multiple fixes

.gitignore

@@ -1,5 +1,6 @@
 # app specific
 outputs/*
+data/stories.db
 
 # Byte-compiled / optimized / DLL files
 __pycache__/

README.md

@@ -74,6 +74,7 @@ It puts the user in charge of a how the story is going to develop.
 - [x] GPT-4 story generation in a gradio interface
 - [x] Do the evaluator (if it's time to end)
 - [x] Inerchange text output for play.ht voice generation
+- [x] Expose switch to the user on what's the max lenght of story and whether ask about details or not
 - [ ] Interchange text input for whisper
 - [ ] Clear input on submit
 - [ ] Dockerfile and deploy (including magic word for access control)

adaptors/llm.py

@@ -10,6 +10,7 @@ openai.api_key = os.getenv('OPENAI_KEY')
 
 MODEL = 'gpt-4'
 #MODEL = 'gpt-3.5-turbo'
+TEMPERATURE = 0.8
 
 @retry(wait=wait_random_exponential(multiplier=1, max=40), stop=stop_after_attempt(3))
 def answer(system_message, user_and_assistant_messages):
@@ -23,7 +24,8 @@ def answer(system_message, user_and_assistant_messages):
     try:
         chat_completion = openai.ChatCompletion.create(
             model=MODEL,
-            messages=messages
+            messages=messages,
+            temperature = TEMPERATURE
         )
         output = chat_completion.choices[0].message.content
         return output

adaptors/voice.py

@@ -46,7 +46,7 @@ def say_new(text, voice="dylan"):
   client = sseclient.SSEClient(resp)
   for event in client.events():
       if event.data:
-          #pprint(event.data)
+          print(event.data)
           if helpers.is_valid_json(event.data): # play.ht api is unrealiable
               e = json.loads(event.data)
               if e["stage"] == "complete":

homeros.py

@@ -3,19 +3,124 @@ from models.story import Story
 from loguru import logger
 import json
 from pprint import pprint
-from helpers import gen_unique_id
+import helpers
 import prompts
+import random
 from adaptors.llm import answer
 from adaptors.voice import say_new
 
