GhostStream — GPU transcoding server (HLS/ABR) now integrated with GhostHub
https://github.com/BleedingXiko/GhostStream

https://redd.it/1pnm61f
@r_opensource

Check out Quantica 0.2.0 With AI/ML Capabilities
https://github.com/Quantica-Foundation/quantica-lang

https://redd.it/1pnp9h8
@r_opensource

Ekphos: A lightweight, fast, terminal-based markdown research tool inspired by Obsidian
https://github.com/hanebox/ekphos

https://redd.it/1pnr97n
@r_opensource

Open-sourced a React PDF annotation library (highlights, notes, drawing, signatures and more)

Hi everyone 👋

I’ve been working on a PDF annotation tool for React and just open-sourced the **first public version**.

Landing page: [https://react-pdf-highlighter-plus-demo.vercel.app/](https://react-pdf-highlighter-plus-demo.vercel.app/)

Npm: [https://www.npmjs.com/package/react-pdf-highlighter-plus](https://www.npmjs.com/package/react-pdf-highlighter-plus)

Github: [https://quocvietha08.github.io/react-pdf-highlighter-plus](https://quocvietha08.github.io/react-pdf-highlighter-plus/docs/)

What it supports right now:

* Text highlighting with notes
* Freehand drawing on PDFs
* Add signatures
* Insert images
* Designed to be embeddable in React apps
* Export PDF
* Free Hand Draw
* Insert a shape like a rectangle, circle, or arrow

It’s still early, but my goal is to make this a solid, flexible base for apps that need PDF interaction (learning tools, research, document review, etc.).

I’d really appreciate:

* Feedback from people who’ve built similar tools
* Feature requests
* Contributions or bug reports

If this looks useful to you, feel free to try it out or contribute.
Thanks for taking a look!

https://redd.it/1pnrkvs
@r_opensource

Looking for open source contributors for Voice AI on Github.
https://github.com/rapidaai/voice-ai

https://redd.it/1pnv4v0
@r_opensource

I built JSONTry, a JSON viewer using Flutter.
https://github.com/rivafarabi/jsontry

https://redd.it/1pnz3qc
@r_opensource

Dealing with open source burnout

I need some advice, as I’m feeling pretty burned out from maintaining my projects.

I created these projects because I personally needed them, and I made them public so others could use them too. One of them gained a lot of traction, which initially made me happy - people were finding something I built genuinely useful. However, that growth was followed by a torrent of issues and feature requests (with no PRs). Many of the ideas were good, but they were impossible to implement because I hadn’t considered scalability when I originally built the project. Again, it was made for myself and my specific use case.

Because of that, I decided to rewrite it to make those features possible. I prefer CLI apps, but a UI was by far the most requested feature, so I started building one as well. The rewrite is about 60% done, but I can’t bring myself to finish it. I stopped needing the project a while ago, and now it feels like I’m sacrificing my limited free time for nothing other than a never-ending list of issues and feature requests. I’m also on the fence about accepting donations, because at that point I think it would stop feeling like a hobby and start feeling like a product.

I’ve recently started working on something new - a CLI app that I actually need. It’s relatively simple for my use case, but I think a lot of people would be interested in it, and it could end up being my biggest project in terms of traction. The potential for features is basically endless, and because of that, I’m dreading making it public. It would be nice to help people, but I’m afraid it would turn into a third full-time job.

At the same time, it feels wrong to abandon the rewrite, given how much time I’ve already spent on it and the fact that many people are waiting for it. I’m honestly tempted to just archive everything and focus on other hobbies, but that would feel wrong too.

Has anyone dealt with something similar?

https://redd.it/1po0502
@r_opensource

Geolocate - A simple IP geolocation CLI tool based on latency
https://github.com/jimaek/geolocation-tool

https://redd.it/1po0t4q
@r_opensource

DebtDrone: An advanced technical debt analysis tool using AST
https://github.com/endrilickollari/debtdrone-cli

https://redd.it/1po175x
@r_opensource

What would make you trust a security browser extension?

Extensions are powerful. That's why people distrust them.


