Pylint & CVE fix

.gitattributes

@@ -33,4 +33,3 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
-*.pptx filter=lfs diff=lfs merge=lfs -text

.streamlit/config.toml

@@ -1,10 +0,0 @@
-[server]
-runOnSave = true
-headless = false
-maxUploadSize = 0
-
-[browser]
-gatherUsageStats = false
-
-[theme]
-base = "dark"

README.md

@@ -4,7 +4,7 @@ emoji: 🏢
 colorFrom: yellow
 colorTo: green
 sdk: streamlit
-sdk_version: 1.32.2
+sdk_version: 1.26.0
 app_file: app.py
 pinned: false
 license: mit
@@ -16,54 +16,36 @@ We spend a lot of time on creating the slides and organizing our thoughts for an
 With SlideDeck AI, co-create slide decks on any topic with Generative Artificial Intelligence.
 Describe your topic and let SlideDeck AI generate a PowerPoint slide deck for you—it's as simple as that!
 
-SlideDeck AI is powered by [Mistral 7B Instruct](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.2).
+SlideDeck AI is powered by [Mistral 7B Instruct](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.1).
 Originally, it was built using the Llama 2 API provided by Clarifai.
 
-*Update (v4.0)*: Legacy SlideDeck AI allowed one-shot generation of a slide deck based on the inputs. 
-In contrast, SlideDeck AI *Reloaded* enables an iterative workflow with a conversational interface,
-where you can create and improve the presentation.
-
-
 # Process
 
 SlideDeck AI works in the following way:
 
-1. Given a topic description, it uses Mistral 7B Instruct to generate the *initial* content of the slides. 
+1. Given a topic description, it uses Mistral 7B Instruct to generate the outline/contents of the slides. 
 The output is generated as structured JSON data based on a pre-defined schema.
 2. Subsequently, it uses the `python-pptx` library to generate the slides, 
 based on the JSON data from the previous step. 
-A user can choose from a set of three pre-defined presentation templates.
-3. At this stage onward, a user can provide additional instructions to *refine* the content.
-For example, one can ask to add another slide or modify an existing slide.
-A history of instructions is maintained.
-4. Every time SlideDeck AI generates a PowerPoint presentation, a download button is provided.
-Clicking on the button will download the file.
-
-
-# Known Issues
-
-- **Connection timeout**: Requests sent to the Hugging Face Inference endpoint might time out.
-A maximum of five retries are attempted. If it still does not work, wait for a while and try again.
+Here, a user can choose from a set of three pre-defined presentation templates.
+3. In addition, it uses Metaphor to fetch Web pages related to the topic.
 
-The following is not an issue but might appear as a strange behavior:
-- **Cannot paste text in the input box**: If the length of the copied text is greater than the maximum
-number of allowed characters in the textbox, pasting would not work.
+4. ~~Finally, it uses Stable Diffusion 2 to generate an image, based on the title and each slide heading.~~
 
 
 # Local Development
 
-SlideDeck AI uses [Mistral 7B Instruct](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.2) 
+SlideDeck AI uses [Mistral 7B Instruct](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.1) 
 via the Hugging Face Inference API.
-To run this project by yourself, you need to provide the `HUGGINGFACEHUB_API_TOKEN` API key,
+To run this project by yourself, you need to provide the `HUGGINGFACEHUB_API_TOKEN` and `METAPHOR_API_KEY` API keys,
 for example, in a `.env` file. Visit the respective websites to obtain the keys.
 
 
 # Live Demo
 
-- [SlideDeck AI](https://huggingface.co/spaces/barunsaha/slide-deck-ai) on Hugging Face Spaces
-- [Demo video](https://youtu.be/QvAKzNKtk9k) of the chat interface on YouTube
+[SlideDeck AI](https://huggingface.co/spaces/barunsaha/slide-deck-ai)
 
 
 # Award
 
-SlideDeck AI has won the 3rd Place in the [Llama 2 Hackathon with Clarifai](https://lablab.ai/event/llama-2-hackathon-with-clarifai) in 2023.
+SlideDeck AI has won the 3rd Place in the [Llama 2 Hackathon with Clarifai](https://lablab.ai/event/llama-2-hackathon-with-clarifai).

app.py

@@ -1,377 +1,310 @@
-import datetime
-import logging
 import pathlib
-import random
+import logging
 import tempfile
-from typing import List
+from typing import List, Tuple
 
 import json5
+import metaphor_python as metaphor
 import streamlit as st
-from langchain_community.chat_message_histories import (
-    StreamlitChatMessageHistory
-)
-from langchain_core.messages import HumanMessage
-from langchain_core.prompts import ChatPromptTemplate
-# from transformers import AutoTokenizer
 
+import llm_helper
+import pptx_helper
 from global_config import GlobalConfig
-from helpers import llm_helper, pptx_helper, text_helper
+
+
+APP_TEXT = json5.loads(open(GlobalConfig.APP_STRINGS_FILE, 'r', encoding='utf-8').read())
+GB_CONVERTER = 2 ** 30
+
+
+logging.basicConfig(
+    level=GlobalConfig.LOG_LEVEL,
+    format='%(asctime)s - %(message)s',
+)
 
 
 @st.cache_data
-def _load_strings() -> dict:
+def get_contents_wrapper(text: str) -> str:
     """
-    Load various strings to be displayed in the app.
-    :return: The dictionary of strings.
+    Fetch and cache the slide deck contents on a topic by calling an external API.
+
+    :param text: The presentation topic
+    :return: The slide deck contents or outline in JSON format
     """
 
-    with open(GlobalConfig.APP_STRINGS_FILE, 'r', encoding='utf-8') as in_file:
-        return json5.loads(in_file.read())
+    logging.info('LLM call because of cache miss...')
+    return llm_helper.generate_slides_content(text).strip()
 
 
-@st.cache_data
-def _get_prompt_template(is_refinement: bool) -> str:
+@st.cache_resource
+def get_metaphor_client_wrapper() -> metaphor.Metaphor:
     """
-    Return a prompt template.
+    Create a Metaphor client for semantic Web search.
 
-    :param is_refinement: Whether this is the initial or refinement prompt.
-    :return: The prompt template as f-string.
+    :return: Metaphor instance
     """
 
-    if is_refinement:
-        with open(GlobalConfig.REFINEMENT_PROMPT_TEMPLATE, 'r', encoding='utf-8') as in_file:
-            template = in_file.read()
-    else:
-        with open(GlobalConfig.INITIAL_PROMPT_TEMPLATE, 'r', encoding='utf-8') as in_file:
-            template = in_file.read()
+    return metaphor.Metaphor(api_key=GlobalConfig.METAPHOR_API_KEY)
 
-    return template
 
+@st.cache_data
+def get_web_search_results_wrapper(text: str) -> List[Tuple[str, str]]:
+    """
+    Fetch and cache the Web search results on a given topic.
 
-# @st.cache_resource
-# def _get_tokenizer() -> AutoTokenizer:
-#     """
-#     Get Mistral tokenizer for counting tokens.
-#
-#     :return: The tokenizer.
-#     """
-#
-#     return AutoTokenizer.from_pretrained(
-#         pretrained_model_name_or_path=GlobalConfig.HF_LLM_MODEL_NAME
-#     )
+    :param text: The topic
+    :return: A list of (title, link) tuples
+    """
 
+    results = []
+    search_results = get_metaphor_client_wrapper().search(
+        text,
+        use_autoprompt=True,
+        num_results=5
+    )
 
-APP_TEXT = _load_strings()
+    for a_result in search_results.results:
+        results.append((a_result.title, a_result.url))
 
-# Session variables
-CHAT_MESSAGES = 'chat_messages'
-DOWNLOAD_FILE_KEY = 'download_file_name'
-IS_IT_REFINEMENT = 'is_it_refinement'
+    return results
 
-logger = logging.getLogger(__name__)
-progress_bar = st.progress(0, text='Setting up SlideDeck AI...')
 
-texts = list(GlobalConfig.PPTX_TEMPLATE_FILES.keys())
-captions = [GlobalConfig.PPTX_TEMPLATE_FILES[x]['caption'] for x in texts]
-pptx_template = st.sidebar.radio(
-    'Select a presentation template:',
-    texts,
-    captions=captions,
-    horizontal=True
-)
+# def get_disk_used_percentage() -> float:
+#     """
+#     Compute the disk usage.
+#
+#     :return: Percentage of the disk space currently used
+#     """
+#
+#     total, used, free = shutil.disk_usage(__file__)
+#     total = total // GB_CONVERTER
+#     used = used // GB_CONVERTER
+#     free = free // GB_CONVERTER
+#     used_perc = 100.0 * used / total
+#
+#     logging.debug(f'Total: {total} GB\n'
+#                   f'Used: {used} GB\n'
+#                   f'Free: {free} GB')
+#
+#     logging.debug('\n'.join(os.listdir()))
+#
+#     return used_perc
 
 
-def display_page_header_content():
+def build_ui():
     """
-    Display content in the page header.
+    Display the input elements for content generation. Only covers the first step.
     """
 
+    # get_disk_used_percentage()
+
     st.title(APP_TEXT['app_name'])
     st.subheader(APP_TEXT['caption'])
     st.markdown(
-        '![Visitors](https://api.visitorbadge.io/api/visitors?path=https%3A%2F%2Fhuggingface.co%2Fspaces%2Fbarunsaha%2Fslide-deck-ai&countColor=%23263759)'  # noqa: E501
+        'Powered by'
+        ' [Mistral-7B-Instruct-v0.2](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.2).'
+    )
+    st.markdown(
+        '*If the JSON is generated or parsed incorrectly, try again later by making minor changes'
+        ' to the input text.*'
     )
 
+    with st.form('my_form'):
+        # Topic input
+        try:
+            with open(GlobalConfig.PRELOAD_DATA_FILE, 'r', encoding='utf-8') as in_file:
+                preload_data = json5.loads(in_file.read())
+        except (FileExistsError, FileNotFoundError):
+            preload_data = {'topic': '', 'audience': ''}
+
+        topic = st.text_area(
+            APP_TEXT['input_labels'][0],
+            value=preload_data['topic']
+        )
 
-def display_page_footer_content():
-    """
-    Display content in the page footer.
-    """
-
-    st.text(APP_TEXT['tos'] + '\n\n' + APP_TEXT['tos2'])
+        texts = list(GlobalConfig.PPTX_TEMPLATE_FILES.keys())
+        captions = [GlobalConfig.PPTX_TEMPLATE_FILES[x]['caption'] for x in texts]
 
+        pptx_template = st.radio(
+            'Select a presentation template:',
+            texts,
+            captions=captions,
+            horizontal=True
+        )
 
-def build_ui():
-    """
-    Display the input elements for content generation.
-    """
+        st.divider()
+        submit = st.form_submit_button('Generate slide deck')
 
-    display_page_header_content()
+    if submit:
+        # st.write(f'Clicked {time.time()}')
+        st.session_state.submitted = True
 
-    with st.expander('Usage Policies and Limitations'):
-        display_page_footer_content()
+    # https://github.com/streamlit/streamlit/issues/3832#issuecomment-1138994421
+    if 'submitted' in st.session_state:
+        progress_text = 'Generating the slides...give it a moment'
+        progress_bar = st.progress(0, text=progress_text)
 
-    progress_bar.progress(50, text='Setting up chat interface...')
-    set_up_chat_ui()
+        topic_txt = topic.strip()
+        generate_presentation(topic_txt, pptx_template, progress_bar)
 
+    st.divider()
+    st.text(APP_TEXT['tos'])
+    st.text(APP_TEXT['tos2'])
 
-def set_up_chat_ui():
-    """
-    Prepare the chat interface and related functionality.
-    """
+    st.markdown(
+        '![Visitors]'
+        '(https://api.visitorbadge.io/api/visitors?path=https%3A%2F%2Fhuggingface.co%2Fspaces%2Fbarunsaha%2Fslide-deck-ai&countColor=%23263759)'
+    )
 
-    with st.expander('Usage Instructions'):
-        st.markdown(GlobalConfig.CHAT_USAGE_INSTRUCTIONS)
-        st.markdown(
-            'SlideDeck AI is powered by'
-            ' [Mistral-7B-Instruct-v0.2](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.2)'
-        )
 
-    # view_messages = st.expander('View the messages in the session state')
+def generate_presentation(topic: str, pptx_template: str, progress_bar):
+    """
+    Process the inputs to generate the slides.
 
