Local LLM benchmark reveals extraction flaws in knowledge graph building
A new reproducible leaderboard tests 22 local models on knowledge graph extraction and chat, showing that model size does not guarantee accuracy.
News for teams that run their own software
A daily digest on self-hosting, DevOps, security, document AI and HR tech. Each story is summarized, explains why it matters and links to the original publisher.
Fixing Keycloak HTTPS and proxy errors in production mode
A technical guide resolves common Keycloak deployment errors like 'HTTPS required' and invalid redirect URIs by configuring reverse proxies and hostname settings correctly.
Backdoored ARTEX copies grant root access to self-hosted consoles
CERT-In reports AI-driven attacks on BFSI sectors and warns of backdoored ARTEX Docker images granting unauthenticated root shells.
Maintenant offers self-hosted monitoring with built-in AI agent support
The open-source tool maintenant provides zero-config monitoring for Docker and Kubernetes, featuring a native Model Context Protocol server for AI integration.
Diagnosing ISP DNS hijacking when standard resolvers fail
Plaintext DNS queries on port 53 are vulnerable to interception. Learn how to detect ISP tampering using encrypted DNS and certificate validation.
Self-hosted AI gateway outage reveals credential handling flaws
A developer shares how a self-hosted model gateway triggered a vendor block due to shared OAuth tokens, leading to a redesign of task dispatch and error handling.
Why measuring one reader led to deleting files needed by tests
A developer quarantined 34,557 HTML files after measuring a single reader, only to find 310 test failures. The incident highlights the risks of incomplete data lifecycle management.
Prompt adware in OpenAI Codex jailbreak modifies agent instructions
Check Point Research found a Codex jailbreak bundle that injects ads into AI agent responses, highlighting new supply chain risks for self-hosted LLM setups.
DreamHost breach exposes customer emails and session tokens via file manager flaw
Hackers exploited a vulnerability in DreamHost's file manager starting in August, stealing session tokens and emails from an undisclosed number of customers before detection in November.
Self-hosted 27B model automates code reviews with transparent fallback
A team replaced expensive hosted LLM reviews with a self-hosted 27B model on existing GPUs, achieving 75% of PRs reviewed within eight minutes.
Critical Rails flaw allows file writes via ImageMagick uploads
Ruby on Rails versions before 8.0.5.2 and 8.1.4.1 contain a critical vulnerability in Active Storage that lets attackers write files via crafted ImageMagick uploads.
Grist introduces window functions to fix spreadsheet data integrity
Grist adds PREVIOUS and NEXT functions to handle running totals and ordered lookups without the fragility of traditional cell references.
When health checks report success but measure nothing
A developer found three monitoring checks that always passed despite being broken, highlighting the danger of silent validation failures in production systems.
Specialized AI model extracts document fields cheaper than frontier APIs
Hugging Face released Dharma Field Extractor, a 2-billion-parameter model that outputs structured JSON from documents. It outperforms general LLMs in benchmarks while reducing inference costs signific
BTCPay Server users urged to patch after hot wallet thefts
Attackers are exploiting unpatched BTCPay Server instances to steal Bitcoin and Litecoin from hot wallets, prompting urgent upgrade recommendations.