We're building Banbo with:

Minimal permissions
Client-side crypto
Zero email hosting
Transparent threat model

What would you personally need to see to trust an extension like this?

Project page: banbo

https://redd.it/1po4auq
@r_opensource

Making the Cyber Resilience Act Work for Open Source
https://thenewstack.io/making-the-cyber-resilience-act-work-for-open-source/

https://redd.it/1po5dwp
@r_opensource

Is there an open source alternative to DAPs like Whatfix?

Digital adoption tools like Whatfix and Pendo are too expensive for what they offer if you think about it. Are there any proper open source replacements for them?

If not would people use it I built one?

https://redd.it/1po4soo
@r_opensource

WhatsApp Wrapped - Every WhatsApp analytics tool wants to upload your chats to their servers. I built one that doesn't

I've always wanted something like Spotify Wrapped but for WhatsApp. There are some tools out there that do this, but every one I found either runs your chat history on their servers or is closed source. I wasn't comfortable with all that, so this year I built my own.

WhatsApp Wrapped generates visual reports for your group chats. You export your chat from WhatsApp (without media), run it through the tool, and get an HTML report with analytics about your conversations. Everything runs locally or in your own Colab session. Nothing gets sent anywhere.

Here is a Sample Report.

What it does:

- Message counts and activity patterns (who texts the most, what time of day, etc.)
- Emoji usage stats and word clouds
- Calendar heatmaps showing activity over time (like github activity)
- Interactive charts you can hover over and explore

How to use it:

The easiest way is through Google Colab, no installation needed. Just upload your chat export and download the report. There's also a CLI if you want to run it locally.

Tech stack: Python, Polars for data processing, Plotly for charts, Jinja2 for templating.

Links:

- GitHub Repository
- Sample Report
- Google Colab

Happy to answer any questions or hear feedback.

https://redd.it/1po8nh2
@r_opensource

I made an open-source macOS app that simulates realistic human typing to expose the limits of AI detection based on document history.
https://github.com/0xff-r4bbit/watchmetype

https://redd.it/1po8bc9
@r_opensource

What is everyone currently working on?



https://redd.it/1poa97f
@r_opensource

BurnBin - Free/Donoware, Open Source, Secure, No file size/bandwidth/speed limits, locally hosted.
https://www.youtube.com/watch?v=mW_qokqThg0

https://redd.it/1poc5mt
@r_opensource

I built an open-source site that lets students play games at school
https://michuscrypt.github.io/classroom20x-unblocked-games/

https://redd.it/1po9xau
@r_opensource

TSZ: Open-Source AI Guardrails & PII Security Gateway

Hi everyone! We’re the team at **Thyris**, focused on open-source AI with the mission **“Making AI Accessible to Everyone, Everywhere.”** Today, we’re excited to share our **first open-source product**, **TSZ (Thyris Safe Zone)**.

We built TSZ to help teams adopt LLMs and Generative AI safely, without compromising on data security, compliance, or control. This project reflects how we think AI should be built: open, secure, and practical for real-world production systems.

**GitHub:**
[https://github.com/thyrisAI/safe-zone](https://github.com/thyrisAI/safe-zone)

**Docs:**
[https://github.com/thyrisAI/safe-zone/tree/main/docs](https://github.com/thyrisAI/safe-zone/tree/main/docs)

# Overview

Modern AI systems introduce new security and compliance risks that traditional tools such as WAFs, static DLP solutions or simple regex filters cannot handle effectively. AI-generated content is contextual, unstructured and often unpredictable.

TSZ (Thyris Safe Zone) is an open-source AI-powered guardrails and data security gateway designed to protect sensitive information while enabling organizations to safely adopt Generative AI, LLMs and third-party APIs.

TSZ acts as a zero-trust policy enforcement layer between your applications and external systems. Every request and response crossing this boundary can be inspected, validated, redacted or blocked according to your security, compliance and AI-safety policies.

TSZ addresses this gap by combining deterministic rule-based controls, AI-powered semantic analysis, and structured format and schema validation. This hybrid approach allows TSZ to provide strong guardrails for AI pipelines while minimizing false positives and maintaining performance.

