<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>DEV Community: Riddhi Rawool</title>
    <description>The latest articles on DEV Community by Riddhi Rawool (@riddhi_rawool_3b9f2e1b2da).</description>
    <link>https://dev.to/riddhi_rawool_3b9f2e1b2da</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4188772%2F3852ea7d-f65b-4628-a597-bca387be1db3.png</url>
      <title>DEV Community: Riddhi Rawool</title>
      <link>https://dev.to/riddhi_rawool_3b9f2e1b2da</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/riddhi_rawool_3b9f2e1b2da"/>
    <language>en</language>
    <item>
      <title>EchoVault: Some Memories Refuse to Disappear 👻</title>
      <dc:creator>Riddhi Rawool</dc:creator>
      <pubDate>Sun, 01 Nov 2026 04:28:50 +0000</pubDate>
      <link>https://dev.to/riddhi_rawool_3b9f2e1b2da/echovault-some-memories-refuse-to-disappear-2eoe</link>
      <guid>https://dev.to/riddhi_rawool_3b9f2e1b2da/echovault-some-memories-refuse-to-disappear-2eoe</guid>
      <description>&lt;p&gt;&lt;em&gt;This is a submission for the &lt;a href="https://dev.to/challenges/hacktoberfest-week4-2026-10-26"&gt;Hacktoberfest Open-Source AI Challenge Week 4: Ghost in the Machine&lt;/a&gt;&lt;/em&gt;&lt;/p&gt;

&lt;h1&gt;
  
  
  EchoVault: Some Memories Refuse to Disappear
&lt;/h1&gt;

&lt;p&gt;&lt;strong&gt;What if the voices of yesterday could still answer the questions of tomorrow?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;A voice recording is more than a collection of sounds. It can hold a grandparent's stories, a forgotten family tradition, the name of a place, or a moment that deserves to be remembered.&lt;/p&gt;

&lt;p&gt;But memories are fragile.&lt;/p&gt;

&lt;p&gt;Recordings get buried in folders. Important details become difficult to find. Stories that once brought people together can slowly fade into silence.&lt;/p&gt;

&lt;p&gt;I started thinking about one possibility:&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;What if we could preserve these voices, organize their stories, and make them searchable without allowing AI to rewrite what was actually said?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;That question became &lt;strong&gt;EchoVault&lt;/strong&gt;.&lt;/p&gt;

&lt;h2&gt;
  
  
  What I Built
&lt;/h2&gt;

&lt;p&gt;EchoVault is a local-first AI memory vault that transforms family voice recordings into searchable, structured memories.&lt;/p&gt;

&lt;p&gt;Instead of simply storing audio, it uses AI to identify meaningful details such as people, places, traditions and important moments. It keeps the original transcript at the centre of the experience so that users can revisit the words behind every preserved memory.&lt;/p&gt;

&lt;p&gt;My interpretation of the theme &lt;em&gt;Ghost in the Machine&lt;/em&gt; is about keeping something deeply human alive through technology.&lt;/p&gt;

&lt;p&gt;The ghost is the echo of a memory that refuses to disappear.&lt;/p&gt;

&lt;h3&gt;
  
  
  The journey of a memory
&lt;/h3&gt;

&lt;p&gt;A recording enters EchoVault as a voice. It passes through transcription, memory extraction and evidence verification before becoming a searchable story.&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;                    VOICE RECORDING
                          |
                          ▼
                   FASTER-WHISPER
                    TRANSCRIPTION
                          |
                          ▼
                    TRANSCRIPT REVIEW
                          |
                          ▼
                     QWEN 2.5
                  MEMORY EXTRACTION
                          |
             ┌────────────┼────────────┐
             ▼            ▼            ▼
          PEOPLE        PLACES     TRADITIONS
                          |
                          ▼
                       MOMENTS
                          |
                          ▼
                    GHOSTLIGHT
                 SOURCE EVIDENCE
                          |
                          ▼
                   ASK THIS MEMORY
                          |
                          ▼
                    MEMORY VAULT
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;

&lt;h3&gt;
  
  
  Four dimensions of every story
&lt;/h3&gt;

&lt;p&gt;EchoVault organizes extracted details into four memory categories.&lt;br&gt;
&lt;/p&gt;
&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;MEMORY CATEGORY       FEATURE MAP

People                ████████████████████
Places                ████████████████████
Traditions            ████████████████████
Moments               ████████████████████
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;


&lt;p&gt;This is a visual representation of the four extraction categories, not a performance benchmark.&lt;/p&gt;
&lt;h3&gt;
  
  
  The features that make EchoVault meaningful
&lt;/h3&gt;

&lt;p&gt;*&lt;em&gt;1. Voice to Memory *&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;Users upload a voice recording. faster-whisper converts the speech into text locally. The transcript remains editable, allowing users to correct recognition errors before preserving the story.&lt;/p&gt;

&lt;p&gt;*&lt;em&gt;2. AI Memory Extraction *&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;Qwen 2.5 identifies details explicitly mentioned in the transcript, including:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;People and names&lt;/li&gt;
&lt;li&gt;Places&lt;/li&gt;
&lt;li&gt;Family traditions&lt;/li&gt;
&lt;li&gt;Important moments&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;The extracted information is organized into a structured memory.&lt;/p&gt;

&lt;p&gt;*&lt;em&gt;3. Ghostlight: The Evidence Layer *&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;This is one of the most important features of EchoVault.&lt;/p&gt;

