Can [AI] be the all in one Assistant?

I started with streamlit with some sample apps for route gpx tracking

But this year I have been testing more features with the Streamlit MultiChat Summary.

Anyways, it was time to level up my Streamlit, again.

And why not doing it while making some Business value driven projects!

Looking at Streamlit Roadmap, more ideas will pop up.

Thanks to the investigations during

Let it be leverage

AI Assistant

Source Code for Streamlit-AIssistant

For Entrepreneurs

A Neuromarketing Agent

It is hard to be a solo-preneur…

unless you have some help.

Like the help of EntrepreAI

Many of these were just Tests during Streamlit-Multichat few months ago

Automatic Blog Creation

Using already a SSG for your websites?

If a LLM can write markdown…why not using it to create posts assisted with AI?

Youtube Stuff

There are great videos out there.

YT Summarizer and More

But there are hidden between…too much fluff.

But the PhiData team put together an awsome idea that I decided to Fork and bundle automatically in a Container for further reach…

…a youtube AI summarization tool which uses Groq for very fast inference.

I explained the process at this blog post

And now have been testing a little bit more the GroqAPI.

Of course there are new models: LLaMa3.1’s and LLaMa 3.2’s

docker pull ghcr.io/jalcocert/phidata:yt-groq #https://github.com/users/JAlcocerT/packages/container/package/phidata

version: '3.8'

services:
  phidata_service:
    image: ghcr.io/jalcocert/phidata:yt-groq #phidata:yt_summary_groq
    container_name: phidata_yt_groq
    ports:
      - "8502:8501"    
    environment:
      - GROQ_API_KEY="gsk_dummy-groq-api" # your_api_key_here
    command: streamlit run cookbook/llms/groq/video_summary/app.py
    restart: always
    networks:
      - cloudflare_tunnel

  streamlit_multichat:
    image: ghcr.io/jalcocert/streamlit-multichat:latest
    container_name: streamlit_multichat
    volumes:
      - ai_streamlit_multichat:/app
    working_dir: /app
    command: /bin/sh -c "\
            mkdir -p /app/.streamlit && \
            echo 'OPENAI_API_KEY = "sk-dummy-openai-key"' > /app/.streamlit/secrets.toml && \
            echo 'GROQ_API_KEY = "dummy-groq-key"' >> /app/.streamlit/secrets.toml && \
            echo 'ANTHROPIC_API_KEY = "sk-dummy-anthropic-key"' >> /app/.streamlit/secrets.toml && \
            streamlit run Z_multichat_Auth.py
    ports:
      - "8501:8501"
    networks:
      - cloudflare_tunnel
    restart: always
      # - nginx_default   

  ollama:
    image: ollama/ollama
    container_name: ollama
    ports:
      - "11434:11434" #Ollama API
    volumes:
      #- ollama_data:/root/.ollama
      - /home/Docker/AI/Ollama:/root/.ollama
    networks:
      - ollama_network      

  ollama-webui:
    image: ghcr.io/ollama-webui/ollama-webui:main
    container_name: ollama-webui
    ports:
      - "3000:8080" # 3000 is the UI port
    environment:
      - OLLAMA_BASE_URL=http://192.168.3.200:11434
    # add-host:
    #   - "host.docker.internal:host-gateway"
    volumes:
      - /home/Docker/AI/OllamaWebUI:/app/backend/data
    restart: always          
    networks:
      - ollama_network    

networks:
  cloudflare_tunnel:
    external: true
  ollama_network:
    external: false
  # nginx_default:
  #   external: true

volumes:
  ai_streamlit_multichat:

But that is old school already.

I have created an improved agent that allows us not just to summarize the YT Video - but to chat with a YouTube video content.

And it can be used via the groq API too.

• https://console.groq.com/docs/examples

• Cool sample: Ask questions with Groq + Function Calling + DuckDB SQl Queries- Example

You can also check this project - yt2doc - for YT to markdown transcription

Creating Audio for YT

Imagine that you have recorded some procedure that you want to share with the rest of the world.

But you want to automate the process of explaining it.

How about using AI to generate a .mp3 that will be explaining the video for you?

And actually…that text could be also AI generated. Also the Youtube Video Description.

Initially, I was using these simple scripts.

But these are much better…

Oh, and it uses OpenAI Text2Speech

You can also have realistic AI audio (TTS) generated with 11ElevenLabs or…Local Realtime voice AI con Ichigo

Video as a Code

Is it even possible to create a full .mp4 with AI?

