SShortSingh.
TopTechnologyAI & MLStartupsProgrammingScienceBusinessFinanceCrypto & Web3World
Back to feed
Programming

AURA v0.1.0 Launches with Deterministic AI Safety Scoring and Self-Healing Analytics

0
·2 views

Developers have released AURA v0.1.0, an open-source AI safety engine designed to replace brittle LLM guardrails with deterministic, fully auditable trigger extraction and risk scoring. The system uses a pipeline-driven architecture that processes source cases through normalization, regex-based rule extraction, and a non-linear exponential confidence formula to produce bounded, reproducible scores. Automated GitHub Actions snapshots preserve daily analytics telemetry, addressing a common problem of lost repository traffic data. A contributor identified an async race condition in batch processing that caused disk artifacts to reflect pre-audit state despite successful CLI logs, and a Promise.all-based fix has been documented. Upcoming versions plan to add TF-IDF prompt tokenization, a visual rule editor, and small supervised models to replace some heuristic components.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Log in to join the discussion and vote.

Log in

Related stories

0
ProgrammingDEV Community ·

Developer Rejects Security Advice on OAuth Endpoint, Finds Real Flaw Elsewhere

A developer running a QR code service received an unsolicited security report last week flagging an open OAuth dynamic client registration endpoint as a vulnerability. The developer argued the endpoint was functioning as intended, since the Model Context Protocol requires open dynamic registration to allow AI clients like Claude to connect without prior approval. The suggested fixes — requiring access tokens or a redirect URI allowlist — would have effectively disabled MCP support, and the redirect concern was already mitigated by strict per-client checks enforced at both authorization and token exchange. However, investigating the report led the developer to spot a genuine risk: the consent screen displayed the attacker-controlled 'client_name' field as trusted text, enabling anyone to register a client named something like 'QRFLOW Official Support' and craft a convincing phishing page on the real domain. The actual fix addressed how unverified client names were rendered on the consent screen, leaving the registration endpoint itself untouched.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Google Gemini Agentic Video Cuts Token Use 88% but Has 4 Silent Failure Conditions

A developer integrating Google's Gemini Agentic Video feature into a LINE Bot found that the mode reduces token consumption by 88% and costs by 66% compared to standard static video processing. Unlike the default static mode, which loads every frame at a fixed sampling rate, agentic mode lets the model selectively scan only relevant video segments to answer specific queries. Testing on a two-hour Google I/O keynote showed the feature could accurately pinpoint timestamped content, such as pricing discussions and product segments. However, the developer discovered four conditions must be met simultaneously for agentic mode to activate: using API version v1beta1, setting the media_processing parameter to AGENTIC, using a supported model, and configuring the thinking_level setting. Critically, if any condition is unmet, the API still returns a successful 200 response with no error, silently falling back to the more expensive static mode.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Steam Deck Guide: How to Install Windows .exe Patches on SteamOS Without Errors

Steam Deck users running SteamOS often struggle to install Windows-based .exe patches, localization mods, or fan updates because the installers cannot locate game directories hidden within Linux's dot-folder structure. The root causes are hidden folder restrictions under SteamOS and sandbox permission isolation from tools like Bottles or Flatpak applications. A workaround called the 'Relay Transfer Method' involves temporarily copying the game folder to a visible public directory such as Downloads or Desktop, running the patch installer from there, and then copying the patched files back to the original Steam game directory. The method avoids complex terminal commands and works with common compatibility tools like Bottles or Protontricks to launch the .exe installer. Users should also check whether the patch generates a new executable and rename files accordingly to ensure the game launches correctly after patching.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Google Gemini 3.8 Flash TTS Goes GA: Developer Builds Japanese Song-Learning Web App

Google officially launched Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS on September 22, marking their general availability. A developer used the new text-to-speech API to build a web app that fetches song lyrics, translates them into Chinese, adds phonetic transcriptions for Japanese or Korean, and teaches pronunciation sentence by sentence. The GA release introduced three notable capabilities: text-description-based voice design, voice replication from short recordings with consent watermarking, and per-sentence tone and style control. The two models differ in focus — the flagship targets expressive vocal performance while the Lite variant is optimized for high-volume, cost-efficient generation. The developer noted that while TTS output quality was strong, undocumented implementation details around the new interactions and voices API structure consumed significant development time and daily API quota.

0 comments Read more at DEV Community