&lt;p&gt;AI can organize and interpret a story, but it can also misunderstand information. How can users verify what it extracts?&lt;/p&gt;

&lt;p&gt;Ghostlight connects extracted details to supporting sentences from the original transcript.&lt;/p&gt;

&lt;p&gt;Instead of asking people to blindly trust an AI-generated result, it gives them a way to revisit the actual words behind a memory.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;4. Ask This Memory 💬&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;What was the name of the place? Who was mentioned? What happened during a particular moment?&lt;/p&gt;

&lt;p&gt;Users can ask questions about a preserved story.&lt;/p&gt;

&lt;p&gt;The answers are designed to remain grounded in the original transcript. When a requested detail is missing, EchoVault responds:&lt;/p&gt;

&lt;blockquote&gt;
&lt;p&gt;That detail isn't present in this memory.&lt;/p&gt;
&lt;/blockquote&gt;

&lt;p&gt;Because an invented memory is not a preserved memory.&lt;/p&gt;

&lt;p&gt;*&lt;em&gt;5. The Memory Vault *&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;Preserved stories can be revisited through the archive, bringing their people, places, traditions and moments together in one place.&lt;/p&gt;
&lt;h2&gt;
  
  
  Demo
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;See how EchoVault works:&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;&lt;a href="https://youtu.be/11lBCkpBox0" rel="noopener noreferrer"&gt;https://youtu.be/11lBCkpBox0&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;The demonstration showcases the application workflow, from voice transcription to AI memory extraction, Ghostlight evidence and question answering.&lt;/p&gt;
&lt;h2&gt;
  
  
  Code
&lt;/h2&gt;

&lt;p&gt;EchoVault is an open-source project.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;GitHub Repository:&lt;/strong&gt;&lt;br&gt;&lt;br&gt;
&lt;a href="https://github.com/R-iddhi/Echovault" rel="noopener noreferrer"&gt;https://github.com/R-iddhi/Echovault&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;GitHub Username:&lt;/strong&gt; &lt;code&gt;R-iddhi&lt;/code&gt;&lt;/p&gt;

&lt;p&gt;Explore the source code and project setup instructions:&lt;/p&gt;


&lt;div class="ltag-github-readme-tag"&gt;
  &lt;div class="readme-overview"&gt;
    &lt;h2&gt;
      &lt;img src="https://assets.dev.to/assets/github-logo-5a155e1f9a670af7944dd5e12375bc76ed542ea80224905ecaf878b9157cdefc.svg" alt="GitHub logo"&gt;
      &lt;a href="https://github.com/R-iddhi" rel="noopener noreferrer"&gt;
        R-iddhi
      &lt;/a&gt; / &lt;a href="https://github.com/R-iddhi/Echovault" rel="noopener noreferrer"&gt;
        Echovault
      &lt;/a&gt;
    &lt;/h2&gt;
    &lt;h3&gt;
      A local-first AI memory vault that preserves family voice recordings, extracts people, places, traditions and moments, and answers questions with source-backed evidence.
    &lt;/h3&gt;
  &lt;/div&gt;
  &lt;div class="ltag-github-body"&gt;
    
&lt;div id="readme" class="md"&gt;&lt;div class="markdown-heading"&gt;
&lt;h1 class="heading-element"&gt;EchoVault&lt;/h1&gt;
&lt;/div&gt;

&lt;p&gt;&lt;strong&gt;Some memories refuse to disappear.&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;EchoVault is a local-first AI memory vault that turns family voice recordings into searchable stories. It transcribes audio, extracts people, places, traditions and moments, and answers questions using evidence from the original transcript.&lt;/p&gt;

&lt;div class="markdown-heading"&gt;
&lt;h2 class="heading-element"&gt;Try EchoVault on your computer&lt;/h2&gt;
&lt;/div&gt;

&lt;blockquote&gt;
&lt;p&gt;&lt;strong&gt;Current availability:&lt;/strong&gt; EchoVault is a local application, not a hosted website yet. The frontend, API, Whisper transcription and Ollama model run on your own computer. This keeps the AI processing local, but it means you need to complete the setup below before opening the app.&lt;/p&gt;
&lt;/blockquote&gt;

&lt;div class="markdown-heading"&gt;
&lt;h3 class="heading-element"&gt;1. Install the prerequisites&lt;/h3&gt;
&lt;/div&gt;

&lt;p&gt;Install:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;&lt;a href="https://www.python.org/downloads/" rel="nofollow noopener noreferrer"&gt;Python&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;
&lt;a href="https://nodejs.org/" rel="nofollow noopener noreferrer"&gt;Node.js&lt;/a&gt; (includes npm)&lt;/li&gt;
&lt;li&gt;&lt;a href="https://ollama.com/download" rel="nofollow noopener noreferrer"&gt;Ollama&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;

&lt;div class="markdown-heading"&gt;
&lt;h3 class="heading-element"&gt;2. Download the project&lt;/h3&gt;

&lt;/div&gt;

&lt;p&gt;Open PowerShell or a terminal:&lt;/p&gt;

&lt;div class="highlight highlight-source-powershell notranslate position-relative overflow-auto js-code-highlight"&gt;
&lt;pre&gt;git clone https:&lt;span class="pl-k"&gt;//&lt;/span&gt;&lt;span class="pl-c1"&gt;github.com&lt;/span&gt;&lt;span class="pl-k"&gt;/&lt;/span&gt;R&lt;span class="pl-k"&gt;-&lt;/span&gt;iddhi&lt;span class="pl-k"&gt;/&lt;/span&gt;Echovault.git
cd Echovault&lt;/pre&gt;