In the meantime, I will stay with real life with the DJI Action cam and my repo for putting the videos together

Work Companion

Slides Creation Agent

There are really cool infographics out there.

Like infograpia.com or others

But better to have a ppt than nothing at all.

Well, but this wont be just “a powerpoint”. They can be sleek. And done with code.

Presentations as a code, noways means AI driven ppt’s.

SliDev is an awsome project to generate PPT as code - leveraging VueJS.

I love the SliDev project so much, that I forked it.

• OpenAI PPT with Slidev
• SliDev PPT for Streamlit Multichat - Deployed here using GH Actions
  • Result
  • Code

Diagrams with AI

Because mermaidJS is so cool not to use it also with AI.

You can use those diagrams with SSG (like HUGO, Astro…) and also for Slides creation with SliDev.

Interesting MermaidJS Features:

1. Make it look handDrawn
2. Markdown Strings
3. Support for FontAwsome Icons

flowchart TD
    B["Mermaid with fa:fa-robot "]
    B-->C["`The **cat**
  in the hat -  fa:fa-ban forbidden`"]
    B-->D(fa:fa-spinner loading)
    B-->E(A fa:fa-camera-retro perhaps?)

Ticket Creation with AI

Email with Ai

CV with AI

It all started with CV with OpenAI

Thanks to this app I have been able to level up my Streamlit with many new features.

And there are people already making money with similar tools:

• https://aiapply.co/
• https://enhancv.com/ai-resume-builder/

Speech Rater with AI

It all started with a friend doing public speaking courses.

And this idea appeared…

And since OpenAI can transcribe (Audio to Text)…

And Text can be summarized / be asked questions…

flowchart TD
    A[User opens Streamlit app] --> B[User records audio]
    B --> C[Audio sent to Whisper model for transcription]
    C --> D[Transcribed text]
    D --> E[Text sent to OpenAI GPT LLM]
    E --> F[Generated response from GPT]
    F --> G[User engages in live chat to improve speech]

Kindle Notes to AI

Kindle notes are saved into a .txt in the device.

I was storing notes for few years already and im really curious to see whats inside.

I gave it a try with Google’s NotebookLLM, but…custom is custom

Creating an Ebook

I mean, an AI Assisted e-book.

Conclusions

Potentially, you can choose any other LLMs via API to reply to you.

Its all about getting proper API calls to each of them.

I have chosen OpenAI API for the project.

But you have sample API calls to other models (via APIs) at the Streamlit Multichar project.

Making Streamlit Better

User Authentication

There will be some diagrams here, with Mermaid ofc.

flowchart TD
    Start([User Arrives at App 🙊]) --> LoginPrompt{User enters
username & password}
    
    LoginPrompt -->|Input redentials| AuthCheck["Check credentials
against Auth_functions.py"]

    AuthCheck -->|Valid credentials| Welcome["Display Welcome Message
and access to the app"]
    AuthCheck -->|Invalid credentials| LoginPrompt

    Welcome -->|Logout| LoginPrompt

But to place real users fast into the features loop…

It is interesting at least to get them to provide the email in exchange for getting early access.

flowchart TD
    Start([User Arrives at App 🙊]) --> DevModeCheck{Check DEV_MODE}
    
    DevModeCheck -->|DEV_MODE=True| MainOptions["App Features Access 😊 "]
    DevModeCheck -->|DEV_MODE=False| EmailEntry["Prompt user for email"]
    
    EmailEntry -->|Email Entered| SubscriptionCheck["Check email subscription with MailerLite API"]
    EmailEntry -->|No Email Entered| EndInfo["Show 'Please enter email' message"]

    SubscriptionCheck -->|Subscribed| SuccessWelcome["Welcome Message and Log Success"]
    SubscriptionCheck -->|Not Subscribed| Warning["Show warning: Email not subscribed and Log Failure"]

    SuccessWelcome --> MainOptions
    MainOptions --> OptionSelected{User selects an option}
    
    OptionSelected -->|Analyze CV Template| AnalyzeCV["Run Analyze CV function"]
    OptionSelected -->|CV To Offer| AdaptCV["Run Adapt CV to Offer function"]
    OptionSelected -->|Offer Overview| OfferOverview["Run Offer Overview function"]

    Warning --> Start

FormBricks is an open source Survey Platform….

flowchart TD
    Start([User Arrives at App 🙊]) --> DevModeCheck{BackEnd Checks DEV_MODE}
    
    DevModeCheck -->|DEV_MODE=True| MainOptions["App Features Access 😍"]
    DevModeCheck -->|DEV_MODE=False| EmailEntry["Prompt user for email"]
    