-MAX_STORY_LEN = 3 #after how many chunks we force the story to end
+DEFAULT_WORLD = "J.R.R. Tolkien's Middle Earth"
+DEFAULT_HERO = "I don't know, please choose something unusual"
+DEFAULT_PLOT = "I don't know, please come up with something unexpected"
+DEFAULT_ENDING = (lambda: random.choice(["happy", "tragic", "funny", "unexpected"]))()
+DEFAULT_STYLE = (lambda: random.choice(["Epic", "Funny", "Poetic"]))()
+
+'''
+Here we manage the flow and state of the story
+'''
+def do_homeros(user_input, story_data, settings):
+
+    #TODO refactor. naming in the function is old
+    story = story_data
+
+    # story hasn't started
+    if story["status"] == "not_started":
+        logger.debug("status: initiating a new story")
+        next_message = helpers.get_fixed_msg("welcome")
+        story["status"] = "checking_magic_word"
+
+    # we are checking the magic word or it is wrong and we need to ask for it again
+    elif story["status"] == "checking_magic_word" or story["status"] == "wrong_magic_word":
+        logger.debug("status: checking magic word")
+        magic_word_correct = helpers.check_magic_word(user_input)
+        if magic_word_correct:
+
+            story = init_story(story)
+            # if default settings is true - skip the asking, including magic word and just start the story
+            if settings["default_settings"] and len(story["chunks"]) == 0:
+                story = define_metadata(DEFAULT_WORLD, "world", story)
+                story = define_metadata(DEFAULT_HERO, "hero", story)
+                story = define_metadata(DEFAULT_PLOT, "plot", story)
+                story = define_metadata(DEFAULT_ENDING, "ending", story)
+                story = define_metadata(DEFAULT_STYLE, "style", story)
+                story["status"] = "ongoing"
+                story = start_story(story)
+                next_message = story["chunks"][-1]["audio_url"]
+                return next_message, story, settings
+
+            else:
+                story["status"] = "defining_metadata_world"
+                next_message = helpers.get_fixed_msg("ask_world")
+
+        else:
+            story["status"] = "wrong_magic_word"
+            next_message = helpers.get_fixed_msg("wrong_magic_word")
+
+    # defining the world
+    elif story["status"] == "defining_metadata_world":
+        logger.debug("status: magic word is wrong")
+        story = define_metadata(user_input, "world", story)
+        story["status"] = "defining_metadata_hero"
+        next_message = helpers.get_fixed_msg("ask_hero")
+
+    # defining the hero
+    elif story["status"] == "defining_metadata_hero":
+        logger.debug("status: defining the hero")
+        story = define_metadata(user_input, "hero", story)
+        story["status"] = "defining_metadata_plot"
+        next_message = helpers.get_fixed_msg("ask_plot")
+
+    # defining the plot
+    elif story["status"] == "defining_metadata_plot":
+        logger.debug("status: defining the plot")
+        story = define_metadata(user_input, "plot", story)
+        story["status"] = "defining_metadata_ending"
+        next_message = helpers.get_fixed_msg("ask_ending")
+
+    # defining the ending
+    elif story["status"] == "defining_metadata_ending":
+        logger.debug("status: defining the ending")
+        story = define_metadata(user_input, "ending", story)
+        story["status"] = "defining_metadata_style"
+        next_message = helpers.get_fixed_msg("ask_style")
+
+    # defining the style and starting the story with the first chunk
+    elif story["status"] == "defining_metadata_style":
+        logger.debug("status: defining the style")
+        story = define_metadata(user_input, "style", story)
+        story["status"] = "ongoing"
+        story = start_story(story)
+        next_message = story["chunks"][-1]["audio_url"]
+
+    # we are in the middle of the story - evaluate if time to end, or continue
+    elif story["status"] == "ongoing":
+        if evaluate_story(story, settings)["is_time_to_end"]:
+            logger.debug("status: activating story finish")
+            story = finish_story(user_input, story)
+            story["status"] = "finished"
+        else:
+            story = continue_story(user_input, story)
+            story["status"] = "ongoing"
+
+        next_message = story["chunks"][-1]["audio_url"]
+
+    # story has ended, but the user still inputting. tell them it's over
+    elif story["status"] == "finished":
+        next_message = helpers.get_fixed_msg("no_more_story")
+        story["status"] = "finished"
+
+    else:
+        raise Exception("strange story status")
+        logger.error(f"we have a story status {story['status']} we didn't catch...")
+        logger.debug(story)
+
+    return next_message, story, settings
 
 '''
 initiates a new story and saves in DB
 '''
 def init_story(story_data):
-    story_data["uuid"] = gen_unique_id()
-    pprint(story_data)
+    story_data["uuid"] = helpers.gen_unique_id()
     story = Story(
         uuid=story_data['uuid'],
         status=story_data['status']
@@ -40,6 +145,7 @@ creates the first chunk of the story
 def start_story(story_data):
     return continue_story("Please begin", story_data)
 
+
 '''
 main function that manages adding the next chunk to a story (first text, then audio)
 '''
@@ -54,9 +160,6 @@ def continue_story(user_input, story_data):
     })
     next_chunk_text = create_next_chunk_text(user_input, story)
 
-    if len(chunks) == 0:
-        next_chunk_text = "May our story begin!\n\n"+next_chunk_text
-
     next_chunk_audio = create_next_chunk_audio(next_chunk_text)
     messages.append({
         "role":"assistant",
@@ -74,6 +177,7 @@ def continue_story(user_input, story_data):
     save_story(story)
     return story.to_dict()
 
