Local LLM Server Pro vs Low

Turn Apple Intelligence into a local API for your scripts and apps. Local LLM Server Pro exposes Apple's on-device language model through selected OpenAI and Ollama-compatible endpoints. Inference stays on your device. PRO: A ONE-TIME PURCHASE No ads and no daily generation request quota. Pro does not expand Apple's model context window or guarantee background service. The free version is ad-supported and allows 100 generation requests per day. START WITH ONE RESPONSE Enable Apple Intelligence and wait for the model to be ready. Start the server, open the in-app playground, and ask a short question. Then use the curl, Python or JavaScript examples in Server Information to connect your client. A step-by-step setup guide is available from the support link. SUPPORTED WORKFLOWS • OpenAI Chat Completions with a custom /v1 base URL, streaming, structured output and tool-call responses. • Selected Ollama chat, generate and model-list endpoints. • Text embeddings using Apple's contextual embedding provider. • Same-device localhost access, or local-network access with an optional API key. • Server logs, request statistics and Shortcuts for server controls. KNOW THE LIMITS This app uses Apple's supported on-device model. It does not import GGUF or Hugging Face models, download arbitrary Ollama models, or provide OpenAI's Responses API. Use model IDs returned by the server, such as base. Compatibility depends on the endpoints your client uses. Requests can contain up to 1 MiB of JSON. This HTTP byte limit is separate from the model's token context. Shorten large conversations or tool definitions; use max_tokens to bound replies. KEEP THE SERVER VISIBLE Keep the app in the foreground while serving. On iPad, use Split View with your client alongside it. Requests may pause when the app is backgrounded or the screen locks. The optional background reminder is advisory, not suspension detection. REQUIREMENTS Apple Intelligence-capable hardware with iOS/iPadOS 26 or later, or macOS 26 or later on a compatible Apple silicon Mac. Apple Intelligence must be enabled, supported in your language and region, and its on-device model ready. Setup may require a model download. Setup and troubleshooting: https://www.kevintang.xyz/apps/Local-LLM-Server-Pro/support.html

LEO puts your cloud accounts, your Docker hosts and a set of network diagnostics on one iPhone or iPad. It talks directly to the AWS, Azure and Google Cloud APIs and to the Docker Engine API, using keys you create, with no LEO server in between. MISSION CONTROL • One Environment Check loads your cloud resources and containers and runs a reachability pulse against every Docker host (on its API port) and every cloud instance with a public address • Counts of running, stopped and busy resources, unreachable hosts, and your most recent diagnostics • Quick jumps into Ping, SSL Check, Port Scan, Traceroute and the instance list ORBIT: CLOUD • AWS: EC2, Auto Scaling groups, ECS services, EKS clusters, Lambda, App Runner and RDS, in the region you choose • Azure: virtual machines, scale sets, Container Instances, AKS, App Service and Functions, Container Apps and SQL Managed Instance, across one subscription • Google Cloud: Compute Engine, GKE, Cloud Run, Cloud Functions and Cloud SQL, in one project • State, type, region, IP addresses, tags and resource details; filter by provider, group by status, region or type, search by name • Metrics from CloudWatch, Azure Monitor and Cloud Monitoring (CPU, network, memory for ECS, invocations and errors for Lambda) • Rightsizing hints from AWS Compute Optimizer, Azure Advisor and Google Recommender • Recent logs: EC2 console output and CloudWatch Logs for Lambda and ECS • Start, stop and reboot or restart where the provider allows it (Compute Engine "Reset" is a hard reset; App Runner pauses and resumes) • AWS access keys, with optional session tokens for temporary credentials; Azure service principal; Google service account JSON key ORBIT: DOCKER • Connect to any host that exposes the Docker Engine API over HTTP or HTTPS (2375 / 2376); the API version is negotiated • Containers: start, stop, restart, pause, remove, inspect, CPU and memory stats, live logs, the process list • Run one-off commands in a container (each runs on its own with sh -c) • Images: list, sort by name, size or date, remove GROUND: 10 NETWORK TOOLS • Ping and MTR: latency, loss and jitter from timed TCP connections to ports 80, 443, 53 and 22 • Traceroute: repeated TCP probes that show whether and how fast a host answers • Port Scan: TCP connect scan of 22 common ports, the top 100, ports 1–1000 or a custom range, with service names • SSL Check: certificate chain, subject, issuer, SANs, serial and days to expiry • Dig: A, AAAA, MX, TXT, CNAME, NS, SOA and PTR queries over UDP to the resolver you pick • DNS Propagation: one record checked on 8 public resolvers • WHOIS via RDAP, Reverse Lookup (PTR) and a CIDR subnet calculator • History of every run, and a hosts address book with favorites, tags, CSV import and export, and import from a JSON or Ansible inventory URL WHAT LEO DOESN'T DO • No ICMP: iOS doesn't let apps send it, so Ping, MTR and Traceroute use TCP and can't show the routers along a path • No interactive shell or SSH client • No background monitoring, alerts or notifications; data refreshes while the app is open • One AWS region, one Azure subscription and one Google Cloud project at a time • No Docker client certificates (TLS checks the server's certificate only), no image pulls • No billing or cost reports PRIVACY Your keys are stored in the iOS Keychain on this device. Requests go only from your device to the services you set up (AWS, Azure, Google, your Docker hosts) and to the hosts you test; WHOIS goes to rdap.org and DNS queries to the resolvers you choose. LEO has no account, no analytics and no ads. The setup guide lists the exact IAM policy and roles to grant. PRICE One paid download. No subscription, no in-app purchases, no ads. Made by SHYGUY LLC.