&lt;/div&gt;
&lt;div class="markdown-heading"&gt;
&lt;h3 class="heading-element"&gt;3. Set up the Python backend&lt;/h3&gt;

&lt;/div&gt;

&lt;p&gt;On Windows PowerShell:&lt;/p&gt;

&lt;div class="highlight highlight-source-powershell notranslate position-relative overflow-auto js-code-highlight"&gt;
&lt;pre&gt;python &lt;span class="pl-k"&gt;-&lt;/span&gt;m venv .venv
.\.venv\Scripts\&lt;span class="pl-c1"&gt;python.exe&lt;/span&gt; &lt;span class="pl-k"&gt;-&lt;/span&gt;m pip install &lt;span class="pl-k"&gt;-&lt;/span&gt;r backend&lt;span class="pl-k"&gt;/&lt;/span&gt;requirements.txt&lt;/pre&gt;

&lt;/div&gt;
&lt;p&gt;On macOS or Linux:&lt;/p&gt;
&lt;div class="highlight highlight-source-shell notranslate position-relative overflow-auto js-code-highlight"&gt;…
&lt;/div&gt;&lt;/div&gt;
  &lt;/div&gt;
  &lt;div class="gh-btn-container"&gt;&lt;a class="gh-btn" href="https://github.com/R-iddhi/Echovault" rel="noopener noreferrer"&gt;View on GitHub&lt;/a&gt;&lt;/div&gt;
&lt;/div&gt;



&lt;h2&gt;
  
  
  How I Built It
&lt;/h2&gt;

&lt;p&gt;EchoVault combines a web interface, a Python API and locally running AI models.&lt;/p&gt;

&lt;h3&gt;
  
  
  System architecture
&lt;/h3&gt;

&lt;p&gt;The application has two main processing pathways that work together.&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;                   USER
                     |
                     ▼
             REACT + VITE
                FRONTEND
                     |
                     ▼
                FASTAPI
                 BACKEND
                     |
          ┌──────────┴──────────┐
          |                     |
          ▼                     ▼
    AUDIO PROCESSING       AI PROCESSING
          |                     |
          ▼                     ▼
    FASTER-WHISPER           OLLAMA
          |                     |
          ▼                     ▼
     TRANSCRIPT             QWEN 2.5
          |                     |
          └──────────┬──────────┘
                     |
                     ▼
              STRUCTURED MEMORY
                     |
            ┌────────┴────────┐
            ▼                 ▼
       GHOSTLIGHT        ASK MEMORY
      SOURCE EVIDENCE   GROUNDED ANSWERS
            |                 |
            └────────┬────────┘
                     |
                     ▼
                MEMORY VAULT
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;h3&gt;
  
  
  Technology stack
&lt;/h3&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Layer&lt;/th&gt;
&lt;th&gt;Technology&lt;/th&gt;
&lt;th&gt;Purpose&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Frontend&lt;/td&gt;
&lt;td&gt;React, Vite&lt;/td&gt;
&lt;td&gt;User interface&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;UI animations&lt;/td&gt;
&lt;td&gt;Framer Motion&lt;/td&gt;
&lt;td&gt;Interactive experience&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;API communication&lt;/td&gt;
&lt;td&gt;Axios&lt;/td&gt;
&lt;td&gt;Frontend and backend communication&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Backend&lt;/td&gt;
&lt;td&gt;Python, FastAPI&lt;/td&gt;
&lt;td&gt;Application logic&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Transcription&lt;/td&gt;
&lt;td&gt;faster-whisper&lt;/td&gt;
&lt;td&gt;Audio-to-text conversion&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Language model&lt;/td&gt;
&lt;td&gt;Qwen 2.5&lt;/td&gt;
&lt;td&gt;Memory extraction and questions&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Local inference&lt;/td&gt;
&lt;td&gt;Ollama&lt;/td&gt;
&lt;td&gt;Running Qwen locally&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;h3&gt;
  
  
  How can someone try EchoVault?
&lt;/h3&gt;

&lt;p&gt;The current version runs locally on the user's computer. It is not yet a publicly hosted, one-click website.&lt;/p&gt;