-    st.chat_message('ai').write(
-        random.choice(APP_TEXT['ai_greetings'])
-    )
-    progress_bar.progress(100, text='Done!')
-    progress_bar.empty()
+    :param topic: The presentation topic based on which contents are to be generated
+    :param pptx_template: The PowerPoint template name to be used
+    :param progress_bar: Progress bar from the page
+    :return:
+    """
 
-    history = StreamlitChatMessageHistory(key=CHAT_MESSAGES)
+    topic_length = len(topic)
+    logging.debug('Input length:: topic: %s', topic_length)
+
+    if topic_length >= 10:
+        logging.debug('Topic: %s', topic)
+        target_length = min(topic_length, GlobalConfig.LLM_MODEL_MAX_INPUT_LENGTH)
+
+        try:
+            # Step 1: Generate the contents in JSON format using an LLM
+            json_str = process_slides_contents(topic[:target_length], progress_bar)
+            logging.debug('Truncated topic: %s', topic[:target_length])
+            logging.debug('Length of JSON: %d', len(json_str))
+
+            # Step 2: Generate the slide deck based on the template specified
+            if len(json_str) > 0:
+                st.info(
+                    'Tip: The generated content doesn\'t look so great?'
+                    ' Need alternatives? Just change your description text and try again.',
+                    icon="💡️"
+                )
+            else:
+                st.error(
+                    'Unfortunately, JSON generation failed, so the next steps would lead'
+                    ' to nowhere. Try again or come back later.'
+                )
+                return
+
+            all_headers = generate_slide_deck(json_str, pptx_template, progress_bar)
+
+            # Step 3: Bonus stuff: Web references and AI art
+            show_bonus_stuff(all_headers)
+
+        except ValueError as ve:
+            st.error(f'Unfortunately, an error occurred: {ve}! '
+                     f'Please change the text, try again later, or report it, sharing your inputs.')
 
-    if _is_it_refinement():
-        template = _get_prompt_template(is_refinement=True)
     else:
-        template = _get_prompt_template(is_refinement=False)
-
-    prompt_template = ChatPromptTemplate.from_template(template)
-
-    # Since Streamlit app reloads at every interaction, display the chat history
-    # from the save session state
-    for msg in history.messages:
-        msg_type = msg.type
-        if msg_type == 'user':
-            st.chat_message(msg_type).write(msg.content)
-        else:
-            st.chat_message(msg_type).code(msg.content, language='json')
-
-    if prompt := st.chat_input(
-        placeholder=APP_TEXT['chat_placeholder'],
-        max_chars=GlobalConfig.LLM_MODEL_MAX_INPUT_LENGTH
-    ):
-
-        progress_bar_pptx = st.progress(0, 'Preparing to run...')
-        if not text_helper.is_valid_prompt(prompt):
-            st.error(
-                'Not enough information provided!'
-                ' Please be a little more descriptive and type a few words'
-                ' with a few characters :)'
-            )
-            return
-
-        logger.info('User input: %s | #characters: %d', prompt, len(prompt))
-        st.chat_message('user').write(prompt)
-
-        user_messages = _get_user_messages()
-        user_messages.append(prompt)
-        list_of_msgs = [
-            f'{idx + 1}. {msg}' for idx, msg in enumerate(user_messages)
-        ]
-        list_of_msgs = '\n'.join(list_of_msgs)
-
-        if _is_it_refinement():
-            formatted_template = prompt_template.format(
-                **{
-                    'instructions': list_of_msgs,
-                    'previous_content': _get_last_response()
-                }
-            )
-        else:
-            formatted_template = prompt_template.format(
-                **{
-                    'question': prompt,
-                }
-            )
-
-        progress_bar_pptx.progress(5, 'Calling LLM...will retry if connection times out...')
-        response: dict = llm_helper.hf_api_query({
-            'inputs': formatted_template,
-            'parameters': {
-                'temperature': GlobalConfig.LLM_MODEL_TEMPERATURE,
-                'min_length': GlobalConfig.LLM_MODEL_MIN_OUTPUT_LENGTH,
-                'max_length': GlobalConfig.LLM_MODEL_MAX_OUTPUT_LENGTH,
-                'max_new_tokens': GlobalConfig.LLM_MODEL_MAX_OUTPUT_LENGTH,
-                'num_return_sequences': 1,
-                'return_full_text': False,
-                # "repetition_penalty": 0.0001
-            },
-            'options': {
-                'wait_for_model': True,
-                'use_cache': True
-            }
-        })
-
-        if len(response) > 0 and 'generated_text' in response[0]:
-            response: str = response[0]['generated_text'].strip()
-
-        st.chat_message('ai').code(response, language='json')
-
-        history.add_user_message(prompt)
-        history.add_ai_message(response)
-
-        # if GlobalConfig.COUNT_TOKENS:
-        #     tokenizer = _get_tokenizer()
-        #     tokens_count_in = len(tokenizer.tokenize(formatted_template))
-        #     tokens_count_out = len(tokenizer.tokenize(response))
-        #     logger.debug(
-        #         'Tokens count:: input: %d, output: %d',
-        #         tokens_count_in, tokens_count_out
-        #     )
-
-        # _display_messages_history(view_messages)
-
-        # The content has been generated as JSON
-        # There maybe trailing ``` at the end of the response -- remove them
-        # To be careful: ``` may be part of the content as well when code is generated
-        progress_bar_pptx.progress(50, 'Analyzing response...')
-        response_cleaned = text_helper.get_clean_json(response)
-
-        logger.info(
-            'Cleaned JSON response:: original length: %d | cleaned length: %d',
-            len(response), len(response_cleaned)
-        )
-        logger.debug('Cleaned JSON: %s', response_cleaned)
+        st.error('Not enough information provided! Please be little more descriptive :)')
 
-        # Now create the PPT file
-        progress_bar_pptx.progress(75, 'Creating the slide deck...give it a moment...')
-        generate_slide_deck(response_cleaned)
-        progress_bar_pptx.progress(100, text='Done!')
 
-        logger.info(
-            '#messages in history / 2: %d',
-            len(st.session_state[CHAT_MESSAGES]) / 2
-        )
-
-
-def generate_slide_deck(json_str: str):
+def process_slides_contents(text: str, progress_bar: st.progress) -> str:
     """
-    Create a slide deck.
+    Convert given text into structured data and display. Update the UI.
 
-    :param json_str: The content in *valid* JSON format.
+    :param text: The topic description for the presentation
+    :param progress_bar: Progress bar for this step
+    :return: The contents as a JSON-formatted string
     """
 
-    if DOWNLOAD_FILE_KEY in st.session_state:
-        path = pathlib.Path(st.session_state[DOWNLOAD_FILE_KEY])
-    else:
-        temp = tempfile.NamedTemporaryFile(delete=False, suffix='.pptx')
-        path = pathlib.Path(temp.name)
-        st.session_state[DOWNLOAD_FILE_KEY] = str(path)
-
-        if temp:
-            temp.close()
-
-    logger.debug('Creating PPTX file: %s...', st.session_state[DOWNLOAD_FILE_KEY])
+    json_str = ''
 
     try:
-        pptx_helper.generate_powerpoint_presentation(
-            json_str,
-            slides_template=pptx_template,
-            output_file_path=path
-        )
-    except ValueError:
-        # st.error(
-        #     f"{APP_TEXT['json_parsing_error']}"
-        #     f"\n\nAdditional error info: {ve}"
-        #     f"\n\nHere are some sample instructions that you could try to possibly fix this error;"
-        #     f" if these don't work, try rephrasing or refreshing:"
-        #     f"\n\n"
-        #     "- Regenerate content and fix the JSON error."
-        #     "\n- Regenerate content and fix the JSON error. Quotes inside quotes should be escaped."
-        # )
-        # logger.error('%s', APP_TEXT['json_parsing_error'])
-        # logger.error('Additional error info: %s', str(ve))
+        logging.info('Calling LLM for content generation on the topic: %s', text)
+        json_str = get_contents_wrapper(text)
+    except Exception as ex:
         st.error(
-            'Encountered error while parsing JSON...will fix it and retry'
-        )
-        logger.error(
-            'Caught ValueError: trying again after repairing JSON...'
+            f'An exception occurred while trying to convert to JSON. It could be because of heavy'
+            f' traffic or something else. Try doing it again or try again later.'
+            f'\nError message: {ex}'
         )
 
-        pptx_helper.generate_powerpoint_presentation(
-            text_helper.fix_malformed_json(json_str),
-            slides_template=pptx_template,
-            output_file_path=path
-        )
-    except Exception as ex:
-        st.error(APP_TEXT['content_generation_error'])
-        logger.error('Caught a generic exception: %s', str(ex))
-    finally:
-        _display_download_button(path)
-
-
-def _is_it_refinement() -> bool:
-    """
-    Whether it is the initial prompt or a refinement.
+    progress_bar.progress(50, text='Contents generated')
 
-    :return: True if it is the initial prompt; False otherwise.
-    """
-
-    if IS_IT_REFINEMENT in st.session_state:
-        return True
-
-    if len(st.session_state[CHAT_MESSAGES]) >= 2:
-        # Prepare for the next call
-        st.session_state[IS_IT_REFINEMENT] = True
-        return True
+    with st.expander('The generated contents (in JSON format)'):
+        st.code(json_str, language='json')
 
-    return False
+    return json_str
 
 
-def _get_user_messages() -> List[str]:
-    """
-    Get a list of user messages submitted until now from the session state.
-
-    :return: The list of user messages.
+def generate_slide_deck(json_str: str, pptx_template: str, progress_bar) -> List:
     """
+    Create a slide deck.
 
-    return [
-        msg.content for msg in st.session_state[CHAT_MESSAGES] if isinstance(msg, HumanMessage)
-    ]
-
-
-def _get_last_response() -> str:
+    :param json_str: The contents in JSON format
+    :param pptx_template: The PPTX template name
+    :param progress_bar: Progress bar
+    :return: A list of all slide headers and the title
     """
-    Get the last response generated by AI.
 