# Why TSZ Exists

As organizations adopt LLMs and AI-driven workflows, they face new classes of risk:

* Leakage of PII and secrets through prompts, logs or model outputs
* Prompt injection and jailbreak attacks
* Toxic, unsafe or non-compliant AI responses
* Invalid or malformed structured outputs that break downstream systems

Traditional security controls either lack context awareness, generate excessive false positives or cannot interpret AI-generated content. TSZ is designed specifically to secure AI-to-AI and human-to-AI interactions.

# Core Capabilities

# PII and Secrets Detection

TSZ detects and classifies sensitive entities including:

* Email addresses, phone numbers and personal identifiers
* Credit card numbers and banking details
* API keys, access tokens and secrets
* Organization-specific or domain-specific identifiers

Each detection includes a confidence score and an explanation of how the detection was performed (regex-based or AI-assisted).

# Redaction and Masking

Before data leaves your environment, TSZ can redact sensitive values while preserving semantic context for downstream systems such as LLMs.

**Example redaction output:**

john.doe@company.com -> [EMAIL]
4111 1111 1111 1111 -> [CREDIT_CARD]


This ensures that raw sensitive data never reaches external providers.

# AI-Powered Guardrails

TSZ supports semantic guardrails that go beyond keyword matching, including:

* Toxic or abusive language detection
* Medical or financial advice restrictions
* Brand safety and tone enforcement
* Domain-specific policy checks

Guardrails are implemented as validators of the following types:

* BUILTIN
* REGEX
* SCHEMA
* AI\_PROMPT

# Structured Output Enforcement

For AI systems that rely on structured outputs, TSZ validates that responses conform to predefined schemas such as JSON or typed objects.

This prevents application crashes caused by invalid JSON and silent failures due to missing or incorrectly typed fields.

# Templates and Reusable Policies

TSZ supports reusable guardrail templates that bundle patterns and validators into portable policy packs.

Examples include:

* PII Starter Pack
* Compliance Pack (PCI, GDPR)
* AI Safety Pack (toxicity, unsafe content)

Templates can be imported via API to quickly bootstrap new environments.

#

Architecture and Deployment

TSZ is typically deployed as a microservice within a private network or VPC.

**High-level request flow:**

1. Your application sends input or output data to the TSZ detect API
2. TSZ applies detection, guardrails and optional schema validation
3. TSZ returns redacted text, detection metadata, guardrail results and a blocked flag with an optional message

Your application decides how to proceed based on the response.

# API Overview

The TSZ REST API centers around the `detect` endpoint.

**Typical response fields include:**

* redacted\_text
* detections
* guardrail\_results
* blocked
* message

The API is designed to be easily integrated into middleware layers, AI pipelines or existing services.

# Quick Start

Clone the repository and run TSZ using Docker Compose.

git clone https://github.com/thyrisAI/safe-zone.git
cd safe-zone
docker compose up -d


Send a request to the detection API.

POST http://localhost:8080/detect
Content-Type: application/json

{"text": "Sensitive content goes here"}


# Use Cases

Common use cases include:

* Secure prompt and response filtering for LLM chatbots
* Centralized guardrails for multiple AI applications
* PII and secret redaction for logs and support tickets
* Compliance enforcement for AI-generated content
* Safe API proxying for third-party model providers

# Who Is TSZ For

TSZ is designed for teams and organizations that:

* Handle regulated or sensitive data
* Deploy AI systems in production environments
* Require consistent guardrails across teams and services
* Care about data minimization and data residency

# Contributing and Feedback

TSZ is an open-source project and contributions are welcome.

You can contribute by reporting bugs, proposing new guardrail templates, improving documentation or adding new validators and integrations.

# License

TSZ is licensed under the Apache License, Version 2.0.

https://redd.it/1pofbz1
@r_opensource

Open Source: Inside 2025’s 4 Biggest Trends
https://thenewstack.io/open-source-inside-2025s-4-biggest-trends/

https://redd.it/1poa89w
@r_opensource

domco@5.0.0 - use your favorite server framework with Vite
https://github.com/rossrobino/domco

https://redd.it/1poifko
@r_opensource