&lt;p&gt;Anyone interested in exploring it can clone the public repository and follow the setup instructions.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Requirements:&lt;/strong&gt;&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Python&lt;/li&gt;
&lt;li&gt;Node.js and npm&lt;/li&gt;
&lt;li&gt;Ollama&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;&lt;strong&gt;Step 1: Clone the repository&lt;/strong&gt;&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;git clone https://github.com/R-iddhi/Echovault.git
&lt;span class="nb"&gt;cd &lt;/span&gt;Echovault
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;&lt;strong&gt;Step 2: Set up the Python backend&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;On Windows PowerShell:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight powershell"&gt;&lt;code&gt;&lt;span class="n"&gt;python&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="nt"&gt;-m&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="nx"&gt;venv&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="o"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;venv&lt;/span&gt;&lt;span class="w"&gt;
&lt;/span&gt;&lt;span class="o"&gt;.&lt;/span&gt;&lt;span class="nx"&gt;\.venv\Scripts\python.exe&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="nt"&gt;-m&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="nx"&gt;pip&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="nx"&gt;install&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="nt"&gt;-r&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="nx"&gt;backend/requirements.txt&lt;/span&gt;&lt;span class="w"&gt;
&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;&lt;strong&gt;Step 3: Download the language model&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Make sure Ollama is running:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;ollama pull qwen2.5:3b
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;&lt;strong&gt;Step 4: Start the backend&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;From the project root:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight powershell"&gt;&lt;code&gt;&lt;span class="o"&gt;.&lt;/span&gt;&lt;span class="n"&gt;\.venv\Scripts\python.exe&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="nt"&gt;-m&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="nx"&gt;uvicorn&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="nx"&gt;backend.main:app&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="nt"&gt;--host&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="nx"&gt;127.0.0.1&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="nt"&gt;--port&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="nx"&gt;8000&lt;/span&gt;&lt;span class="w"&gt;
&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Keep this terminal running.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Step 5: Start the frontend&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Open a second terminal:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight powershell"&gt;&lt;code&gt;&lt;span class="n"&gt;cd&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="nx"&gt;frontend&lt;/span&gt;&lt;span class="w"&gt;
&lt;/span&gt;&lt;span class="n"&gt;npm&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="nx"&gt;install&lt;/span&gt;&lt;span class="w"&gt;
&lt;/span&gt;&lt;span class="n"&gt;npm&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="nx"&gt;run&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="nx"&gt;dev&lt;/span&gt;&lt;span class="w"&gt;
&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Open the local address shown by Vite, usually:&lt;/p&gt;

&lt;p&gt;&lt;code&gt;http://localhost:5173&lt;/code&gt;&lt;/p&gt;

&lt;p&gt;Keep the frontend, backend and Ollama running while using EchoVault.&lt;/p&gt;

&lt;p&gt;On the first run, faster-whisper may download its transcription model. An internet connection is required for the initial model downloads.&lt;/p&gt;

&lt;h3&gt;
  
  
  The user experience
&lt;/h3&gt;

&lt;p&gt;Once the application is running, the actual journey is simple:&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Upload Recording → Review Transcript → Preserve Memory → Explore Details → Verify Evidence → Ask Questions&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;The current implementation performs its core AI processing on the user's own computer rather than depending on a publicly hosted AI backend.&lt;/p&gt;

&lt;h2&gt;
  
  
  Why Does Open Innovation Matter?
&lt;/h2&gt;

&lt;p&gt;Family memories can contain personal stories, names, relationships and details that people may not want to send to external AI services.&lt;/p&gt;

&lt;p&gt;That made local inference an important design decision for EchoVault.&lt;/p&gt;

&lt;p&gt;I used:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;faster-whisper&lt;/strong&gt; for local speech transcription.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Qwen 2.5&lt;/strong&gt;, an open-weight language model, through &lt;strong&gt;Ollama&lt;/strong&gt; for local memory extraction and question answering.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Open innovation gives developers the freedom to experiment with models, modify prompts, explore different approaches and build around their own requirements.&lt;/p&gt;

&lt;p&gt;For EchoVault, this means the core AI processing does not require a remote AI inference API once the necessary models and dependencies have been installed.&lt;/p&gt;

&lt;p&gt;It also reduces dependence on closed AI services and makes experimentation more accessible.&lt;/p&gt;

&lt;p&gt;Most importantly, open innovation allowed me to explore a principle that became central to the project:&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;AI can interpret the memory. It cannot rewrite the memory.&lt;/strong&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  💭 The Spirit of It
&lt;/h2&gt;

&lt;p&gt;Imagine someone telling a story today.&lt;/p&gt;

&lt;p&gt;A grandparent speaks about a place that no longer looks the same. Someone remembers a family celebration, a childhood tradition, or a person whose voice is no longer heard every day.&lt;/p&gt;

&lt;p&gt;Years later, another family member might wonder:&lt;/p&gt;

&lt;p&gt;&lt;em&gt;Where did that story happen? Who was there? What was that tradition?&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;The recording still exists, but finding one particular detail inside hours of audio can be difficult.&lt;/p&gt;

&lt;p&gt;EchoVault tries to bridge that gap.&lt;/p&gt;

&lt;p&gt;The voice remains the original memory. The transcript makes it readable. AI makes it searchable. Ghostlight brings the user back to the actual words.&lt;/p&gt;

&lt;p&gt;I did not want to build another chatbot that confidently answers everything.&lt;/p&gt;

&lt;p&gt;I wanted to build something that understands the importance of saying, &lt;em&gt;I cannot find that detail in this story.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;Because sometimes, preserving what was never said is just as important as preserving what was.&lt;/p&gt;

&lt;p&gt;This project is about more than storing recordings. It is about giving forgotten stories another chance to be heard.&lt;/p&gt;

&lt;p&gt;*&lt;em&gt;Some memories refuse to disappear. *&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Final Thought
&lt;/h2&gt;

&lt;p&gt;A machine can process thousands of words, but the value of a memory is not measured by the number of words it contains.&lt;/p&gt;

&lt;p&gt;It is measured by the people, places and moments it keeps alive.&lt;/p&gt;

&lt;p&gt;Built for &lt;strong&gt;Hacktoberfest 2026, Week 4: Ghost in the Machine&lt;/strong&gt;.&lt;/p&gt;

&lt;h1&gt;
  
  
  devchallenge #hf26challenge #Hacktoberfest2026 #OpenSource #AI #EchoVault
&lt;/h1&gt;