-    :return: The response text.
-    """
+    progress_text = 'Creating the slide deck...give it a moment'
+    progress_bar.progress(75, text=progress_text)
 
-    return st.session_state[CHAT_MESSAGES][-1].content
+    # # Get a unique name for the file to save -- use the session ID
+    # ctx = st_sr.get_script_run_ctx()
+    # session_id = ctx.session_id
+    # timestamp = time.time()
+    # output_file_name = f'{session_id}_{timestamp}.pptx'
 
+    temp = tempfile.NamedTemporaryFile(delete=False, suffix='.pptx')
+    path = pathlib.Path(temp.name)
 
-def _display_messages_history(view_messages: st.expander):
-    """
-    Display the history of messages.
+    logging.info('Creating PPTX file...')
+    all_headers = pptx_helper.generate_powerpoint_presentation(
+        json_str,
+        as_yaml=False,
+        slides_template=pptx_template,
+        output_file_path=path
+    )
+    progress_bar.progress(100, text='Done!')
 
-    :param view_messages: The list of AI and Human messages.
-    """
+    with open(path, 'rb') as f:
+        st.download_button('Download PPTX file', f, file_name='Presentation.pptx')
 
-    with view_messages:
-        view_messages.json(st.session_state[CHAT_MESSAGES])
+    return all_headers
 
 
-def _display_download_button(file_path: pathlib.Path):
+def show_bonus_stuff(ppt_headers: List[str]):
     """
-    Display a download button to download a slide deck.
+    Show bonus stuff for the presentation.
 
-    :param file_path: The path of the .pptx file.
+    :param ppt_headers: A list of the slide headings.
     """
 
-    with open(file_path, 'rb') as download_file:
-        st.download_button(
-            'Download PPTX file ⬇️',
-            data=download_file,
-            file_name='Presentation.pptx',
-            key=datetime.datetime.now()
-        )
+    # Use the presentation title and the slide headers to find relevant info online
+    logging.info('Calling Metaphor search...')
+    ppt_text = ' '.join(ppt_headers)
+    search_results = get_web_search_results_wrapper(ppt_text)
+    md_text_items = []
+
+    for (title, link) in search_results:
+        md_text_items.append(f'[{title}]({link})')
+
+    with st.expander('Related Web references'):
+        st.markdown('\n\n'.join(md_text_items))
+
+    logging.info('Done!')
+
+    # # Avoid image generation. It costs time and an API call, so just limit to the text generation.
+    # with st.expander('AI-generated image on the presentation topic'):
+    #     logging.info('Calling SDXL for image generation...')
+    #     # img_empty.write('')
+    #     # img_text.write(APP_TEXT['image_info'])
+    #     image = get_ai_image_wrapper(ppt_text)
+    #
+    #     if len(image) > 0:
+    #         image = base64.b64decode(image)
+    #         st.image(image, caption=ppt_text)
+    #         st.info('Tip: Right-click on the image to save it.', icon="💡️")
+    #         logging.info('Image added')
 
 
 def main():

clarifai_grpc_helper.py

@@ -0,0 +1,71 @@
+from clarifai_grpc.channel.clarifai_channel import ClarifaiChannel
+from clarifai_grpc.grpc.api import resources_pb2, service_pb2, service_pb2_grpc
+from clarifai_grpc.grpc.api.status import status_code_pb2
+
+from global_config import GlobalConfig
+
+
+CHANNEL = ClarifaiChannel.get_grpc_channel()
+STUB = service_pb2_grpc.V2Stub(CHANNEL)
+
+METADATA = (
+    ('authorization', 'Key ' + GlobalConfig.CLARIFAI_PAT),
+)
+
+USER_DATA_OBJECT = resources_pb2.UserAppIDSet(
+    user_id=GlobalConfig.CLARIFAI_USER_ID,
+    app_id=GlobalConfig.CLARIFAI_APP_ID
+)
+
+RAW_TEXT = '''You are a helpful, intelligent chatbot. Create the slides for a presentation on the given topic. Include main headings for each slide, detailed bullet points for each slide. Add relevant content to each slide. Do not output any blank line.
+
+Topic:
+Talk about AI, covering what it is and how it works. Add its pros, cons, and future prospects. Also, cover its job prospects.
+'''
+
+
+def get_text_from_llm(prompt: str) -> str:
+    post_model_outputs_response = STUB.PostModelOutputs(
+        service_pb2.PostModelOutputsRequest(
+            user_app_id=USER_DATA_OBJECT,  # The userDataObject is created in the overview and is required when using a PAT
+            model_id=GlobalConfig.CLARIFAI_MODEL_ID,
+            # version_id=MODEL_VERSION_ID,  # This is optional. Defaults to the latest model version
+            inputs=[
+                resources_pb2.Input(
+                    data=resources_pb2.Data(
+                        text=resources_pb2.Text(
+                            raw=prompt
+                        )
+                    )
+                )
+            ]
+        ),
+        metadata=METADATA
+    )
+
+    if post_model_outputs_response.status.code != status_code_pb2.SUCCESS:
+        print(post_model_outputs_response.status)
+        raise Exception(f"Post model outputs failed, status: {post_model_outputs_response.status.description}")
+
+    # Since we have one input, one output will exist here
+    output = post_model_outputs_response.outputs[0]
+
+    # print("Completion:\n")
+    # print(output.data.text.raw)
+
+    return output.data.text.raw
+
+
+if __name__ == '__main__':
+    topic = ('Talk about AI, covering what it is and how it works.'
+             ' Add its pros, cons, and future prospects.'
+             ' Also, cover its job prospects.'
+             )
+    print(topic)
+
+    with open(GlobalConfig.SLIDES_TEMPLATE_FILE, 'r') as in_file:
+        prompt_txt = in_file.read()
+        prompt_txt = prompt_txt.replace('{topic}', topic)
+        response_txt = get_text_from_llm(prompt_txt)
+
+        print('Output:\n', response_txt)

examples/example_04.json

@@ -1,3 +0,0 @@
-{
-    "topic": "12 slides on a basic tutorial on Python along with examples"
-}

global_config.py

@@ -1,4 +1,3 @@
-import logging
 import os
 
 from dataclasses import dataclass
@@ -13,25 +12,22 @@ class GlobalConfig:
     HF_LLM_MODEL_NAME = 'mistralai/Mistral-7B-Instruct-v0.2'
     LLM_MODEL_TEMPERATURE: float = 0.2
     LLM_MODEL_MIN_OUTPUT_LENGTH: int = 50
-    LLM_MODEL_MAX_OUTPUT_LENGTH: int = 4096
-    LLM_MODEL_MAX_INPUT_LENGTH: int = 750
+    LLM_MODEL_MAX_OUTPUT_LENGTH: int = 2000
+    LLM_MODEL_MAX_INPUT_LENGTH: int = 300
 
     HUGGINGFACEHUB_API_TOKEN = os.environ.get('HUGGINGFACEHUB_API_TOKEN', '')
     METAPHOR_API_KEY = os.environ.get('METAPHOR_API_KEY', '')
 
     LOG_LEVEL = 'DEBUG'
-    COUNT_TOKENS = False
     APP_STRINGS_FILE = 'strings.json'
     PRELOAD_DATA_FILE = 'examples/example_02.json'
     SLIDES_TEMPLATE_FILE = 'langchain_templates/template_combined.txt'
-    # JSON_TEMPLATE_FILE = 'langchain_templates/text_to_json_template_02.txt'
-    INITIAL_PROMPT_TEMPLATE = 'langchain_templates/chat_prompts/initial_template_v3_two_cols.txt'
-    REFINEMENT_PROMPT_TEMPLATE = 'langchain_templates/chat_prompts/refinement_template_v3_two_cols.txt'
+    JSON_TEMPLATE_FILE = 'langchain_templates/text_to_json_template_02.txt'
 
     PPTX_TEMPLATE_FILES = {
-        'Basic': {
+        'Blank': {
             'file': 'pptx_templates/Blank.pptx',
-            'caption': 'A good start (Uses [photos](https://unsplash.com/photos/AFZ-qBPEceA) by [cetteup](https://unsplash.com/@cetteup?utm_content=creditCopyText&utm_medium=referral&utm_source=unsplash) on [Unsplash](https://unsplash.com/photos/a-foggy-forest-filled-with-lots-of-trees-d3ci37Gcgxg?utm_content=creditCopyText&utm_medium=referral&utm_source=unsplash))'
+            'caption': 'A good start'
         },
         'Ion Boardroom': {
             'file': 'pptx_templates/Ion_Boardroom.pptx',
@@ -42,30 +38,3 @@ class GlobalConfig:
             'caption': 'Marvel in a monochrome dream'
         }
     }
-
-    # This is a long text, so not incorporated as a string in `strings.json`
-    CHAT_USAGE_INSTRUCTIONS = (
-        'Briefly describe your topic of presentation in the textbox provided below.'
-        ' For example:\n'
-        '- Make a slide deck on AI.'
-        '\n\n'
-        'Subsequently, you can add follow-up instructions, e.g.:\n'
-        '- Can you add a slide on GPUs?'
-        '\n\n'
-        ' You can also ask it to refine any particular slide, e.g.:\n'
-        '- Make the slide with title \'Examples of AI\' a bit more descriptive.'
-        '\n\n'
-        'See this [demo video](https://youtu.be/QvAKzNKtk9k) for a brief walkthrough.'
-        'SlideDeck AI does not have access to the Web.'
-        '\n\n'
-        'If you like SlideDeck AI, please consider leaving a heart ❤️ on the'
-        ' [Hugging Face Space](https://huggingface.co/spaces/barunsaha/slide-deck-ai/) or'
-        ' a star ⭐ on [GitHub](https://github.com/barun-saha/slide-deck-ai).'
-    )
-
-
-logging.basicConfig(
-    level=GlobalConfig.LOG_LEVEL,
-    format='%(asctime)s - %(levelname)s - %(name)s - %(message)s',
-    datefmt='%Y-%m-%d %H:%M:%S'
-)

