils.dvmconfig import DVMConfig
+from utils.mediasource_utils import organize_input_media_data
+from utils.nip89_utils import NIP89Config
+from utils.definitions import EventDefinitions
+
+"""
+This File contains a Module to transform a media file input on Google Cloud
+
+Accepted Inputs: Url to media file (url)
+Outputs: Transcribed text
+
+"""
+
+
+class SpeechToTextGoogle(DVMTaskInterface):
+    KIND: int = EventDefinitions.KIND_NIP90_EXTRACT_TEXT
+    TASK: str = "speech-to-text"
+    FIX_COST: float = 10
+    PER_UNIT_COST: float = 0.1
+
+    def __init__(self, name, dvm_config: DVMConfig, nip89config: NIP89Config,
+                 admin_config: AdminConfig = None, options=None):
+        super().__init__(name, dvm_config, nip89config, admin_config, options)
+        if options is None:
+            options = {}
+
+    def is_input_supported(self, tags):
+        for tag in tags:
+            if tag.as_vec()[0] == 'i':
+                input_value = tag.as_vec()[1]
+                input_type = tag.as_vec()[2]
+                if input_type != "url":
+                    return False
+
+            elif tag.as_vec()[0] == 'output':
+                output = tag.as_vec()[1]
+                if output == "" or not (output == "text/plain"):
+                    print("Output format not supported, skipping..")
+                    return False
+
+        return True
+
+    def create_request_form_from_nostr_event(self, event, client=None, dvm_config=None):
+        request_form = {"jobID": event.id().to_hex() + "_" + self.NAME.replace(" ", "")}
+
+        url = ""
+        input_type = "url"
+        start_time = 0
+        end_time = 0
+        media_format = "audio/wav"
+        language = "en-US"
+
+        for tag in event.tags():
+            if tag.as_vec()[0] == 'i':
+                input_type = tag.as_vec()[2]
+                if input_type == "url":
+                    url = tag.as_vec()[1]
+
+            elif tag.as_vec()[0] == 'param':
+                print("Param: " + tag.as_vec()[1] + ": " + tag.as_vec()[2])
+                if tag.as_vec()[1] == "language":
+                    language = tag.as_vec()[2]
+                elif tag.as_vec()[1] == "range":
+                    try:
+                        t = time.strptime(tag.as_vec()[2], "%H:%M:%S")
+                        seconds = t.tm_hour * 60 * 60 + t.tm_min * 60 + t.tm_sec
+                        start_time = float(seconds)
+                    except:
+                        try:
+                            t = time.strptime(tag.as_vec()[2], "%M:%S")
+                            seconds = t.tm_min * 60 + t.tm_sec
+                            start_time = float(seconds)
+                        except:
+                            start_time = tag.as_vec()[2]
+                            try:
+                                t = time.strptime(tag.as_vec()[3], "%H:%M:%S")
+                                seconds = t.tm_hour * 60 * 60 + t.tm_min * 60 + t.tm_sec
+                                end_time = float(seconds)
+                            except:
+                                try:
+                                    t = time.strptime(tag.as_vec()[3], "%M:%S")
+                                    seconds = t.tm_min * 60 + t.tm_sec
+                                    end_time = float(seconds)
+                                except:
+                                    end_time = float(tag.as_vec()[3])
+
+        filepath = organize_input_media_data(url, input_type, start_time, end_time, dvm_config, client, True,
+                                             media_format)
+        options = {
+            "filepath": filepath,
+            "language": language,
+        }
+        request_form['options'] = json.dumps(options)
+        return request_form
+
+    def process(self, request_form):
+        import speech_recognition as sr
+        if self.options.get("api_key"):
+            api_key = self.options['api_key']
+        else:
+            api_key = None
+        options = DVMTaskInterface.set_options(request_form)
+        # Speech recognition instance
+        asr = sr.Recognizer()
+        with sr.AudioFile(options["filepath"]) as source:
+            audio = asr.record(source)  # read the entire audio file
+
+        try:
+            # Use Google Web Speech API to recognize speech from audio data
+            result = asr.recognize_google(audio, language=options["language"], key=api_key)
+        except Exception as e:
+            print(e)
+            # If an error occurs during speech recognition, return False and the type of the exception
+            return "error"
+
+        return result