</description>
      <category>devchallenge</category>
      <category>hf26challenge</category>
      <category>ai</category>
      <category>machinelearning</category>
    </item>
    <item>
      <title>The App Worked. The Launch Wasn't Ready. So I Built LaunchLazy</title>
      <dc:creator>Riddhi Rawool</dc:creator>
      <pubDate>Thu, 22 Oct 2026 14:23:09 +0000</pubDate>
      <link>https://dev.to/riddhi_rawool_3b9f2e1b2da/the-app-worked-the-launch-wasnt-ready-so-i-built-launchlazy-25b4</link>
      <guid>https://dev.to/riddhi_rawool_3b9f2e1b2da/the-app-worked-the-launch-wasnt-ready-so-i-built-launchlazy-25b4</guid>
      <description>&lt;p&gt;&lt;em&gt;This is a submission for the &lt;a href="https://dev.to/challenges/hacktoberfest-week3-2026-10-19"&gt;Hacktoberfest Open-Source AI Challenge Week 3: Slow and Steady&lt;/a&gt;.&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  What I Built
&lt;/h2&gt;

&lt;p&gt;You know that moment when your project finally runs and you immediately feel like:&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;“Okay. Done.”&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Then five minutes later:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Did I leave a localhost URL somewhere?&lt;/li&gt;
&lt;li&gt;Is &lt;code&gt;.env&lt;/code&gt; ignored?&lt;/li&gt;
&lt;li&gt;Did I document Python dependencies?&lt;/li&gt;
&lt;li&gt;Is there a build script?&lt;/li&gt;
&lt;li&gt;Is CORS too open?&lt;/li&gt;
&lt;li&gt;Did I actually explain how to run this project?&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;That was the problem I wanted to solve.&lt;/p&gt;

&lt;p&gt;So I built &lt;strong&gt;LaunchLazy&lt;/strong&gt; — a local launch-readiness auditor that checks a software project before deployment and tells you what deserves attention first.&lt;/p&gt;

&lt;p&gt;The flow is simple:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;Project Folder
     ↓
Deterministic Checks
     ↓
Score + Risks
     ↓
Qwen 2.5 via Ollama
     ↓
Explanation + Next Actions
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;


&lt;p&gt;LaunchLazy checks things like README quality, environment files, &lt;code&gt;.gitignore&lt;/code&gt;, build scripts, tests, dependency files, localhost URLs, CORS settings, basic secret patterns, and deployment configuration.&lt;/p&gt;

&lt;p&gt;It then gives a readiness score, priority level, affected files, recommended fixes, and a downloadable report.&lt;/p&gt;

&lt;p&gt;The idea is not to replace a real security review.&lt;/p&gt;

&lt;p&gt;It is simply to catch the boring things developers often forget before launch.&lt;/p&gt;


&lt;h2&gt;
  
  
  Demo
&lt;/h2&gt;

&lt;p&gt;For the demo, I scanned another project instead of only testing LaunchLazy on itself.&lt;/p&gt;

&lt;p&gt;It found:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;28/100 readiness score&lt;/strong&gt;&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;High priority&lt;/strong&gt;&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;7 automated risks&lt;/strong&gt;&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;including missing build/test setup, undocumented Python dependencies, a hard-coded localhost URL, and missing deployment configuration.&lt;/p&gt;

&lt;p&gt;  &lt;iframe src="https://www.youtube.com/embed/LSZs7G0Xq0k" width="710" height="399"&gt;
  &lt;/iframe&gt;
&lt;/p&gt;


&lt;h2&gt;
  
  
  Code
&lt;/h2&gt;


&lt;div class="ltag-github-readme-tag"&gt;
  &lt;div class="readme-overview"&gt;
    &lt;h2&gt;
      &lt;img src="https://assets.dev.to/assets/github-logo-5a155e1f9a670af7944dd5e12375bc76ed542ea80224905ecaf878b9157cdefc.svg" alt="GitHub logo"&gt;
      &lt;a href="https://github.com/R-iddhi" rel="noopener noreferrer"&gt;
        R-iddhi
      &lt;/a&gt; / &lt;a href="https://github.com/R-iddhi/launchlazy" rel="noopener noreferrer"&gt;
        launchlazy
      &lt;/a&gt;
    &lt;/h2&gt;
    &lt;h3&gt;
      Build LaunchLazy AI-assisted launch readiness auditor
    &lt;/h3&gt;
  &lt;/div&gt;
  &lt;div class="ltag-github-body"&gt;
    
&lt;div id="readme" class="md"&gt;&lt;div class="markdown-heading"&gt;
&lt;h1 class="heading-element"&gt;LaunchLazy&lt;/h1&gt;
&lt;/div&gt;

&lt;p&gt;LaunchLazy is an AI-assisted software launch-readiness auditor.&lt;/p&gt;

&lt;p&gt;It scans a local software project, detects common launch risks, calculates a readiness score, and uses a local open-weight model through Ollama to explain what should be fixed next.&lt;/p&gt;

&lt;div class="markdown-heading"&gt;
&lt;h2 class="heading-element"&gt;What LaunchLazy Does&lt;/h2&gt;
&lt;/div&gt;