helpers/pptx_helper.py

@@ -1,540 +0,0 @@
-import logging
-import pathlib
-import re
-import tempfile
-
-from typing import List, Tuple
-
-import json5
-import pptx
-from pptx.enum.shapes import MSO_AUTO_SHAPE_TYPE
-
-from global_config import GlobalConfig
-
-
-# English Metric Unit (used by PowerPoint) to inches
-EMU_TO_INCH_SCALING_FACTOR = 1.0 / 914400
-INCHES_1_5 = pptx.util.Inches(1.5)
-INCHES_1 = pptx.util.Inches(1)
-INCHES_0_5 = pptx.util.Inches(0.5)
-INCHES_0_4 = pptx.util.Inches(0.4)
-INCHES_0_3 = pptx.util.Inches(0.3)
-
-STEP_BY_STEP_PROCESS_MARKER = '>> '
-
-PATTERN = re.compile(r"^slide[ ]+\d+:", re.IGNORECASE)
-SAMPLE_JSON_FOR_PPTX = '''
-{
-    "title": "Understanding AI",
-    "slides": [
-        {
-            "heading": "Introduction",
-            "bullet_points": [
-                "Brief overview of AI",
-                [
-                    "Importance of understanding AI"
-                ]
-            ]
-        }
-    ]
-}
-'''
-
-logger = logging.getLogger(__name__)
-
-
-def remove_slide_number_from_heading(header: str) -> str:
-    """
-    Remove the slide number from a given slide header.
-
-    :param header: The header of a slide.
-    """
-
-    if PATTERN.match(header):
-        idx = header.find(':')
-        header = header[idx + 1:]
-
-    return header
-
-
-def generate_powerpoint_presentation(
-        structured_data: str,
-        slides_template: str,
-        output_file_path: pathlib.Path
-) -> List:
-    """
-    Create and save a PowerPoint presentation file containing the content in JSON format.
-
-    :param structured_data: The presentation contents as "JSON" (may contain trailing commas).
-    :param slides_template: The PPTX template to use.
-    :param output_file_path: The path of the PPTX file to save as.
-    :return A list of presentation title and slides headers.
-    """
-
-    # The structured "JSON" might contain trailing commas, so using json5
-    parsed_data = json5.loads(structured_data)
-
-    logger.debug(
-        '*** Using PPTX template: %s',
-        GlobalConfig.PPTX_TEMPLATE_FILES[slides_template]['file']
-    )
-    presentation = pptx.Presentation(GlobalConfig.PPTX_TEMPLATE_FILES[slides_template]['file'])
-    slide_width_inch, slide_height_inch = _get_slide_width_height_inches(presentation)
-
-    # The title slide
-    title_slide_layout = presentation.slide_layouts[0]
-    slide = presentation.slides.add_slide(title_slide_layout)
-    title = slide.shapes.title
-    subtitle = slide.placeholders[1]
-    title.text = parsed_data['title']
-    logger.info(
-        'PPT title: %s | #slides: %d',
-        title.text, len(parsed_data['slides'])
-    )
-    subtitle.text = 'by Myself and SlideDeck AI :)'
-    all_headers = [title.text, ]
-
-    # Add content in a loop
-    for a_slide in parsed_data['slides']:
-        is_processing_done = _handle_double_col_layout(
-            presentation=presentation,
-            slide_json=a_slide,
-            slide_width_inch=slide_width_inch,
-            slide_height_inch=slide_height_inch
-        )
-
-        if not is_processing_done:
-            is_processing_done = _handle_step_by_step_process(
-                presentation=presentation,
-                slide_json=a_slide,
-                slide_width_inch=slide_width_inch,
-                slide_height_inch=slide_height_inch
-            )
-
-        if not is_processing_done:
-            _handle_default_display(
-                presentation=presentation,
-                slide_json=a_slide,
-                slide_width_inch=slide_width_inch,
-                slide_height_inch=slide_height_inch
-            )
-
-    # The thank-you slide
-    last_slide_layout = presentation.slide_layouts[0]
-    slide = presentation.slides.add_slide(last_slide_layout)
-    title = slide.shapes.title
-    title.text = 'Thank you!'
-
-    presentation.save(output_file_path)
-
-    return all_headers
-
-
-def get_flat_list_of_contents(items: list, level: int) -> List[Tuple]:
-    """
-    Flatten a (hierarchical) list of bullet points to a single list containing each item and
-    its level.
-
-    :param items: A bullet point (string or list).
-    :param level: The current level of hierarchy.
-    :return: A list of (bullet item text, hierarchical level) tuples.
-    """
-
-    flat_list = []
-
-    for item in items:
-        if isinstance(item, str):
-            flat_list.append((item, level))
-        elif isinstance(item, list):
-            flat_list = flat_list + get_flat_list_of_contents(item, level + 1)
-
-    return flat_list
-
-
-def _handle_default_display(
-        presentation: pptx.Presentation,
-        slide_json: dict,
-        slide_width_inch: float,
-        slide_height_inch: float
-):
-    """
-    Display a list of text in a slide.
-
-    :param presentation: The presentation object.
-    :param slide_json: The content of the slide as JSON data.
-    :param slide_width_inch: The width of the slide in inches.
-    :param slide_height_inch: The height of the slide in inches.
-    """
-
-    bullet_slide_layout = presentation.slide_layouts[1]
-    slide = presentation.slides.add_slide(bullet_slide_layout)
-
-    shapes = slide.shapes
-    title_shape = shapes.title
-    body_shape = shapes.placeholders[1]
-    title_shape.text = remove_slide_number_from_heading(slide_json['heading'])
-    text_frame = body_shape.text_frame
-
-    # The bullet_points may contain a nested hierarchy of JSON arrays
-    # In some scenarios, it may contain objects (dictionaries) because the LLM generated so
-    #  ^ The second scenario is not covered
-
-    flat_items_list = get_flat_list_of_contents(slide_json['bullet_points'], level=0)
-
-    for idx, an_item in enumerate(flat_items_list):
-        if idx == 0:
-            text_frame.text = an_item[0].removeprefix(STEP_BY_STEP_PROCESS_MARKER)
-        else:
-            paragraph = text_frame.add_paragraph()
-            paragraph.text = an_item[0].removeprefix(STEP_BY_STEP_PROCESS_MARKER)
-            paragraph.level = an_item[1]
-
-    _handle_key_message(
-        the_slide=slide,
-        slide_json=slide_json,
-        slide_height_inch=slide_height_inch,
-        slide_width_inch=slide_width_inch
-    )
-
-
-def _handle_double_col_layout(
-        presentation: pptx.Presentation(),
-        slide_json: dict,
-        slide_width_inch: float,
-        slide_height_inch: float
-) -> bool:
-    """
-    Add a slide with a double column layout for comparison.
-
-    :param presentation: The presentation object.
-    :param slide_json: The content of the slide as JSON data.
-    :param slide_width_inch: The width of the slide in inches.
-    :param slide_height_inch: The height of the slide in inches.
-    :return: True if double col layout has been added; False otherwise.
-    """
-
-    if 'bullet_points' in slide_json and slide_json['bullet_points']:
-        double_col_content = slide_json['bullet_points']
-
-        if double_col_content and (
-                len(double_col_content) == 2
-        ) and isinstance(double_col_content[0], dict) and isinstance(double_col_content[1], dict):
-            slide = presentation.slide_layouts[4]
-            slide = presentation.slides.add_slide(slide)
-
-            shapes = slide.shapes
-            title_placeholder = shapes.title
-            title_placeholder.text = remove_slide_number_from_heading(slide_json['heading'])
-
-            left_heading, right_heading = shapes.placeholders[1], shapes.placeholders[3]
-            left_col, right_col = shapes.placeholders[2], shapes.placeholders[4]
-            left_col_frame, right_col_frame = left_col.text_frame, right_col.text_frame
-
-            if 'heading' in double_col_content[0]:
-                left_heading.text = double_col_content[0]['heading']
-            if 'bullet_points' in double_col_content[0]:
-                flat_items_list = get_flat_list_of_contents(
-                    double_col_content[0]['bullet_points'], level=0
-                )
-
-                for idx, an_item in enumerate(flat_items_list):
-                    if idx == 0:
-                        left_col_frame.text = an_item[0].removeprefix(STEP_BY_STEP_PROCESS_MARKER)
-                    else:
-                        paragraph = left_col_frame.add_paragraph()
-                        paragraph.text = an_item[0].removeprefix(STEP_BY_STEP_PROCESS_MARKER)
-                        paragraph.level = an_item[1]
-
-            if 'heading' in double_col_content[1]:
-                right_heading.text = double_col_content[1]['heading']
-            if 'bullet_points' in double_col_content[1]:
-                flat_items_list = get_flat_list_of_contents(
-                    double_col_content[1]['bullet_points'], level=0
-                )
-
-                for idx, an_item in enumerate(flat_items_list):
-                    if idx == 0:
-                        right_col_frame.text = an_item[0].removeprefix(STEP_BY_STEP_PROCESS_MARKER)
-                    else:
-                        paragraph = right_col_frame.add_paragraph()
-                        paragraph.text = an_item[0].removeprefix(STEP_BY_STEP_PROCESS_MARKER)
-                        paragraph.level = an_item[1]
-
-            _handle_key_message(
-                the_slide=slide,
-                slide_json=slide_json,
-                slide_height_inch=slide_height_inch,
-                slide_width_inch=slide_width_inch
-            )
-
-            return True
-
-    return False
-
-
-def _handle_step_by_step_process(
-        presentation: pptx.Presentation,
-        slide_json: dict,
-        slide_width_inch: float,
-        slide_height_inch: float
-) -> bool:
-    """
-    Add shapes to display a step-by-step process in the slide, if available.
-
-    :param presentation: The presentation object.
-    :param slide_json: The content of the slide as JSON data.
-    :param slide_width_inch: The width of the slide in inches.
-    :param slide_height_inch: The height of the slide in inches.
-    :return True if this slide has a step-by-step process depiction added; False otherwise.
-    """
-
-    if 'bullet_points' in slide_json and slide_json['bullet_points']:
-        steps = slide_json['bullet_points']
-
-        no_marker_count = 0.0
-        n_steps = len(steps)
-
-        # Ensure that it is a single list of strings without any sub-list
-        for step in steps:
-            if not isinstance(step, str):
-                return False
-
-            # In some cases, one or two steps may not begin with >>, e.g.:
-            # {
-            #     "heading": "Step-by-Step Process: Creating a Legacy",
-            #     "bullet_points": [
-            #         "Identify your unique talents and passions",
-            #         ">> Develop your skills and knowledge",
-            #         ">> Create meaningful work",
-            #         ">> Share your work with the world",
-            #         ">> Continuously learn and adapt"
-            #     ],
-            #     "key_message": ""
-            # },
-            #
-            # Use a threshold, e.g., at most 20%
-            if not step.startswith(STEP_BY_STEP_PROCESS_MARKER):
-                no_marker_count += 1
-
-        slide_header = slide_json['heading'].lower()
-        if (no_marker_count / n_steps > 0.25) and not (
-                ('step-by-step' in slide_header) or ('step by step' in slide_header)
-        ):
-            return False
-
-        bullet_slide_layout = presentation.slide_layouts[1]
-        slide = presentation.slides.add_slide(bullet_slide_layout)
-        shapes = slide.shapes
-        shapes.title.text = remove_slide_number_from_heading(slide_json['heading'])
-
-        if 3 <= n_steps <= 4:
-            # Horizontal display
-            height = INCHES_1_5
-            width = pptx.util.Inches(slide_width_inch / n_steps - 0.01)
-            top = pptx.util.Inches(slide_height_inch / 2)
-            left = pptx.util.Inches((slide_width_inch - width.inches * n_steps) / 2 + 0.05)
-
-            for step in steps:
-                shape = shapes.add_shape(MSO_AUTO_SHAPE_TYPE.CHEVRON, left, top, width, height)
-                shape.text = step.removeprefix(STEP_BY_STEP_PROCESS_MARKER)
-                left += width - INCHES_0_4
-        elif 4 < n_steps <= 6:
-            # Vertical display
-            height = pptx.util.Inches(0.65)
-            top = pptx.util.Inches(slide_height_inch / 4)
-            left = INCHES_1  # slide_width_inch - width.inches)
-
-            # Find the close to median width, based on the length of each text, to be set
-            # for the shapes
-            width = pptx.util.Inches(slide_width_inch * 2 / 3)
-            lengths = [len(step) for step in steps]
-            font_size_20pt = pptx.util.Pt(20)
-            widths = sorted(
-                [
-                    min(
-                        pptx.util.Inches(font_size_20pt.inches * a_len),
-                        width
-                    ) for a_len in lengths
-                ]
-            )
-            width = widths[len(widths) // 2]
-
-            for step in steps:
-                shape = shapes.add_shape(MSO_AUTO_SHAPE_TYPE.PENTAGON, left, top, width, height)
-                shape.text = step.removeprefix(STEP_BY_STEP_PROCESS_MARKER)
-                top += height + INCHES_0_3
-                left += INCHES_0_5
-        else:
-            # Two steps -- probably not a process
-            # More than 5--6 steps -- would likely cause a visual clutter
-            return False
-
-    return True
-
-
-def _handle_key_message(
-        the_slide: pptx.slide.Slide,
-        slide_json: dict,
-        slide_width_inch: float,
-        slide_height_inch: float
-):
-    """
-    Add a shape to display the key message in the slide, if available.
-
-    :param the_slide: The slide to be processed.
-    :param slide_json: The content of the slide as JSON data.
-    :param slide_width_inch: The width of the slide in inches.
-    :param slide_height_inch: The height of the slide in inches.
-    """
-
-    if 'key_message' in slide_json and slide_json['key_message']:
-        height = pptx.util.Inches(1.6)
-        width = pptx.util.Inches(slide_width_inch / 2.3)
-        top = pptx.util.Inches(slide_height_inch - height.inches - 0.1)
-        left = pptx.util.Inches((slide_width_inch - width.inches) / 2)
-        shape = the_slide.shapes.add_shape(
-            MSO_AUTO_SHAPE_TYPE.ROUNDED_RECTANGLE,
-            left=left,
-            top=top,
-            width=width,
-            height=height
-        )
-        shape.text = slide_json['key_message']
-
-
-def _get_slide_width_height_inches(presentation: pptx.Presentation) -> Tuple[float, float]:
-    """
-    Get the dimensions of a slide in inches.
-
-    :param presentation: The presentation object.
-    :return: The width and the height.
-    """
-
-    slide_width_inch = EMU_TO_INCH_SCALING_FACTOR * presentation.slide_width
-    slide_height_inch = EMU_TO_INCH_SCALING_FACTOR * presentation.slide_height
-    # logger.debug('Slide width: %f, height: %f', slide_width_inch, slide_height_inch)
-
-    return slide_width_inch, slide_height_inch
-
-
-if __name__ == '__main__':
-    _JSON_DATA = '''
-    {
-    "title": "Understanding AI",
-    "slides": [
-        {
-            "heading": "Introduction",
-            "bullet_points": [
-                "Brief overview of AI",
-                [
-                    "Importance of understanding AI"
-                ]
-            ],
-            "key_message": ""
-        },
-        {
-            "heading": "What is AI?",
-            "bullet_points": [
-                "Definition of AI",
-                [
-                    "Types of AI",
-                    [
-                        "Narrow or weak AI",
-                        "General or strong AI"
-                    ]
-                ],
-                "Differences between AI and machine learning"
-            ],
-            "key_message": ""
-        },
-        {
-            "heading": "How AI Works",
-            "bullet_points": [
-                "Overview of AI algorithms",
-                [
-                    "Types of AI algorithms",
-                    [
-                        "Rule-based systems",
-                        "Decision tree systems",
-                        "Neural networks"
-                    ]
-                ],
-                "How AI processes data"
-            ],
-            "key_message": ""
-        },
-        {
-            "heading": "Building AI Models",
-            "bullet_points": [
-                ">> Collect data",
-                ">> Select model or architecture to use",
-                ">> Set appropriate parameters",
-                ">> Train model with data",
-                ">> Run inference",
-            ],
-            "key_message": ""
-        },
-        {
-            "heading": "Pros and Cons: Deep Learning vs. Classical Machine Learning",
-            "bullet_points": [
-                {
-                    "heading": "Classical Machine Learning",
-                    "bullet_points": [
-                        "Interpretability: Easy to understand the model",
-                        "Faster Training: Quicker to train models",
-                        "Scalability: Can handle large datasets"
-                    ]
-                },
-                {
-                    "heading": "Deep Learning",
-                    "bullet_points": [
-                        "Handling Complex Data: Can learn from raw data",
-                        "Feature Extraction: Automatically learns features",
-                        "Improved Accuracy: Achieves higher accuracy"
-                    ]
-                }
-            ],
-            "key_message": ""
-        },
-        {
-            "heading": "Pros of AI",
-            "bullet_points": [
-                "Increased efficiency and productivity",
-                "Improved accuracy and precision",
-                "Enhanced decision-making capabilities",
-                "Personalized experiences"
-            ],
-            "key_message": "AI can be used for many different purposes"
-        },
-        {
-            "heading": "Cons of AI",
-            "bullet_points": [
-                "Job displacement and loss of employment",
-                "Bias and discrimination",
-                "Privacy and security concerns",
-                "Dependence on technology"
-            ],
-            "key_message": ""
-        },
-        {
-            "heading": "Future Prospects of AI",
-            "bullet_points": [
-                "Advancements in fields such as healthcare and finance",
-                "Increased use"
-            ],
-            "key_message": ""
-        }
-    ]
-}'''
-
-    temp = tempfile.NamedTemporaryFile(delete=False, suffix='.pptx')
-    path = pathlib.Path(temp.name)
-
-    generate_powerpoint_presentation(
-        json5.loads(_JSON_DATA),
-        output_file_path=path,
-        slides_template='Basic'
-    )
-    print(f'File path: {path}')
-
-    temp.close()

