⚡️ InfinityAI v1.0

"An assistant that can build stuff, imporve it and fix it without needing follow-up."

🧠 What is InfinityAI

InfinityAI is an AI systems that can:

• Understand what you say (voice or text)
• plan what needs to be done through python logic
• build real projects
• fix errors automatically
• run code
• open websites
• search the internet
• and give you result

It is your very own JARVIS. Plan to add more and get as close as to real JARVIS.

🚀 Features

🎤 Voice Mode

• Its more like how you talk to Chatgpt's voice but you can be able to build and search with just talking. Which I think Chat can't do and is very cool.
• Full voice interaction (will be added in v1.2)(It is going to be really good. Trust!)
• You can talk -> it listens -> it responds, searches and builds in a matter of seconds.

🧠 AI brain (This the m5 chip)

• Uses the incredible fast OpenAI models for part of reasoning. (No Glaze!)
• Maintains memory so it doesn't feel out of context (in \data folder)
• Fast responses for simple inputs through cache files
• Understands context and intent (Because of the memory, duh!)

⚙️ Smart Command System

• Commands feel instant and eaasy to remember
• Corrects when needed to using rapidfuzz library, this is only when trying to open webpages. I actually thought to send everything to OpenAI and then realized how slow (ms) it was and expensive it was. Pretty cool feature if wanting to reduce costs and save time. I also plan to expand the list more than it is now (coming right up v1.3)
• Uses AI only when needed

🌐 Web Control + Browsing

• Opens wbesites from the CLI and the very cool Glass Cockpit (UI)

🌐 Web Search + Data Extraction

• Searches the internet(no heavy APIs), Again for cost saving and latency delays.
• Reads real web pages so it is up to date
• Extracts useful information using a python logic (check core/web_agent.py)
• Summarizes results clearly using OpenAI. Again to save up costs I thought to use the least expensive version and about time delay, it takes time cause its web search.

📁 Project Generator (File Agent)

• Creates full projects from one prompt and hosts if it can be hosted or else simply say it can't be hosted.
• Multi-file structure - can create files and folders and even subfolders and even sub-subfolders and ...
• Clean and usable code that is bullet proof

🔧 Self-Healing System

• Runs generated code using subprocess
• Detects errors if any tend to exist (this sentence is really AI looking for me personally. This is the best sentence in the whole readme, Trust!)
• Fixes them automatically
• Reruns until it works

Basically a clean supervisor that checks for bugs and makes your idea project better.

▶️ Run & Preview

• Automatically runs generated apps
• Detects server (Flask, etc.) using keyword python logic (check out core/runner.py)
• Opens them in your browser

Boom! You just turned your idea into a project.

🧠 Autonomous Agent Loop

• Think -> Observe -> Improve (Inspired from JARVIS and OpenAI)
• Handles multistep tasks seemlessly
• Works toward goals without needing any follow_ups

⚡️ Perfomance Optimization

• Async voice (no latency delay). So this is a very unique, smart and thoughtful feature
• • So when using voice mode you have to wait for OpenAI to finish sending normal text and process it back to OpenAI TTS API and wait idle until it returns the audio before anything can be done(fyi this takes about a minute). Knowing that it is time consuming. It is also makes it feel laggy and less smart and engaging. Basically making it not a conversation.
• • Instead of waiting for the full response to come back I made them into chunks of text (based upon '.') and buffering them into a variable and sending to the TTS model without waiting for the rest and making it speak while the response model is still responding
• • The result after this made voice replies feel instant and more engaging.
• Fast-path responses (no unnecessary API calls)
• Smart execution system
• Designed for speed and effeciency by the creator of IfinityAI (thats me)

💻 UI Interface (Ispired totally from Stark Techology)

• Streamlit-based interface
• Chat-style interaction with stats and command pallete
• Clean and simple design, very hard to get lost
• Real-time feedback
• Used Codex for creating part of it

How the system works

Voice / Text Input
      ↓
        AI Brain (LLM)
            ↓
            Agent System    
                    ↓
                    File Generator / Web / Commands
                                    ↓
                                    Self Healing Looop
                                        ↓
                                        Voice Output

🛠️ Made Of

• Python
• OpenAI (LLM + TTS +STT)
• Streamlit
• SoundDevice (audio parsing)
• BeautifulSoup (web parsing)
• Subprocess (execution engine)

⁉️ Why InfinityAI?

Most AI tools:

• give answers

InfintiyAI:

• builds real projects
• fixes project automatically
• runs code
• opens websites
• searches the web
• completes tasks all from one command.

🤖 AI USAGE

InfintyAI uses AI tools such as OpenAI (including Codex and language models) for:

• (part of) reasoning and reponses
• code generation
• speech processing (Both STT and TTS)

The system design, architecture, agent loop, file generation system, self-healing mechanism, exectution pipeline, and integrations were designed and implemented by me.

AI is used as a tool inside the system rather than replacement for the system itself.

Shipping 🚀 InfintyAI v1.1.9 - minor bug fixes and reworked most of the files

• shipped v1.1
• shipped subversion v1.1 - major bug fixes

Final Line

"Don't just ask AI. Let it build, fix, run and find" - InfinityAI