&lt;p&gt;LaunchLazy checks project structure and configuration for issues such as:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;missing README documentation&lt;/li&gt;
&lt;li&gt;missing environment variable templates&lt;/li&gt;
&lt;li&gt;missing or weak &lt;code&gt;.gitignore&lt;/code&gt;
&lt;/li&gt;
&lt;li&gt;missing build scripts&lt;/li&gt;
&lt;li&gt;missing test scripts&lt;/li&gt;
&lt;li&gt;missing Node lockfiles&lt;/li&gt;
&lt;li&gt;undocumented Python dependencies&lt;/li&gt;
&lt;li&gt;hard-coded localhost URLs&lt;/li&gt;
&lt;li&gt;risky wildcard CORS configuration&lt;/li&gt;
&lt;li&gt;possible hard-coded secrets&lt;/li&gt;
&lt;li&gt;missing deployment configuration&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;It then generates:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;detected technology stack&lt;/li&gt;
&lt;li&gt;frontend/backend component breakdown&lt;/li&gt;
&lt;li&gt;launch-readiness score&lt;/li&gt;
&lt;li&gt;prioritized risk list&lt;/li&gt;
&lt;li&gt;AI-assisted explanation&lt;/li&gt;
&lt;li&gt;downloadable Markdown launch report&lt;/li&gt;
&lt;/ul&gt;

&lt;div class="markdown-heading"&gt;
&lt;h2 class="heading-element"&gt;Tech Stack&lt;/h2&gt;
&lt;/div&gt;

&lt;div class="markdown-heading"&gt;
&lt;h3 class="heading-element"&gt;Frontend&lt;/h3&gt;

&lt;/div&gt;

&lt;ul&gt;
&lt;li&gt;React&lt;/li&gt;
&lt;li&gt;Vite&lt;/li&gt;
&lt;li&gt;Axios&lt;/li&gt;
&lt;/ul&gt;

&lt;div class="markdown-heading"&gt;
&lt;h3 class="heading-element"&gt;Backend&lt;/h3&gt;

&lt;/div&gt;

&lt;ul&gt;
&lt;li&gt;Python&lt;/li&gt;
&lt;li&gt;FastAPI&lt;/li&gt;
&lt;li&gt;Uvicorn&lt;/li&gt;
&lt;li&gt;Requests&lt;/li&gt;
&lt;/ul&gt;

&lt;div class="markdown-heading"&gt;
&lt;h3 class="heading-element"&gt;Local AI&lt;/h3&gt;

&lt;/div&gt;

&lt;ul&gt;
&lt;li&gt;Ollama&lt;/li&gt;
&lt;li&gt;Qwen 2.5 3B&lt;/li&gt;
&lt;/ul&gt;

&lt;div class="markdown-heading"&gt;
&lt;h2 class="heading-element"&gt;Project Structure&lt;/h2&gt;

&lt;/div&gt;

&lt;div class="snippet-clipboard-content notranslate position-relative overflow-auto"&gt;
&lt;pre class="notranslate"&gt;&lt;code&gt;launchlazy/
├── backend/
│   ├── main.py
│   └── requirements.txt
├── frontend/
│   ├── src/
│   ├── package.json
│   ├── package-lock.json
│   ├── .env.example
│   └──&lt;/code&gt;&lt;/pre&gt;…&lt;/div&gt;&lt;/div&gt;
  &lt;/div&gt;
  &lt;div class="gh-btn-container"&gt;&lt;a class="gh-btn" href="https://github.com/R-iddhi/launchlazy" rel="noopener noreferrer"&gt;View on GitHub&lt;/a&gt;&lt;/div&gt;
&lt;/div&gt;






&lt;h2&gt;
  
  
  How I Built It
&lt;/h2&gt;

&lt;p&gt;The most important design decision was actually deciding &lt;strong&gt;what the AI should NOT do&lt;/strong&gt;.&lt;/p&gt;

&lt;p&gt;I did not want to send a project to an LLM and ask:&lt;/p&gt;

&lt;blockquote&gt;
&lt;p&gt;“Tell me everything wrong with this.”&lt;/p&gt;
&lt;/blockquote&gt;

&lt;p&gt;That can easily become guesswork.&lt;/p&gt;

&lt;p&gt;So LaunchLazy separates facts from reasoning.&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Part&lt;/th&gt;
&lt;th&gt;Job&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Python scanner&lt;/td&gt;
&lt;td&gt;Find what actually exists&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Risk engine&lt;/td&gt;
&lt;td&gt;Score verified findings&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Qwen 2.5&lt;/td&gt;
&lt;td&gt;Explain and prioritize&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;React dashboard&lt;/td&gt;
&lt;td&gt;Show everything clearly&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;If &lt;code&gt;requirements.txt&lt;/code&gt; is missing, normal code can detect that.&lt;/p&gt;

&lt;p&gt;If CORS is set to &lt;code&gt;*&lt;/code&gt;, normal code can detect that.&lt;/p&gt;

&lt;p&gt;The AI only gets the findings after they are verified.&lt;/p&gt;

&lt;p&gt;That became my favourite rule from this project:&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Code finds the facts. AI explains why they matter.&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;I also optimized the scanner so the project filesystem is walked once instead of repeatedly searching the same folders for every check.&lt;/p&gt;

&lt;p&gt;And I separated scanning from AI generation, so the user sees the score and risks first while Qwen prepares the explanation.&lt;/p&gt;




&lt;h2&gt;
  
  
  Why Does Open Innovation Matter?
&lt;/h2&gt;

&lt;p&gt;LaunchLazy uses &lt;strong&gt;Qwen 2.5 3B through Ollama&lt;/strong&gt;, locally.&lt;/p&gt;

&lt;p&gt;That matters because project structure and configuration can contain information I would rather keep on the developer's machine.&lt;/p&gt;