helpers/text_helper.py

@@ -1,89 +0,0 @@
-import json_repair as jr
-
-
-def is_valid_prompt(prompt: str) -> bool:
-    """
-    Verify whether user input satisfies the concerned constraints.
-
-    :param prompt: The user input text.
-    :return: True if all criteria are satisfied; False otherwise.
-    """
-
-    if len(prompt) < 7 or ' ' not in prompt:
-        return False
-
-    return True
-
-
-def get_clean_json(json_str: str) -> str:
-    """
-    Attempt to clean a JSON response string from the LLM by removing the trailing ```
-    and any text beyond that.
-    CAUTION: May not be always accurate.
-
-    :param json_str: The input string in JSON format.
-    :return: The "cleaned" JSON string.
-    """
-
-    # An example of response containing JSON and other text:
-    # {
-    #     "title": "AI and the Future: A Transformative Journey",
-    #     "slides": [
-    #       ...
-    #     ]
-    # }    <<---- This is end of valid JSON content
-    # ```
-    #
-    # ```vbnet
-    # Please note that the JSON output is in valid format but the content of the "Role of GPUs in AI" slide is just an example and may not be factually accurate. For accurate information, you should consult relevant resources and update the content accordingly.
-    # ```
-    response_cleaned = json_str
-
-    while True:
-        idx = json_str.rfind('```')  # -1 on failure
-
-        if idx <= 0:
-            break
-
-        # In the ideal scenario, the character before the last ``` should be
-        # a new line or a closing bracket }
-        prev_char = json_str[idx - 1]
-
-        if (prev_char == '}') or (prev_char == '\n' and json_str[idx - 2] == '}'):
-            response_cleaned = json_str[:idx]
-
-        json_str = json_str[:idx]
-
-    return response_cleaned
-
-
-def fix_malformed_json(json_str: str) -> str:
-    """
-    Try and fix the syntax error(s) in a JSON string.
-
-    :param json_str: The input JSON string.
-    :return: The fixed JSOn string.
-    """
-
-    return jr.repair_json(json_str, skip_json_loads=True)
-
-
-if __name__ == '__main__':
-    json1 = '''{
-    "key": "value"
-    }
-    '''
-    json2 = '''["Reason": "Regular updates help protect against known vulnerabilities."]'''
-    json3 = '''["Reason" Regular updates help protect against known vulnerabilities."]'''
-    json4 = '''
-    {"bullet_points": [
-        ">> Write without stopping or editing",
-        >> Set daily writing goals and stick to them,
-        ">> Allow yourself to make mistakes"
-    ],}
-    '''
-
-    print(fix_malformed_json(json1))
-    print(fix_malformed_json(json2))
-    print(fix_malformed_json(json3))
-    print(fix_malformed_json(json4))

langchain_templates/chat_prompts/initial_template.txt

@@ -1,41 +0,0 @@
-You are a helpful, intelligent chatbot. Create the slides for a presentation on the given topic.
-Include main headings for each slide, detailed bullet points for each slide.
-Add relevant content to each slide.
-The content of each slide should be verbose, descriptive, and very detailed.
-If relevant, add one or two examples to illustrate the concept.
-Unless explicitly specified with the topic, create about 10 slides.
-
-
-### Topic:
-{question}
-
-
-The output must be only a valid and syntactically correct JSON adhering to the following schema:
-{{
-    "title": "Presentation Title",
-    "slides": [
-        {{
-            "heading": "Heading for the First Slide",
-            "bullet_points": [
-                "First bullet point",
-                [
-                    "Sub-bullet point 1",
-                    "Sub-bullet point 2"
-                ],
-                "Second bullet point"
-            ]
-        }},
-        {{
-            "heading": "Heading for the Second Slide",
-            "bullet_points": [
-                "First bullet point",
-                "Second bullet item",
-                "Third bullet point"
-            ]
-        }}
-    ]
-}}
-
-
-### Output:
-```json

langchain_templates/chat_prompts/initial_template_v2_steps.txt

@@ -1,59 +0,0 @@
-You are a helpful, intelligent chatbot. Create the slides for a presentation on the given topic.
-Include main headings for each slide, detailed bullet points for each slide.
-Add relevant content to each slide.
-The content of each slide should be verbose, descriptive, and very detailed.
-If relevant, add one or two examples to illustrate the concept.
-Unless explicitly specified with the topic, create about 10 slides.
-
-
-### Topic:
-{question}
-
-
-The output must be only a valid and syntactically correct JSON adhering to the following schema:
-{{
-    "title": "Presentation Title",
-    "slides": [
-        {{
-            "heading": "Heading for the First Slide",
-            "bullet_points": [
-                "First bullet point",
-                [
-                    "Sub-bullet point 1",
-                    "Sub-bullet point 2"
-                ],
-                "Second bullet point"
-            ],
-            "key_message": ""
-        }},
-        {{
-            "heading": "Heading for the Second Slide",
-            "bullet_points": [
-                "First bullet point",
-                "Second bullet item",
-                "Third bullet point"
-            ],
-            "key_message": "The key message conveyed in this slide"
-        }},
-        {{
-            "heading": "A slide that describes a step-by-step/sequential process",
-            "bullet_points": [
-                ">> The first step of the process (begins with special marker >>)",
-                ">> A second step (begins with >>)",
-                ">> Third step",
-            ],
-            "key_message": ""
-        }}
-    ]
-}}
-
-
-### Some more hints on the slide content and JSON output format:
-- For two or three important slides, generate the key message that those slides convey and assign
-them to the `key_message` elements of JSON output.
-- Identify if a slide describes a step-by-step/sequential process, then begin the bullet points
-with a special marker >>. Limit this to max two or three slides.
-
-
-### Output:
-```json

