Cognocient offers proxy-free LLM cost tracking via async Python wrapper

·4 views

A developer has released Cognocient, an open-source Python wrapper for OpenAI and Anthropic SDKs that tracks LLM usage costs without routing requests through a proxy. The tool wraps existing client calls transparently and fires cost reports on a background thread only after the actual API response has already been returned. A key design priority was fault isolation: if Cognocient's reporting endpoint is unavailable, the failure remains invisible to the host application and does not affect real API calls. The wrapper supports per-call cost attribution with optional tags for teams, features, and users to enable chargeback-style reporting. Currently at v0.1.x, it does not yet support streaming responses and intentionally omits pre-call blocking, positioning it for teams that want cost visibility without adding a critical dependency to their request path.

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)

Developer finds 8 WCAG errors in own accessibility plugin after reviewer flags 3

A WordPress accessibility scanner developer discovered eight incorrectly mapped WCAG criteria in their plugin after a reviewer flagged three issues last week. Problems included citing WCAG 4.1.1, a criterion removed in WCAG 2.2, and mislabeling best practices as formal conformance violations. Of the plugin's 25 automated checks, only 18 actually correspond to WCAG 2.2 success criteria, while the remaining seven are best practices. The developer corrected all eight rules, updated marketing copy to reflect the accurate 18-and-7 split, and tightened the accessibility statement generator to prevent false conformance claims. The incident highlighted how overstated audit reports can undermine legal accessibility declarations rather than support them.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Metabase patches CVSS 10.0 SQL injection zero-day exploited before disclosure

Metabase has disclosed a critical unauthenticated SQL injection vulnerability in its business intelligence platform, rated CVSS 10.0, that was actively exploited in the wild before the patch was released. The flaw allows remote attackers without any credentials to inject SQL into the Metabase application database and grant themselves administrator access. Because Metabase holds stored credentials to connected data warehouses and production databases, a successful compromise effectively exposes all linked data sources. Six version branches from 1.58 onwards are affected, and self-hosted users must manually apply the exact fixed build for their branch, as Metabase Cloud instances have already been updated by the vendor. As a temporary workaround, administrators who cannot patch immediately are advised to block the /api/session/reset_password endpoint at their reverse proxy or WAF.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

IT Student from Indonesia Joins DEV Community to Share Web Dev Journey

Delrico, an Information Technology student at the University of North Sumatra (USU) in Indonesia, has introduced himself to the DEV Community platform. He has a keen interest in web development and enjoys learning through hands-on experimentation and building projects. He joined the community to learn from other developers and share his own discoveries along the way. He plans to write about lessons learned, projects built, and problems solved that may benefit other developers. He views sharing his journey publicly as an integral part of his own learning process.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Developer Builds Sentinel, a Manifest V3 Browser Extension for Phishing Detection

Yerevan-based full-stack developer Artyom Mamyan has released Sentinel, an open-source browser extension built on Chrome's Manifest V3 platform. The tool combines multiple detection layers, including URL heuristics, typosquat-style checks, optional Google Safe Browsing lookups, and brand or page content signals. When a threat is identified, users are alerted through a popup dashboard, an in-page banner, or a full block interstitial. Sentinel is written in vanilla JavaScript with configurable sensitivity settings and includes architecture documentation in its GitHub repository. Mamyan describes it as a portfolio-grade security UX project and is open to remote work opportunities.

0 comments Read more at DEV Community