&lt;p&gt;Using a local open-weight model means:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;project evidence can stay local&lt;/li&gt;
&lt;li&gt;there is no per-scan hosted API cost&lt;/li&gt;
&lt;li&gt;the reasoning model can be replaced later without rebuilding the scanner&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;More importantly, it pushed me to think about where AI was genuinely useful instead of using it for every step.&lt;/p&gt;




&lt;h2&gt;
  
  
  The Lazy Payoff
&lt;/h2&gt;

&lt;p&gt;Without LaunchLazy, the workflow is basically:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;README → package.json → .gitignore → env files
→ source code → backend config → deployment files
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;With LaunchLazy:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;Paste path → Inspect → Fix the important things first
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;I am not claiming it magically makes a project production-ready.&lt;/p&gt;

&lt;p&gt;It just removes a lot of repetitive checking and gives the developer a much clearer starting point.&lt;/p&gt;

&lt;p&gt;That is my version of &lt;strong&gt;Slow and Steady&lt;/strong&gt;.&lt;/p&gt;




&lt;p&gt;One slightly funny part of building this:&lt;/p&gt;

&lt;p&gt;I eventually made &lt;strong&gt;LaunchLazy audit LaunchLazy&lt;/strong&gt;.&lt;/p&gt;

&lt;p&gt;And yes, it complained about my own project.&lt;/p&gt;

&lt;p&gt;Which is probably a good sign. 😅&lt;/p&gt;




&lt;p&gt;The biggest thing I learned was this:&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;An AI project does not become better by giving more work to the AI.&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Sometimes the better design is:&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Let software establish what is true.&lt;br&gt;&lt;br&gt;
Let AI help decide what to do about it.&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;And if that lets me be a little lazier before deployment, even better. 🦥&lt;/p&gt;

</description>
      <category>devchallenge</category>
      <category>hf26challenge</category>
      <category>ai</category>
      <category>opensource</category>
    </item>
    <item>
      <title>SideLedger — AI-Assisted Financial Review for Freelancers</title>
      <dc:creator>Riddhi Rawool</dc:creator>
      <pubDate>Tue, 20 Oct 2026 13:17:11 +0000</pubDate>
      <link>https://dev.to/riddhi_rawool_3b9f2e1b2da/sideledger-ai-assisted-financial-review-for-freelancers-2n7a</link>
      <guid>https://dev.to/riddhi_rawool_3b9f2e1b2da/sideledger-ai-assisted-financial-review-for-freelancers-2n7a</guid>
      <description>&lt;p&gt;This is a submission for the &lt;a href="https://dev.to/challenges/hacktoberfest-week2-2026-10-12"&gt;Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle&lt;/a&gt;&lt;/p&gt;

&lt;h1&gt;
  
  
  SideLedger — AI-Assisted Financial Review for Freelancers
&lt;/h1&gt;

&lt;h2&gt;
  
  
  What I Built
&lt;/h2&gt;

&lt;p&gt;I built &lt;strong&gt;SideLedger&lt;/strong&gt;, an AI-assisted financial statement review tool designed for freelancers and independent workers.&lt;/p&gt;

&lt;p&gt;Freelancers often receive transaction statements containing dozens or hundreds of payments, subscriptions, purchases, and repeated charges. Reviewing these manually can be time-consuming, especially when trying to understand where money went or which transactions deserve attention.&lt;/p&gt;

&lt;p&gt;SideLedger lets a user upload a CSV transaction statement and turns it into a structured financial review.&lt;/p&gt;

&lt;p&gt;It can:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Categorize transactions&lt;/li&gt;
&lt;li&gt;Detect possible duplicate payments&lt;/li&gt;
&lt;li&gt;Identify recurring expenses&lt;/li&gt;
&lt;li&gt;Flag unusual spending&lt;/li&gt;
&lt;li&gt;Build a prioritized Review Inbox&lt;/li&gt;
&lt;li&gt;Summarize spending by category&lt;/li&gt;
&lt;li&gt;Answer questions about the uploaded statement through the SideLedger Assistant&lt;/li&gt;
&lt;li&gt;Export analyzed results&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;The idea is to reduce the time a freelancer spends manually inspecting transactions while keeping the final financial decisions with the user.&lt;/p&gt;

&lt;p&gt;Rather than positioning AI as an autonomous financial adviser, SideLedger uses AI as a &lt;strong&gt;review assistant&lt;/strong&gt;. Potential users could include freelancers, independent consultants, creators, and other self-employed professionals who regularly need to understand their business spending.&lt;/p&gt;

&lt;h2&gt;
  
  
  Demo
&lt;/h2&gt;

&lt;p&gt;  &lt;iframe src="https://www.youtube.com/embed/WgfHpDq_6S0" width="710" height="399"&gt;
  &lt;/iframe&gt;
&lt;/p&gt;

&lt;p&gt;The demo shows SideLedger analyzing a transaction statement, displaying the financial dashboard and Review Inbox, and using the AI assistant to answer questions about the analyzed transactions.&lt;/p&gt;

&lt;h2&gt;
  
  
  Code
&lt;/h2&gt;