langchain_templates/chat_prompts/initial_template_v3_two_cols.txt

@@ -1,78 +0,0 @@
-You are a helpful, intelligent chatbot. Create the slides for a presentation on the given topic.
-Include main headings for each slide, detailed bullet points for each slide.
-Add relevant content to each slide.
-The content of each slide should be VERBOSE, DESCRIPTIVE, and very DETAILED.
-If relevant, add one or two EXAMPLES to illustrate the concept.
-For two or three important slides, generate the key message that those slides convey.
-Identify if a slide describes a step-by-step/sequential process, then begin the bullet points with a special marker >>. Limit this to max two or three slides.
-Also, add at least one slide with a double column layout by generating appropriate content based on the description in the JSON schema provided below.
-ALWAYS add a concluding slide at the end, containing a list of the key takeways and an optional call-to-action if relevant to the context.
-Unless explicitly instructed, create 10 TO 12 SLIDES in total.
-
-
-### Topic:
-{question}
-
-
-The output must be only a valid and syntactically correct JSON adhering to the following schema:
-{{
-    "title": "Presentation Title",
-    "slides": [
-        {{
-            "heading": "Heading for the First Slide",
-            "bullet_points": [
-                "First bullet point",
-                [
-                    "Sub-bullet point 1",
-                    "Sub-bullet point 2"
-                ],
-                "Second bullet point"
-            ],
-            "key_message": ""
-        }},
-        {{
-            "heading": "Heading for the Second Slide",
-            "bullet_points": [
-                "First bullet point",
-                "Second bullet item",
-                "Third bullet point"
-            ],
-            "key_message": "The key message conveyed in this slide"
-        }},
-        {{
-            "heading": "A slide that describes a step-by-step/sequential process",
-            "bullet_points": [
-                ">> The first step of the process (begins with special marker >>)",
-                ">> A second step (begins with >>)",
-                ">> Third step",
-            ],
-            "key_message": ""
-        }},
-        {{
-            "heading": "A slide with a double column layout (useful for side-by-side comparison/contrasting of two related concepts, e.g., pros & cons, advantages & risks, old approach vs. modern approach, and so on)",
-            "bullet_points": [
-                {{
-                    "heading": "Heading of the left column",
-                    "bullet_points": [
-                        "First bullet point",
-                        "Second bullet item",
-                        "Third bullet point"
-                    ]
-                }},
-                {{
-                    "heading": "Heading of the right column",
-                    "bullet_points": [
-                        "First bullet point",
-                        "Second bullet item",
-                        "Third bullet point"
-                    ]
-                }}
-            ],
-            "key_message": ""
-        }}
-    ]
-}}
-
-
-### Output:
-```json

langchain_templates/chat_prompts/refinement_template.txt

@@ -1,49 +0,0 @@
-You are a helpful, intelligent chatbot. You follow instructions to refine an existing slide deck.
-A list of user instructions is provided below in sequential order -- from the oldest to the latest.
-The previously generated content of the slide deck in JSON format is also provided.
-Follow the instructions to revise the content of the previously generated slides of the presentation on the given topic.
-Include main headings for each slide, detailed bullet points for each slide.
-Add relevant content to each slide.
-The content of the slides should be descriptive, verbose, and detailed.
-If relevant, add one or two examples to illustrate the concept.
-Unless explicitly specified with the topic, create about 10 slides.
-You also fix any syntax error that may be present in the JSON-formatted content.
-
-
-### List of instructions:
-{instructions}
-
-
-### Previously generated slide deck content as JSON:
-{previous_content}
-
-
-The output must be only a valid and syntactically correct JSON adhering to the following schema:
-{{
-    "title": "Presentation Title",
-    "slides": [
-        {{
-            "heading": "Heading for the First Slide",
-            "bullet_points": [
-                "First bullet point",
-                [
-                    "Sub-bullet point 1",
-                    "Sub-bullet point 2"
-                ],
-                "Second bullet point"
-            ]
-        }},
-        {{
-            "heading": "Heading for the Second Slide",
-            "bullet_points": [
-                "First bullet point",             
-                "Second bullet item",
-                "Third bullet point"
-            ]
-        }}
-    ]
-}}
-
-
-### Output:
-```json

langchain_templates/chat_prompts/refinement_template_v2_steps.txt

@@ -1,70 +0,0 @@
-You are a helpful, intelligent chatbot. You follow instructions to refine an existing slide deck.
-A list of user instructions is provided below in sequential order -- from the oldest to the latest.
-The previously generated content of the slide deck in JSON format is also provided.
-Follow the instructions to revise the content of the previously generated slides of the presentation on the given topic.
-Include main headings for each slide, detailed bullet points for each slide.
-Add relevant content to each slide.
-The content of the slides should be descriptive, verbose, and detailed.
-If relevant, add one or two examples to illustrate the concept.
-Unless explicitly specified with the topic, create about 10 slides.
-You also fix any syntax error that may be present in the JSON-formatted content.
-
-A slide that describes a step-by-step/sequential process begins the bullet points
-with a special marker >>
-
-
-### List of instructions:
-{instructions}
-
-
-### Previously generated slide deck content as JSON:
-{previous_content}
-
-
-The output must be only a valid and syntactically correct JSON adhering to the following schema:
-{{
-    "title": "Presentation Title",
-    "slides": [
-        {{
-            "heading": "Heading for the First Slide",
-            "bullet_points": [
-                "First bullet point",
-                [
-                    "Sub-bullet point 1",
-                    "Sub-bullet point 2"
-                ],
-                "Second bullet point"
-            ],
-            "key_message": ""
-        }},
-        {{
-            "heading": "Heading for the Second Slide",
-            "bullet_points": [
-                "First bullet point",
-                "Second bullet item",
-                "Third bullet point"
-            ],
-            "key_message": "The key message conveyed in this slide"
-        }},
-        {{
-            "heading": "A slide that describes a step-by-step/sequential process",
-            "bullet_points": [
-                ">> The first step of the process (begins with special marker >>)",
-                ">> A second step (begins with >>)",
-                ">> Third step",
-            ],
-            "key_message": ""
-        }}
-    ]
-}}
-
-
-### Some more hints on the slide content and JSON output format:
-- For two or three important slides, generate the key message that those slides convey and assign
-them to the `key_message` elements of JSON output.
-- Identify if a slide describes a step-by-step/sequential process, then begin the bullet points
-with a special marker >>. Limit this to max two or three slides.
-
-
-### Output:
-```json

langchain_templates/chat_prompts/refinement_template_v3_two_cols.txt

@@ -1,85 +0,0 @@
-You are a helpful, intelligent chatbot. You follow instructions to refine an existing slide deck.
-A list of user instructions is provided below in sequential order -- from the oldest to the latest.
-The previously generated content of the slide deck in JSON format is also provided.
-Follow the instructions to revise the content of the previously generated slides of the presentation on the given topic.
-Include main headings for each slide, detailed bullet points for each slide.
-Add relevant content to each slide.
-The content of each slide should be VERBOSE, DESCRIPTIVE, and very DETAILED.
-If relevant, add one or two EXAMPLES to illustrate the concept.
-For two or three important slides, generate the key message that those slides convey.
-Identify if a slide describes a step-by-step/sequential process, then begin the bullet points with a special marker >>. Limit this to max two or three slides.
-Also, add at least one slide with a double column layout by generating appropriate content based on the description in the JSON schema provided below.
-ALWAYS add a concluding slide at the end, containing a list of the key takeways and an optional call-to-action if relevant to the context.
-Unless explicitly instructed, create 10 TO 12 SLIDES in total.
-
-
-### List of instructions:
-{instructions}
-
-
-### Previously generated slide deck content as JSON:
-{previous_content}
-
-
-The output must be only a valid and syntactically correct JSON adhering to the following schema:
-{{
-    "title": "Presentation Title",
-    "slides": [
-        {{
-            "heading": "Heading for the First Slide",
-            "bullet_points": [
-                "First bullet point",
-                [
-                    "Sub-bullet point 1",
-                    "Sub-bullet point 2"
-                ],
-                "Second bullet point"
-            ],
-            "key_message": ""
-        }},
-        {{
-            "heading": "Heading for the Second Slide",
-            "bullet_points": [
-                "First bullet point",
-                "Second bullet item",
-                "Third bullet point"
-            ],
-            "key_message": "The key message conveyed in this slide"
-        }},
-        {{
-            "heading": "A slide that describes a step-by-step/sequential process",
-            "bullet_points": [
-                ">> The first step of the process (begins with special marker >>)",
-                ">> A second step (begins with >>)",
-                ">> Third step",
-            ],
-            "key_message": ""
-        }},
-        {{
-            "heading": "A slide with a double column layout (useful for side-by-side comparison/contrasting of two related concepts, e.g., pros & cons, advantages & risks, old approach vs. modern approach, and so on)",
-            "bullet_points": [
-                {{
-                    "heading": "Heading of the left column",
-                    "bullet_points": [
-                        "First bullet point",
-                        "Second bullet item",
-                        "Third bullet point"
-                    ]
-                }},
-                {{
-                    "heading": "Heading of the right column",
-                    "bullet_points": [
-                        "First bullet point",
-                        "Second bullet item",
-                        "Third bullet point"
-                    ]
-                }}
-            ],
-            "key_message": ""
-        }}
-    ]
-}}
-
-
-### Output:
-```json

langchain_templates/template_combined.txt

@@ -1,9 +1,5 @@
-You are a helpful, intelligent chatbot. Create the slides for a presentation on the given topic.
-Include main headings for each slide, detailed bullet points for each slide.
-Add relevant content to each slide.
-The content should be descriptive, verbose, and detailed as much as possible.
+You are a helpful, intelligent chatbot. Create the slides for a presentation on the given topic. Include main headings for each slide, detailed bullet points for each slide. Add relevant content to each slide.
 If relevant, add one or two examples to illustrate the concept.
-Unless explicitly specified with the topic, create about 10 slides.
 
