SShortSingh.
Back to feed

How to Run Debian 13 and a Local LLM on Apple Silicon Using Apple's Container CLI

0
·70 views

A developer has published a technical guide detailing how to set up a Debian 13 virtual machine using Apple's container CLI on an Apple Silicon Mac, while offloading AI inference to the Mac's GPU via Ollama. Because Linux VMs lack direct Metal GPU access, the guide proposes running the application inside the Debian VM while routing model calls to Ollama running on the host macOS system at 192.168.64.1. A key pitfall documented is that the official Debian 13 image lacks /sbin/init, causing the machine to silently fail to boot — a fix involving a custom Dockerfile with systemd-sysv is provided. The setup was tested on an Apple M3 Mac running macOS 27.0 with container runtime 1.4.1 and Ollama 0.33.3, achieving 45.2 tokens per second with the Gemma 4 model loaded entirely on the GPU. Supporting files and configuration tips have been published on GitHub to help others replicate the workflow.

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 ·

What Clients Check Before Hiring a Freelancer: Proof, Pricing and Response Time

This piece looks at personal branding from the buyer's perspective, focusing on what potential clients scan in the minutes before deciding whether to contact or hire a freelancer. It argues that buyers mainly want evidence that someone has solved similar problems, so documented case studies with outcomes outweigh follower counts or frequent posting. The author suggests answering a buyer's core questions directly: positioning, past results, a starting price range, and response time. It also recommends owning a custom domain, a fast mobile-friendly page, and a visible contact email, since these signal long-term reliability.

0
ProgrammingDEV Community ·

TeeRouter Proposes Signed Batches for Auditable AI API Billing Without a Blockchain

A developer building the TeeRouter AI API router describes how to make billing records verifiable without running a blockchain. The design freezes pricing rules at request admission so historical charges can be reproduced later. It also sets a policy for interrupted requests, such as timeouts after the request was fully sent, where estimated input charges may apply. The author states that the approach publishes cryptographically verifiable records but makes no claim of decentralized consensus. The article is a case study of one public codebase, not an independent certification of the deployed service.

0
ProgrammingDEV Community ·

Fixing SMTP 554 5.7.1 blocklist rejections: causes and remedies

Mail receivers can permanently reject messages with SMTP 554 5.7.1 when the sending IP appears on a DNS blocklist or the content is flagged as spam. Common triggers include a compromised mailbox or web form relaying spam, a dynamic IP range listed under Spamhaus PBL, or another tenant on a shared IP causing the listing. The guide recommends first stopping the spam source by locking accounts, changing passwords, and clearing the queue before requesting delisting, since otherwise the listing tends to recur. It also explains that a 127.255.255.254 response from Spamhaus indicates a query made through a public resolver rather than a true listing, and notes that SBL/XBL delistings typically take minutes to hours once spam stops.

0
ProgrammingDEV Community ·

TeeRouter uses trusted execution environments to make an AI API router verifiable

An AI API router must read requests to route them, enforce permissions, and calculate usage, which gives its operator room to alter behavior or misreport prices. The open-source project TeeRouter explores whether trusted execution environments and verifiable receipts can answer whether the expected program is actually handling requests. Its design binds a hardware attestation report to the live TLS certificate and a client challenge, so the evidence is tied to the connection carrying the prompts. Verification also requires a verifier-chosen policy for accepted measurements, since an authentic report can still describe untrustworthy code. The analysis, based on commit 5c89b551, notes that enclave isolation does not fix flawed authorization logic and that user authentication additionally requires a client-held Ed25519 signature.