+
 '''
 generates the last part of the story and changes status to "finished"
 '''
@@ -138,11 +242,11 @@ def create_next_chunk_text(user_input, story):
 '''
 evaluates the story up until now and returns a dict with the result of the evaluation
 '''
-def evaluate_story(story):
+def evaluate_story(story, settings):
     evaluation = {}
     story_len = len(story["chunks"])
     logger.debug(story_len)
-    evaluation["is_time_to_end"] = story_len >= MAX_STORY_LEN
+    evaluation["is_time_to_end"] = story_len >= settings["max_len"]
 
     return evaluation
 

main.py

@@ -6,105 +6,23 @@ from dotenv import load_dotenv
 load_dotenv()
 import helpers
 
-from homeros import init_story, start_story, continue_story, finish_story, define_metadata, evaluate_story
+from homeros import do_homeros
 
-DEFAULT_PARAMS = True
+DEFAULT_STORY_LEN = 3 #default : after how many chunks we force the story to end
+USE_DEFAULT_SETTINGS = "Go with the defaults" #default : are we using default story config or asking the user
 
 '''
-Here we manage the flow and state of the story
+update settings
 '''
-def do_homeros(user_input, story):
-
-    # if default params is true - skip the asking, including magic word and just start the story
-    if DEFAULT_PARAMS and len(story["chunks"]) == 0:
-        story = init_story(story)
-        story = define_metadata("J.R.R. Tolkien's Middle Earth", "world", story)
-        story = define_metadata("I don't know. Please choose something unusual.", "hero", story)
-        story = define_metadata("I don't know. Please choose something unusual.", "plot", story)
-        story = define_metadata("Happy", "ending", story)
-        story = define_metadata("epic", "style", story)
-        story["status"] = "ongoing"
-        story = start_story(story)
-        next_message = story["chunks"][-1]["audio_url"]
-        return next_message, story
-
-    # story hasn't started
-    if story["status"] == "not_started":
-        logger.debug("status: initiating a new story")
-        next_message = helpers.get_fixed_msg("welcome")
-        story["status"] = "checking_magic_word"
-
-    # we are checking the magic word or it is wrong and we need to ask for it again
-    elif story["status"] == "checking_magic_word" or story["status"] == "wrong_magic_word":
-        logger.debug("status: checking magic word")
-        magic_word_correct = helpers.check_magic_word(user_input)
-        if magic_word_correct:
-            story["status"] = "defining_metadata_world"
-            story = init_story(story)
-            next_message = helpers.get_fixed_msg("ask_world")
-        else:
-            story["status"] = "wrong_magic_word"
-            next_message = helpers.get_fixed_msg("wrong_magic_word")
-
-    # defining the world
-    elif story["status"] == "defining_metadata_world":
-        logger.debug("status: magic word is wrong")
-        story = define_metadata(user_input, "world", story)
-        story["status"] = "defining_metadata_hero"
-        next_message = helpers.get_fixed_msg("ask_hero")
-
-    # defining the hero
-    elif story["status"] == "defining_metadata_hero":
-        logger.debug("status: defining the hero")
-        story = define_metadata(user_input, "hero", story)
-        story["status"] = "defining_metadata_plot"
-        next_message = helpers.get_fixed_msg("ask_plot")
-
-    # defining the plot
-    elif story["status"] == "defining_metadata_plot":
-        logger.debug("status: defining the plot")
-        story = define_metadata(user_input, "plot", story)
-        story["status"] = "defining_metadata_ending"
-        next_message = helpers.get_fixed_msg("ask_ending")
-
-    # defining the ending
-    elif story["status"] == "defining_metadata_ending":
-        logger.debug("status: defining the ending")
-        story = define_metadata(user_input, "ending", story)
-        story["status"] = "defining_metadata_style"
-        next_message = helpers.get_fixed_msg("ask_style")
-
-    # defining the style and starting the story with the first chunk
-    elif story["status"] == "defining_metadata_style":
-        logger.debug("status: defining the style")
-        story = define_metadata(user_input, "style", story)
-        story["status"] = "ongoing"
-        story = start_story(story)
-        next_message = story["chunks"][-1]["audio_url"]
-
-    # we are in the middle of the story - evaluate if time to end, or continue
-    elif story["status"] == "ongoing":
-        if evaluate_story(story)["is_time_to_end"]:
-            logger.debug("status: activating story finish")
-            story = finish_story(user_input, story)
-            story["status"] = "finished"
-        else:
-            story = continue_story(user_input, story)
-            story["status"] = "ongoing"
-
-        next_message = story["chunks"][-1]["audio_url"]
-
-    # story has ended, but the user still inputting. tell them it's over
-    elif story["status"] == "finished":
-        next_message = helpers.get_fixed_msg("no_more_story")
-        story["status"] = "finished"
-
-    else:
-        raise Exception("strange story status")
-        logger.error(f"we have a story status {story['status']} we didn't catch...")
-        logger.debug(story)
-
-    return next_message, story
+def save_settings(how_many_chunks, go_with_the_defaults, save_params_btn, go_btn, text_input, story_chunk, settings, story):
+
+    #save settings
+    settings["default_settings"] = go_with_the_defaults == "Go with the defaults"
+    settings["max_len"] = how_many_chunks + 1
+
+    #update ui
+    return gr.update(visible=False), gr.update(visible=False), gr.update(visible=False), gr.update(visible=True), gr.update(visible=True), gr.update(visible=True), settings, story
+
 