 
 Topic:

legacy_app.py

@@ -1,294 +0,0 @@
-import pathlib
-import logging
-import tempfile
-from typing import List, Tuple
-
-import json5
-import metaphor_python as metaphor
-import streamlit as st
-
-from helpers import llm_helper, pptx_helper
-from global_config import GlobalConfig
-
-
-APP_TEXT = json5.loads(open(GlobalConfig.APP_STRINGS_FILE, 'r', encoding='utf-8').read())
-GB_CONVERTER = 2 ** 30
-
-
-logger = logging.getLogger(__name__)
-
-
-@st.cache_data
-def get_contents_wrapper(text: str) -> str:
-    """
-    Fetch and cache the slide deck contents on a topic by calling an external API.
-
-    :param text: The presentation topic.
-    :return: The slide deck contents or outline in JSON format.
-    """
-
-    logger.info('LLM call because of cache miss...')
-    return llm_helper.generate_slides_content(text).strip()
-
-
-@st.cache_resource
-def get_metaphor_client_wrapper() -> metaphor.Metaphor:
-    """
-    Create a Metaphor client for semantic Web search.
-
-    :return: Metaphor instance.
-    """
-
-    return metaphor.Metaphor(api_key=GlobalConfig.METAPHOR_API_KEY)
-
-
-@st.cache_data
-def get_web_search_results_wrapper(text: str) -> List[Tuple[str, str]]:
-    """
-    Fetch and cache the Web search results on a given topic.
-
-    :param text: The topic.
-    :return: A list of (title, link) tuples.
-    """
-
-    results = []
-    search_results = get_metaphor_client_wrapper().search(
-        text,
-        use_autoprompt=True,
-        num_results=5
-    )
-
-    for a_result in search_results.results:
-        results.append((a_result.title, a_result.url))
-
-    return results
-
-
-def build_ui():
-    """
-    Display the input elements for content generation. Only covers the first step.
-    """
-
-    # get_disk_used_percentage()
-
-    st.title(APP_TEXT['app_name'])
-    st.subheader(APP_TEXT['caption'])
-    st.markdown(
-        'Powered by'
-        ' [Mistral-7B-Instruct-v0.2](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.2).'
-    )
-    st.markdown(
-        '*If the JSON is generated or parsed incorrectly, try again later by making minor changes'
-        ' to the input text.*'
-    )
-
-    with st.form('my_form'):
-        # Topic input
-        try:
-            with open(GlobalConfig.PRELOAD_DATA_FILE, 'r', encoding='utf-8') as in_file:
-                preload_data = json5.loads(in_file.read())
-        except (FileExistsError, FileNotFoundError):
-            preload_data = {'topic': '', 'audience': ''}
-
-        topic = st.text_area(
-            APP_TEXT['input_labels'][0],
-            value=preload_data['topic']
-        )
-
-        texts = list(GlobalConfig.PPTX_TEMPLATE_FILES.keys())
-        captions = [GlobalConfig.PPTX_TEMPLATE_FILES[x]['caption'] for x in texts]
-
-        pptx_template = st.radio(
-            'Select a presentation template:',
-            texts,
-            captions=captions,
-            horizontal=True
-        )
-
-        st.divider()
-        submit = st.form_submit_button('Generate slide deck')
-
-    if submit:
-        # st.write(f'Clicked {time.time()}')
-        st.session_state.submitted = True
-
-    # https://github.com/streamlit/streamlit/issues/3832#issuecomment-1138994421
-    if 'submitted' in st.session_state:
-        progress_text = 'Generating the slides...give it a moment'
-        progress_bar = st.progress(0, text=progress_text)
-
-        topic_txt = topic.strip()
-        generate_presentation(topic_txt, pptx_template, progress_bar)
-
-    st.divider()
-    st.text(APP_TEXT['tos'])
-    st.text(APP_TEXT['tos2'])
-
-    st.markdown(
-        '![Visitors]'
-        '(https://api.visitorbadge.io/api/visitors?path=https%3A%2F%2Fhuggingface.co%2Fspaces%2Fbarunsaha%2Fslide-deck-ai&countColor=%23263759)'
-    )
-
-
-def generate_presentation(topic: str, pptx_template: str, progress_bar):
-    """
-    Process the inputs to generate the slides.
-
-    :param topic: The presentation topic based on which contents are to be generated.
-    :param pptx_template: The PowerPoint template name to be used.
-    :param progress_bar: Progress bar from the page.
-    """
-
-    topic_length = len(topic)
-    logger.debug('Input length:: topic: %s', topic_length)
-
-    if topic_length >= 10:
-        logger.debug('Topic: %s', topic)
-        target_length = min(topic_length, GlobalConfig.LLM_MODEL_MAX_INPUT_LENGTH)
-
-        try:
-            # Step 1: Generate the contents in JSON format using an LLM
-            json_str = process_slides_contents(topic[:target_length], progress_bar)
-            logger.debug('Truncated topic: %s', topic[:target_length])
-            logger.debug('Length of JSON: %d', len(json_str))
-
-            # Step 2: Generate the slide deck based on the template specified
-            if len(json_str) > 0:
-                st.info(
-                    'Tip: The generated content doesn\'t look so great?'
-                    ' Need alternatives? Just change your description text and try again.',
-                    icon="💡️"
-                )
-            else:
-                st.error(
-                    'Unfortunately, JSON generation failed, so the next steps would lead'
-                    ' to nowhere. Try again or come back later.'
-                )
-                return
-
-            all_headers = generate_slide_deck(json_str, pptx_template, progress_bar)
-
-            # Step 3: Bonus stuff: Web references and AI art
-            show_bonus_stuff(all_headers)
-
-        except ValueError as ve:
-            st.error(f'Unfortunately, an error occurred: {ve}! '
-                     f'Please change the text, try again later, or report it, sharing your inputs.')
-
-    else:
-        st.error('Not enough information provided! Please be little more descriptive :)')
-
-
-def process_slides_contents(text: str, progress_bar: st.progress) -> str:
-    """
-    Convert given text into structured data and display. Update the UI.
-
-    :param text: The topic description for the presentation.
-    :param progress_bar: Progress bar for this step.
-    :return: The contents as a JSON-formatted string.
-    """
-
-    json_str = ''
-
-    try:
-        logger.info('Calling LLM for content generation on the topic: %s', text)
-        json_str = get_contents_wrapper(text)
-    except Exception as ex:
-        st.error(
-            f'An exception occurred while trying to convert to JSON. It could be because of heavy'
-            f' traffic or something else. Try doing it again or try again later.'
-            f'\nError message: {ex}'
-        )
-
-    progress_bar.progress(50, text='Contents generated')
-
-    with st.expander('The generated contents (in JSON format)'):
-        st.code(json_str, language='json')
-
-    return json_str
-
-
-def generate_slide_deck(json_str: str, pptx_template: str, progress_bar) -> List:
-    """
-    Create a slide deck.
-
-    :param json_str: The contents in JSON format.
-    :param pptx_template: The PPTX template name.
-    :param progress_bar: Progress bar.
-    :return: A list of all slide headers and the title.
-    """
-
-    progress_text = 'Creating the slide deck...give it a moment'
-    progress_bar.progress(75, text=progress_text)
-
-    # # Get a unique name for the file to save -- use the session ID
-    # ctx = st_sr.get_script_run_ctx()
-    # session_id = ctx.session_id
-    # timestamp = time.time()
-    # output_file_name = f'{session_id}_{timestamp}.pptx'
-
-    temp = tempfile.NamedTemporaryFile(delete=False, suffix='.pptx')
-    path = pathlib.Path(temp.name)
-
-    logger.info('Creating PPTX file...')
-    all_headers = pptx_helper.generate_powerpoint_presentation(
-        json_str,
-        slides_template=pptx_template,
-        output_file_path=path
-    )
-    progress_bar.progress(100, text='Done!')
-
-    with open(path, 'rb') as f:
-        st.download_button('Download PPTX file', f, file_name='Presentation.pptx')
-
-    if temp:
-        temp.close()
-
-    return all_headers
-
-
-def show_bonus_stuff(ppt_headers: List[str]):
-    """
-    Show bonus stuff for the presentation.
-
-    :param ppt_headers: A list of the slide headings.
-    """
-
-    # Use the presentation title and the slide headers to find relevant info online
-    logger.info('Calling Metaphor search...')
-    ppt_text = ' '.join(ppt_headers)
-    search_results = get_web_search_results_wrapper(ppt_text)
-    md_text_items = []
-
-    for (title, link) in search_results:
-        md_text_items.append(f'[{title}]({link})')
-
-    with st.expander('Related Web references'):
-        st.markdown('\n\n'.join(md_text_items))
-
-    logger.info('Done!')
-
-    # # Avoid image generation. It costs time and an API call, so just limit to the text generation.
-    # with st.expander('AI-generated image on the presentation topic'):
-    #     logger.info('Calling SDXL for image generation...')
-    #     # img_empty.write('')
-    #     # img_text.write(APP_TEXT['image_info'])
-    #     image = get_ai_image_wrapper(ppt_text)
-    #
-    #     if len(image) > 0:
-    #         image = base64.b64decode(image)
-    #         st.image(image, caption=ppt_text)
-    #         st.info('Tip: Right-click on the image to save it.', icon="💡️")
-    #         logger.info('Image added')
-
-
-def main():
-    """
-    Trigger application run.
-    """
-
-    build_ui()
-
-
-if __name__ == '__main__':
-    main()

helpers/llm_helper.py → llm_helper.py

@@ -1,10 +1,5 @@
 import logging
 import requests
-from requests.adapters import HTTPAdapter
-from urllib3.util import Retry
-
-from langchain_community.llms.huggingface_endpoint import HuggingFaceEndpoint
-from langchain_core.language_models import LLM
 
 from global_config import GlobalConfig
 
@@ -12,58 +7,28 @@ from global_config import GlobalConfig
 HF_API_URL = f"https://api-inference.huggingface.co/models/{GlobalConfig.HF_LLM_MODEL_NAME}"
 HF_API_HEADERS = {"Authorization": f"Bearer {GlobalConfig.HUGGINGFACEHUB_API_TOKEN}"}
 
-logger = logging.getLogger(__name__)
-
-retries = Retry(
-    total=5,
-    backoff_factor=0.25,
-    backoff_jitter=0.3,
-    status_forcelist=[502, 503, 504],
-    allowed_methods={'POST'},
+logging.basicConfig(
+    level=GlobalConfig.LOG_LEVEL,
+    format='%(asctime)s - %(message)s',
 )
-adapter = HTTPAdapter(max_retries=retries)
-http_session = requests.Session()
-http_session.mount('https://', adapter)
-http_session.mount('http://', adapter)
-
-
-def get_hf_endpoint() -> LLM:
-    """
-    Get an LLM via the HuggingFaceEndpoint of LangChain.
-
-    :return: The LLM.
-    """
-
-    logger.debug('Getting LLM via HF endpoint')
 
-    return HuggingFaceEndpoint(
-        repo_id=GlobalConfig.HF_LLM_MODEL_NAME,
-        max_new_tokens=GlobalConfig.LLM_MODEL_MAX_OUTPUT_LENGTH,
-        top_k=40,
-        top_p=0.95,
-        temperature=GlobalConfig.LLM_MODEL_TEMPERATURE,
-        repetition_penalty=1.03,
-        streaming=True,
-        huggingfacehub_api_token=GlobalConfig.HUGGINGFACEHUB_API_TOKEN,
-        return_full_text=False,
-        stop_sequences=['</s>'],
-    )
+# llm = None
 
 
-def hf_api_query(payload: dict) -> dict:
+def hf_api_query(payload: dict):
     """
     Invoke HF inference end-point API.
 
-    :param payload: The prompt for the LLM and related parameters.
-    :return: The output from the LLM.
+    :param payload: The prompt for the LLM and related parameters
+    :return: The output from the LLM
     """
 
     try:
-        response = http_session.post(HF_API_URL, headers=HF_API_HEADERS, json=payload, timeout=15)
+        response = requests.post(HF_API_URL, headers=HF_API_HEADERS, json=payload, timeout=15)
         result = response.json()
     except requests.exceptions.Timeout as te:
-        logger.error('*** Error: hf_api_query timeout! %s', str(te))
-        result = []
+        logging.error('*** Error: hf_api_query timeout! %s', str(te))
+        result = {}
 
     return result
 
@@ -72,8 +37,8 @@ def generate_slides_content(topic: str) -> str:
     """
     Generate the outline/contents of slides for a presentation on a given topic.
 
