Developer runs Claude Code on Grok 4.5 via API gateway with three environment variables

·2 views

A developer who regularly uses Claude Code with Opus 4.8 decided to test xAI's Grok 4.5, released on July 8, within the same terminal environment. Two factors motivated the experiment: Grok 4.5 is xAI's first model explicitly built for coding and agentic tasks, and third-party benchmarks showed it completing an agentic indexing task using roughly 14,000 output tokens compared to around 67,000 for a rival flagship model. The setup relied on Claude Code's ability to route requests to any endpoint that speaks Anthropic's native messaging protocol, using an API gateway called byesu that exposes Anthropic-compatible access to multiple models including Grok 4.5. Just three environment variables — setting the base URL, API key, and model name — were enough to redirect Claude Code's requests to Grok 4.5 without any additional client software. The developer noted a few minor gotchas, including stripping trailing path segments from the base URL and trying an alternate variable name for the API key if authentication fails.

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)

How to Audit Trial Reactivation Emails Before Blaming the Copy

Many SaaS teams rush to rewrite subject lines or add discounts when trial reactivation emails underperform, but the root cause is often poor segmentation, delayed event tracking, or test accounts mixed with real users. A structured review should confirm three things: who received the email, why they were targeted, and what they did in the product afterward. Key technical checks include validating the segment query, ensuring job idempotency, verifying the timestamp that marks a trial as inactive, and confirming the final event that signals a user returned. A common bug involves converted users receiving win-back emails because billing status syncs with a slight delay, making campaigns appear intrusive when the real issue is a data pipeline problem. Defining what 'reactivation' means before testing — whether a login or a key product action — is also critical, as a shifting definition mid-test undermines confidence in any results.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Three Ways to Forward WhatsApp Messages to Telegram With Media Intact

There is no built-in method to forward messages between WhatsApp and Telegram, leaving users to choose from three available workarounds. WhatsApp's manual chat export can be imported into Telegram, but it creates only a one-time snapshot and frequently drops or degrades media files. Open-source self-hosted bots like watgbridge offer real-time syncing but require a continuously running server and ongoing technical maintenance. A third option, managed bridge services such as wa2tg, handles the infrastructure remotely and relays messages live with media preserved in both directions. The choice largely depends on the user's technical comfort and need for reliability, with managed bridges being the most practical solution for businesses and communities that depend on media-rich communication.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Shopify Shipping Policy Split Triggered Mass Google Merchant Center Disapprovals

A small Shopify store owner accidentally caused 621 of 739 active products to be disapproved in Google Merchant Center after splitting a single shipping policy into separate domestic and international rules. Google returned a 'missing_shipping' error, which was misleading — the real cause was a sync-order race condition where Google's feed re-synced before both new policies had fully propagated rates for all products. The fix was straightforward: consolidating both policies back into one universal rule, which restored approvals from 582 to 739 within a day. The store owner noted that Merchant Center's UI reports symptoms at the product level rather than identifying account-level root causes, making diagnosis unnecessarily time-consuming. To address this, they built a diagnostic tool using the Merchant Center API that groups disapprovals by root cause, filters out non-impactful warnings, and checks account-level settings as a primary step.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Why Microservices Need Journey-Based SLOs, Not Service-Level Uptime Targets

A technical guide published on DEV Community by Dr. Samson Tanimawo explains why traditional Service Level Objectives often fail in microservices architectures. When a service depends on multiple downstream services, compound availability math means reliability degrades rapidly — five services each at 99.9% yield a realistic ceiling of only 99.5%. The author argues that SLOs should be defined around user journeys such as checkout or login, rather than individual service health checks, to better reflect real user experience. Teams are advised to track error budget burn rates using Prometheus metrics and enforce deployment freezes when budgets are consumed, creating shared accountability between engineering and product teams. The piece recommends starting with a single critical user journey, measuring it for 30 days, and gradually expanding SLO coverage from there.

0 comments Read more at DEV Community