 demo = gr.Blocks()
 
@@ -126,6 +44,12 @@ with demo:
         "full_story_text": ""
     })
 
+    settings = gr.State(value={
+        "max_len" : DEFAULT_STORY_LEN,
+        "default_settings": USE_DEFAULT_SETTINGS
+    })
+
+
     with gr.Row():
         gr.Markdown('''
 # HOMEROS
@@ -133,36 +57,64 @@ with demo:
 This demo is exploring the future of interactive storytelling.
 It puts the user in charge and blurs the boundary between the reader and the author.
 
-Hit "Tell me!" to get started.
-
-When Homeros asks you something - hit record, answer with your voice and then hit "Tell me!" again.
-
         ''')
 
     with gr.Row():
+        how_many_chunks = gr.Slider(
+            minimum = 1,
+            value = DEFAULT_STORY_LEN,
+            step = 1,
+            maximum = 10,
+            label = "How long would you like your story to be?",
+            interactive = True
+        )
+
+    with gr.Row():
+        go_with_defaults = gr.Radio(
+            label = "Would you like to go with the defaults or should the storyteller ask you about the details?",
+            value = USE_DEFAULT_SETTINGS,
+            choices = [
+                "Go with the defaults",
+                "I want full control"
+            ],
+            interactive = True
+        )
 
+
+    with gr.Row():
         text_input = gr.Textbox(
-            label="you say"
+            label="you say",
+            visible=False
         )
 
     with gr.Row():
         story_chunk = gr.Audio(
             label="storyteller says",
-            autoplay=True
+            interactive=False,
+            autoplay=True,
+            visible=False
         )
 
     with gr.Row():
-
         go_btn = gr.Button(
-            "Tell me!",
+            value="Tell the story!",
+            visible=False
         )
 
     go_btn.click(
         do_homeros,
-        inputs=[text_input, story],
-        outputs=[story_chunk, story]
+        inputs=[text_input, story, settings],
+        outputs=[story_chunk, story, settings]
     )
 
+    with gr.Row():
+        save_params_btn = gr.Button("Save Settings")
+        save_params_btn.click(
+            save_settings,
+            inputs=[how_many_chunks, go_with_defaults, save_params_btn, go_btn, story_chunk, text_input, settings, story],
+            outputs=[how_many_chunks, go_with_defaults, save_params_btn, go_btn, story_chunk, text_input, settings, story]
+        )
+
     demo.queue(
         concurrency_count=5
     )