-    :param topic: Topic on which slides are to be generated.
-    :return: The content in JSON format.
+    :param topic: Topic on which slides are to be generated
+    :return: The content in JSON format
     """
 
     with open(GlobalConfig.SLIDES_TEMPLATE_FILE, 'r', encoding='utf-8') as in_file:
@@ -81,8 +46,8 @@ def generate_slides_content(topic: str) -> str:
         template_txt = template_txt.replace('<REPLACE_PLACEHOLDER>', topic)
 
     output = hf_api_query({
-        'inputs': template_txt,
-        'parameters': {
+        "inputs": template_txt,
+        "parameters": {
             'temperature': GlobalConfig.LLM_MODEL_TEMPERATURE,
             'min_length': GlobalConfig.LLM_MODEL_MIN_OUTPUT_LENGTH,
             'max_length': GlobalConfig.LLM_MODEL_MAX_OUTPUT_LENGTH,
@@ -91,7 +56,7 @@ def generate_slides_content(topic: str) -> str:
             'return_full_text': False,
             # "repetition_penalty": 0.0001
         },
-        'options': {
+        "options": {
             'wait_for_model': True,
             'use_cache': True
         }
@@ -105,7 +70,7 @@ def generate_slides_content(topic: str) -> str:
         # logging.debug(f'{json_end_idx=}')
         output = output[:json_end_idx]
 
-    logger.debug('generate_slides_content: output: %s', output)
+    logging.debug('generate_slides_content: output: %s', output)
 
     return output
 

pptx_helper.py

@@ -0,0 +1,254 @@
+import logging
+import pathlib
+import re
+import tempfile
+from typing import List, Tuple
+
+import json5
+import pptx
+import yaml
+
+from global_config import GlobalConfig
+
+
+PATTERN = re.compile(r"^slide[ ]+\d+:", re.IGNORECASE)
+SAMPLE_JSON_FOR_PPTX = '''
+{
+    "title": "Understanding AI",
+    "slides": [
+        {
+            "heading": "Introduction",
+            "bullet_points": [
+                "Brief overview of AI",
+                [
+                    "Importance of understanding AI"
+                ]
+            ]
+        }
+    ]
+}
+'''
+
+logging.basicConfig(
+    level=GlobalConfig.LOG_LEVEL,
+    format='%(asctime)s - %(message)s',
+)
+
+
+def remove_slide_number_from_heading(header: str) -> str:
+    """
+    Remove the slide number from a given slide header.
+
+    :param header: The header of a slide
+    """
+
+    if PATTERN.match(header):
+        idx = header.find(':')
+        header = header[idx + 1:]
+
+    return header
+
+
+def generate_powerpoint_presentation(
+        structured_data: str,
+        as_yaml: bool,
+        slides_template: str,
+        output_file_path: pathlib.Path
+) -> List:
+    """
+    Create and save a PowerPoint presentation file containing the contents in JSON or YAML format.
+
+    :param structured_data: The presentation contents as "JSON" (may contain trailing commas) or
+    YAML
+    :param as_yaml: True if the input data is in YAML format; False if it is in JSON format
+    :param slides_template: The PPTX template to use
+    :param output_file_path: The path of the PPTX file to save as
+    :return A list of presentation title and slides headers
+    """
+
+    if as_yaml:
+        # Avoid YAML mode: nested bullets can lead to incorrect YAML generation
+        try:
+            parsed_data = yaml.safe_load(structured_data)
+        except yaml.parser.ParserError as ype:
+            logging.error('*** YAML parse error: %s', str(ype))
+            parsed_data = {'title': '', 'slides': []}
+    else:
+        # The structured "JSON" might contain trailing commas, so using json5
+        parsed_data = json5.loads(structured_data)
+
+    logging.debug(
+        "*** Using PPTX template: %s",
+        GlobalConfig.PPTX_TEMPLATE_FILES[slides_template]['file']
+    )
+    presentation = pptx.Presentation(GlobalConfig.PPTX_TEMPLATE_FILES[slides_template]['file'])
+
+    # The title slide
+    title_slide_layout = presentation.slide_layouts[0]
+    slide = presentation.slides.add_slide(title_slide_layout)
+    title = slide.shapes.title
+    subtitle = slide.placeholders[1]
+    title.text = parsed_data['title']
+    logging.debug('Presentation title is: %s', title.text)
+    subtitle.text = 'by Myself and SlideDeck AI :)'
+    all_headers = [title.text, ]
+
+    # background = slide.background
+    # background.fill.solid()
+    # background.fill.fore_color.rgb = RGBColor.from_string('C0C0C0')  # Silver
+    # title.text_frame.paragraphs[0].font.color.rgb = RGBColor(0, 0, 128)  # Navy blue
+
+    # Add contents in a loop
+    for a_slide in parsed_data['slides']:
+        bullet_slide_layout = presentation.slide_layouts[1]
+        slide = presentation.slides.add_slide(bullet_slide_layout)
+        shapes = slide.shapes
+
+        title_shape = shapes.title
+        body_shape = shapes.placeholders[1]
+        title_shape.text = remove_slide_number_from_heading(a_slide['heading'])
+        all_headers.append(title_shape.text)
+        text_frame = body_shape.text_frame
+
+        # The bullet_points may contain a nested hierarchy of JSON arrays
+        # In some scenarios, it may contain objects (dictionaries) because the LLM generated so
+        #  ^ The second scenario is not covered
+
+        flat_items_list = get_flat_list_of_contents(a_slide['bullet_points'], level=0)
+
+        for an_item in flat_items_list:
+            paragraph = text_frame.add_paragraph()
+            paragraph.text = an_item[0]
+            paragraph.level = an_item[1]
+
+    # The thank-you slide
+    last_slide_layout = presentation.slide_layouts[0]
+    slide = presentation.slides.add_slide(last_slide_layout)
+    title = slide.shapes.title
+    title.text = 'Thank you!'
+
+    presentation.save(output_file_path)
+
+    return all_headers
+
+
+def get_flat_list_of_contents(items: list, level: int) -> List[Tuple]:
+    """
+    Flatten a (hierarchical) list of bullet points to a single list containing each item and its level.
+
+    :param items: A bullet point (string or list)
+    :param level: The current level of hierarchy
+    :return: A list of (bullet item text, hierarchical level) tuples
+    """
+
+    flat_list = []
+
+    for item in items:
+        if isinstance(item, str):
+            flat_list.append((item, level))
+        elif isinstance(item, list):
+            flat_list = flat_list + get_flat_list_of_contents(item, level + 1)
+
+    return flat_list
+
+
+if __name__ == '__main__':
+    # bullets = [
+    #     'Description',
+    #     'Types',
+    #     [
+    #         'Type A',
+    #         'Type B'
+    #     ],
+    #     'Grand parent',
+    #     [
+    #         'Parent',
+    #         [
+    #             'Grand child'
+    #         ]
+    #     ]
+    # ]
+
+    # output = get_flat_list_of_contents(bullets, level=0)
+    # for x in output:
+    #     print(x)
+
+    json_data = '''
+    {
+    "title": "Understanding AI",
+    "slides": [
+        {
+            "heading": "Introduction",
+            "bullet_points": [
+                "Brief overview of AI",
+                [
+                    "Importance of understanding AI"
+                ]
+            ]
+        },
+        {
+            "heading": "What is AI?",
+            "bullet_points": [
+                "Definition of AI",
+                [
+                    "Types of AI",
+                    [
+                        "Narrow or weak AI",
+                        "General or strong AI"
+                    ]
+                ],
+                "Differences between AI and machine learning"
+            ]
+        },
+        {
+            "heading": "How AI Works",
+            "bullet_points": [
+                "Overview of AI algorithms",
+                [
+                    "Types of AI algorithms",
+                    [
+                        "Rule-based systems",
+                        "Decision tree systems",
+                        "Neural networks"
+                    ]
+                ],
+                "How AI processes data"
+            ]
+        },
+        {
+            "heading": "Pros of AI",
+            "bullet_points": [
+                "Increased efficiency and productivity",
+                "Improved accuracy and precision",
+                "Enhanced decision-making capabilities",
+                "Personalized experiences"
+            ]
+        },
+        {
+            "heading": "Cons of AI",
+            "bullet_points": [
+                "Job displacement and loss of employment",
+                "Bias and discrimination",
+                "Privacy and security concerns",
+                "Dependence on technology"
+            ]
+        },
+        {
+            "heading": "Future Prospects of AI",
+            "bullet_points": [
+                "Advancements in fields such as healthcare and finance",
+                "Increased use"
+            ]
+        }
+    ]
+}'''
+
+    temp = tempfile.NamedTemporaryFile(delete=False, suffix='.pptx')
+    path = pathlib.Path(temp.name)
+
+    generate_powerpoint_presentation(
+        json5.loads(json_data),
+        as_yaml=False,
+        output_file_path=path,
+        slides_template='Blank'
+    )

requirements.txt

@@ -1,22 +1,12 @@
-aiohttp==3.9.5
 python-dotenv[cli]~=1.0.0
-gitpython==3.1.43
-json_repair==0.15.3
-idna==3.7
-jinja2==3.1.3
-Pillow==10.3.0
-pyarrow~=16.0.0
-pydantic==2.4.0
-langchain~=0.1.16
-langchain-core~=0.1.46
+langchain~=0.1.13
+# huggingface_hub
 streamlit~=1.32.2
+clarifai==9.7.4
 
 python-pptx
 metaphor-python
 json5~=0.9.14
-requests~=2.31.0
-
-transformers~=4.39.2
-langchain-community
-
-urllib3~=2.2.1
+PyYAML~=6.0.1
+# curlify
+requests~=2.31.0

strings.json

@@ -1,6 +1,6 @@
 {
-    "app_name": ":green[SlideDeck AI $^{[Reloaded]}$]",
-    "caption": "*Converse, create, and improve your next PowerPoint slide deck*",
+    "app_name": "SlideDeck AI",
+    "caption": "*:green[Co-create your next PowerPoint slide deck with AI]*",
     "section_headers": [
         "Step 1: Generate your content",
         "Step 2: Make it structured",
@@ -23,15 +23,7 @@
     ],
     "urls_info": "Here is a list of some online resources that you can consult for further information on this topic:",
     "image_info": "Got some more minutes? We are also trying to deliver an AI-generated art on the presentation topic, fresh off the studio, just for you!",
-    "content_generation_error": "Unfortunately, SlideDeck AI failed to generate any content for you! Please try again later.",
-    "json_parsing_error": "Unfortunately, SlideDeck AI failed to parse the response from LLM! Please try again by rephrasing the query or refreshing the page.",
+    "content_generation_failure_error": "Unfortunately, SlideDeck AI failed to generate any content for you! Please try again later.",
     "tos": "SlideDeck AI is an experimental prototype, and it has its limitations.\nPlease carefully review any and all AI-generated content.",
-    "tos2": "By using SlideDeck AI, you agree to fair and responsible usage.\nNo liability assumed by any party.",
-    "ai_greetings": [
-        "How may I help you today?",
-        "Stuck with creating your presentation? Let me help you.",
-        "Looks like you have a looming deadline. Can I help you get started with your slide deck?",
-        "Hello! What topic do you have on your mind today?"
-    ],
-    "chat_placeholder": "Write the topic or instructions here"
+    "tos2": "By using SlideDeck AI, you agree to fair and responsible usage.\nNo liability assumed by any party."
 }