Haunted History: Git Logs Brought to Life

#devchallenge #hf26challenge

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 4: Ghost in the Machine

What I Built

I built Haunted history, an offline-first, privacy-focused tool that uses local open-weight AI models to [insert core function, e.g., automatically organize and summarize unstructured receipts and invoices]. It runs entirely on your local machine without sending data to cloud servers.

Demo

Animated GIF / Video Clip: A quick 10–15 second screen recording showing the main flow (e.g., uploading a document/image \rightarrow local AI processing \rightarrow structured output).
Annotated Screenshots: Static images highlighting key UI elements or terminal outputs.

Code

The Code / Implementation

The core logic of [Project Name] focuses on seamless local inference without hitting external APIs. Here is a look at how the application interacts with the open-weight model:

Connecting to the Local AI Model


python
import ollama

def process_local_inference(prompt_text, image_path=None):
    """
    Sends data to the local open-weight model via Ollama 
    for offline processing and analysis.
    """
    response = ollama.chat(
        model='llama3.2-vision',  # or your chosen model
        messages=[
            {
                'role': 'user',
                'content': prompt_text,
                'images': [image_path] if image_path else []
            }
        ]
    )
    return response['message']['content']

## How I Built It
<!-- Which open-source AI did you use (open-weight models, agent harnesses, frameworks, local inference), and how is your project built around it? -->
1. The Tech Stack
Briefly list the core technologies, frameworks, and libraries you used:
Frontend / UI: Streamlit / React / Terminal CLI
AI Engine / Runner: Ollama / Hugging Face / Llama.cpp
Language: Python / JavaScript / TypeScript
2. Architecture & Design Decisions
Explain why you built it this way (e.g., focusing on privacy, zero-cost, offline-first design):
Why you chose open-weight models instead of cloud APIs.
How data flows from the user interface to the local inference engine.
3. Challenges & Solutions (The Fun Part)
Judges love reading about what went wrong and how you fixed it! Mention 1 or 2 minor technical challenges (e.g., handling local image compression, parsing raw model outputs into structured JSON, or managing prompt engineering).

## Why Does Open Innovation Matter?
<!-- Why does open innovation matter for what you built?  What did it make possible that a closed API wouldn't? -->
Open innovation matters because the pace of technological change and problem-solving today is simply too fast for any single organization—no matter how large, wealthy, or talented—to innovate entirely in isolation.
Originally coined by Henry Chesbrough, open innovation is the practice of combining internal and external ideas, technologies, and pathways to market

## My Agent Session
<!-- Optional, but judges love it.  Save your session with DevRelay and embed it with the agent_session tag (see the challenge page), or link to it. -->
Agent / Tool Used: State which AI coding assistant or agent framework you used to help write, debug, or architect your project (e.g., Claude Code, Backboard R-CLI, GitHub Copilot, Cursor, Roo Code, or custom scripts).
Session Transcript / Export: Provide a link (e.g., to an Entire session log, GitHub Gist, or repo file) or embed a collapsible block showing the prompts and responses that generated your core code.
Key Prompt / Aha! Moment: Highlight 1–2 specific prompts where the agent solved a tricky bug, refactored a function, or created a key feature.

## The Spirit of It
<!-- Optional, but this is where you can make your post stand out.  Did it give someone chills, bring back a memory, or keep a story alive?  Show us the moment it landed.  If not, remove this section. -->

## Prize Categories
<!-- Which partner categories are you entering?  List every one that applies, or remove this section. -->

<!-- Team Submissions: Please pick one member to publish the submission and credit teammates by listing their DEV usernames directly in the body of the post. -->

<!-- Thanks for participating! -->

DEV Community