&lt;div class="ltag-github-readme-tag"&gt;
  &lt;div class="readme-overview"&gt;
    &lt;h2&gt;
      &lt;img src="https://assets.dev.to/assets/github-logo-5a155e1f9a670af7944dd5e12375bc76ed542ea80224905ecaf878b9157cdefc.svg" alt="GitHub logo"&gt;
      &lt;a href="https://github.com/R-iddhi" rel="noopener noreferrer"&gt;
        R-iddhi
      &lt;/a&gt; / &lt;a href="https://github.com/R-iddhi/hacktoberfest-expense-sorter" rel="noopener noreferrer"&gt;
        hacktoberfest-expense-sorter
      &lt;/a&gt;
    &lt;/h2&gt;
    &lt;h3&gt;
      SideLedger - AI-assisted financial statement review for freelancers
    &lt;/h3&gt;
  &lt;/div&gt;
&lt;/div&gt;


&lt;p&gt;The project is open source and the repository contains both the React frontend and FastAPI backend.&lt;/p&gt;

&lt;h2&gt;
  
  
  How I Built It
&lt;/h2&gt;

&lt;p&gt;SideLedger uses a hybrid approach instead of asking a language model to perform every task.&lt;/p&gt;

&lt;h3&gt;
  
  
  Open-weight AI
&lt;/h3&gt;

&lt;p&gt;I used &lt;strong&gt;Qwen 2.5 3B through Ollama&lt;/strong&gt; for local language-model inference.&lt;/p&gt;

&lt;p&gt;Qwen is used for AI-assisted transaction categorization and for the SideLedger Assistant, which answers questions using context from the uploaded financial statement.&lt;/p&gt;

&lt;p&gt;The assistant is deliberately grounded in the analyzed statement instead of being used as a general-purpose chatbot.&lt;/p&gt;

&lt;h3&gt;
  
  
  Deterministic financial analysis
&lt;/h3&gt;

&lt;p&gt;Python and Pandas handle operations where deterministic computation is more appropriate, including transaction processing, totals, duplicate checks, recurring-pattern analysis, and preparation of structured financial data.&lt;/p&gt;

&lt;p&gt;This prevents simple financial calculations from depending entirely on generative AI.&lt;/p&gt;

&lt;h3&gt;
  
  
  Anomaly detection
&lt;/h3&gt;

&lt;p&gt;SideLedger combines statistical anomaly detection with an additional tabular-model signal using &lt;strong&gt;TabPFN&lt;/strong&gt;.&lt;/p&gt;

&lt;p&gt;An IQR-based statistical method provides a simple and explainable baseline for identifying unusually large transactions, while TabPFN provides an additional signal for tabular anomaly analysis.&lt;/p&gt;

&lt;h3&gt;
  
  
  Human-in-the-loop review
&lt;/h3&gt;

&lt;p&gt;Instead of automatically deciding that a transaction is wrong, SideLedger places suspicious or noteworthy transactions into a &lt;strong&gt;Review Inbox&lt;/strong&gt;.&lt;/p&gt;

&lt;p&gt;Users can filter the inbox by:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Unusual spending&lt;/li&gt;
&lt;li&gt;Possible duplicates&lt;/li&gt;
&lt;li&gt;Recurring expenses&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;This keeps the user in control of the final decision.&lt;/p&gt;

&lt;h3&gt;
  
  
  Technology stack
&lt;/h3&gt;

&lt;ul&gt;
&lt;li&gt;React + Vite&lt;/li&gt;
&lt;li&gt;FastAPI&lt;/li&gt;
&lt;li&gt;Python&lt;/li&gt;
&lt;li&gt;Pandas&lt;/li&gt;
&lt;li&gt;Qwen 2.5 3B&lt;/li&gt;
&lt;li&gt;Ollama&lt;/li&gt;
&lt;li&gt;TabPFN&lt;/li&gt;
&lt;li&gt;Lucide icons&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;The architecture separates deterministic financial processing from AI reasoning so that each component is used for the type of task it handles best.&lt;/p&gt;

&lt;h2&gt;
  
  
  Why Does Open Innovation Matter?
&lt;/h2&gt;

&lt;p&gt;Open innovation made SideLedger possible because financial data can be highly personal.&lt;/p&gt;

&lt;p&gt;Using an open-weight model such as Qwen through Ollama makes it possible to experiment with local inference rather than designing the entire product around sending transaction information to a proprietary hosted language-model API.&lt;/p&gt;

&lt;p&gt;It also gives developers greater control over how the model is used, how prompts are constructed, and how AI is combined with deterministic software.&lt;/p&gt;

&lt;p&gt;For SideLedger, open innovation is particularly valuable because the goal is not to let an AI make financial decisions. The goal is to combine transparent software, open models, statistical methods, and human review into a system where the user remains in control.&lt;/p&gt;

&lt;p&gt;Open models also make this architecture easier for other developers to inspect, modify, reproduce, and extend.&lt;/p&gt;

&lt;h2&gt;
  
  
  Prize Categories
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;TabPFN&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;SideLedger integrates TabPFN as an additional tabular anomaly-detection signal alongside a statistical IQR-based method.&lt;/p&gt;

</description>
      <category>devchallenge</category>
      <category>hf26challenge</category>
    </item>
    <item>
      <title>Enjoying HAcktoberfest</title>
      <dc:creator>Riddhi Rawool</dc:creator>
      <pubDate>Fri, 16 Oct 2026 11:42:25 +0000</pubDate>
      <link>https://dev.to/riddhi_rawool_3b9f2e1b2da/enjoying-hacktoberfest-3fp3</link>
      <guid>https://dev.to/riddhi_rawool_3b9f2e1b2da/enjoying-hacktoberfest-3fp3</guid>
      <description>&lt;p&gt;helllllo!!!!&lt;/p&gt;

</description>
    </item>
  </channel>
</rss>