    EmailEntry -->|Email Entered| SubscriptionCheck
    EmailEntry -->|No Email Entered| EndInfo["Show 'Please enter email' message"]

    %% Subscription checks surrounded in a box
    subgraph BackendElements [BE Subscription Logic]
        SubscriptionCheck["Check email subscription with MailerLite API"]
        SubscriptionCheck -->|Subscribed via MailerLite| SuccessWelcome["Welcome Message and Log Success"]
        SubscriptionCheck -->|Not in MailerLite| FormBricksCheck["Check email in FormBricks Google Sheet"]
        
        FormBricksCheck -->|Subscribed in FormBricks| SuccessWelcome
        FormBricksCheck -->|Not Subscribed in Either| Warning["Show warning: Email not subscribed and Log Failure"]
    end

    SuccessWelcome --> MainOptions
    MainOptions --> OptionSelected{User selects an option}
    
    OptionSelected -->|Analyze CV Template| AnalyzeCV["Run Analyze CV function"]
    OptionSelected -->|CV To Offer| AdaptCV["Run Adapt CV to Offer function"]
    OptionSelected -->|Offer Overview| OfferOverview["Run Offer Overview function"]

    Warning --> Start

But…how about paywalls?

Some time ago I tried this streamlit paywall simple example

but it didnt quite work for me.

It was the time to check Stripe API together with Streamlit

In the end, I made a EmailWall Module that combines the 3 checks, according to what the user wants to check agains in the .env variable

graph TD
    A[is_email_subscribed]
    
    A -->|Check MailerLite| ML[check_mailerlite_subscription]
    ML -->|Subscribed| ML_Yes["MailerLite: Subscribed"]
    ML -->|Not Subscribed| ML_No["MailerLite: Not Subscribed"]
    
    A -->|Check FormBricks| FB[check_formbricks_subscription]
    FB -->|Subscribed| FB_Yes["FormBricks: Subscribed"]
    FB -->|Not Subscribed| FB_No["FormBricks: Not Subscribed"]
    
    A -->|Check Stripe| Stripe

    subgraph Orchestrator Stripe Check
        Stripe[orchestrator_stripe]
        Stripe -->|Check if customer email exists| SE[check_stripe_email]
        SE -- Yes --> SS[has_specific_active_or_trialing_subscription]
        SE -- No --> SNE["Stripe: Customer does not exist"]
        
        SS -->|Active/Trialing subscription found| SAT["Stripe: Subscribed"]
        SS -->|No active or trialing subscription found| SNA["Stripe: Not Subscribed"]
    end

If you are on a PoC stage, having a very simple and realiable way to bypass from any auth an email, is really helpful.

This is why I added such option with the module described below.

Streamlit New Features

Big thanks to Fanilo Andrianasolo for the great Streamlit Videos

• Keep up to date with the latest streamlit utilities
• And streamlit components

There is also this video explaining whats going to be released…now.

What Ive learnt with this one

#streamlit run st_scrapnsummarize.py --server.address 100.104.tail.scale
#streamlit run st_scrapnsummarize.py --server.address 100.104.tail.scale --server.port 8533

#pip install -r requirements.txt

#pip freeze | grep -E '^(anthropic|streamlit)=='
#pip freeze | grep -Ff requirements.txt
pip freeze | grep -E "^($(paste -sd '|' requirements.txt))=="

source .env
#export GROQ_API_KEY="your-api-key-here"
#set GROQ_API_KEY=your-api-key-here
#$env:GROQ_API_KEY="your-api-key-here"
echo $GROQ_API_KEY $OPENAI_API_KEY $ANTHROPIC_API_KEY

FAQ

How to Query Different APIs

You can also use OpenRouter API to query many different model providers…

LLMs - OpenAI,Anthropic, Groq…

• https://console.anthropic.com/workbench/
• https://console.groq.com/keys
• https://platform.openai.com/api-keys

Normally the latest Anthropic (Claude Models) are accesible very soon on AWS Bedrock API - As they invested a lot in the company

Streamlit CI/CD

• With Github - using dockerx to get a multiarch (x86 & ARM64) Container Image

To use Github Actions to create automatic container images for your streamlit projects, you will need such configuration file.

• You can also build the container image manually
  • Following this sample steps

Deploying Streamlit

You can use Streamlit Cloud Services.

But you can also host Streamlit with your own Domain.

And without paying extra.

Remember to use: container_name:container_port, not the port on the host.