Windows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallCrashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteiTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
Submit the long request with ?async=true, keep the request_id the gateway returns, and poll /v1/requests/{request_id} until the job reaches a terminal state. Your application never has to hold one HTTP connection open for the whole model run, and stopping the poll does not stop the work on the gateway.
How the asynchronous flow works
A normal synchronous call sends a request and waits on the same connection until the model finishes. For a long chat completion, a vision task or an image job, that connection may be open for minutes. Off Grid AI’s own article on asynchronous API requests describes a different pattern: the first call returns quickly with HTTP 202 and a request_id, and the client checks on the job separately. The article states that the server keeps processing if the client stops polling, so a client-side timeout only ends your waiting, not the task.
The gateway is the Off Grid AI Development Kit (OGAD) project’s local, OpenAI-compatible API. Its repository describes one gateway for chat, vision, image, audio and embeddings, with inference running on your own hardware. Asynchronous submission changes how your client waits. It does not give a model or feature that needs internet access the ability to run offline.
Step-by-step setup
- Start OGAD and load a local model. The repository lists
http://127.0.0.1:7878/v1as the default base URL. Check the port shown by your running installation, because a local setup can differ. - Discover an installed model. Call
GET /v1/modelsand take the model ID from that list. Do not copy a model ID from a sample, since the IDs on your machine depend on what you have installed.curl -s http://127.0.0.1:7878/v1/models - Submit the request to the asynchronous endpoint. Send the same body you would send to a synchronous call, with
?async=trueadded toPOST /v1/chat/completions. Expect HTTP 202 and arequest_idin the response.curl -s -X POST "http://127.0.0.1:7878/v1/chat/completions?async=true" -H "Content-Type: application/json" -d "{"model": "$MODEL_ID", "messages": [{"role": "user", "content": "Summarise this file."}]}" - Store the ID immediately. Write the
request_idto your job record before doing anything else, so a crash or timeout in your client does not leave you with an orphaned job you cannot find. - Poll the request resource. Call
GET /v1/requests/{request_id}with the same ID each time.curl -s http://127.0.0.1:7878/v1/requests/$REQUEST_IDThe Off Grid AI article’s sample waits two seconds between polls, uses a 30-second timeout for each HTTP call, and stops polling after five minutes. Those values are choices in that sample, not limits of the gateway, so set your own interval and deadline to suit your model and hardware.
- Read the result using the originating endpoint’s schema. The polling route is shared, but a completed chat result and a completed image result do not have the same payload. Parse the output according to the endpoint you submitted to.
- Save the output in your own storage. Keep completed results, and the inputs you need to rerun a job, outside the gateway (see the restart section below).
Job states and what your client should do
The article’s sample treats queued and running as pending, completed as success and failed as failure. Your client should map each state to an action:
#1 Best Overall
- EVOLUTION AMD RYZEN AI MAX+ 395 MINI PC - GMKtec EVO-X2 is the next evolution in AI mini PC Ryzen Strix Halo series. Thanks to AMD Simultaneous Multithreading (SMT) the core-count is effectively doubled, to 32 threads. Ryzen AI Max+ 395 has 64 MB of L3 cache and can boost up to 5.1 GHz, depending on the workload. The Ryzen AI Max+ 395 is currently rated as the "most powerful x86 APU" on the market for AI computing.
- AI NPU with XDNA 2 ARCHITECTURE - Powered by 16 “Zen 5” CPU cores, 50+ peak AI TOPS XDNA 2 NPU and a truly massive integrated GPU driven by 40 AMD RDNA 3.5 CUs, the Ryzen AI MAX+ 395 is a transformative upgrade and delivers a significant performance boost over the competition. The Ryzen AI Max+ 395 excels in consumer AI workloads like the llama.cpp-powered application: LM Studio. Shaping up to be the must-have app for client LLM workloads, LM Studio allows users to locally run the latest language model without any technical knowledge required and unleash their creativity and productivity.
- AMD RADEON 8090S iGPU GAMING PC - The AMD Radeon RX 8060S offers all 40 CUs with up to 2.9 GHz graphics clock and uses the new RDNA 3.5 architecture. The powerful iGPU is positioned between an RTX 4060 and 4070 laptop GPU and therefore enables gaming in FHD at maximum details in most demanding games. The 8060S can also utilize the full 128GB pool, which is perfect for running LLMs such as Deepseek 70B Q8, which runs comfortably on this machine.
- EIGHT CHANNEL LPDDR5X - LPDDR5X is a new ground breaking memory small form factor installed on-board. With blazing speeds up to to 8000MT/s, it runs 1.5x faster than the DDR5 SODIMMs; 90% better performance over DDR5 SODIMMs in video conferencing and photo editing; 30% better performance in productivity apps; 12% better performance in digital content workloads.
- QUAD SCREEN 8K DISPLAY SUPPORT - EVO-X2 AI Mini PC support 4-screen 4K/8K output via HDMI 2.1 (8K@60Hz), DisplayPort 1.4 (4K@60Hz), and dual USB 4 40Gbps Transfer speed (supporting PD3.0/DP1.4/DATA). Ideal for gaming, video editing, and multitasking, it provides expansive and crisp multi-display support.
| State or condition | Meaning in the article’s model | Recommended client action |
|---|---|---|
queued |
The job is accepted and waiting. | Keep the same ID and poll again later. Do not resubmit. |
running |
The model is working on the job. | Keep the same ID and poll again later. Show a pending indicator to the user. |
completed |
The job has finished successfully. | Parse the result for the endpoint you used, then save it. |
failed |
The job ended with an error. | Show the returned error. Check the model choice and input before retrying, rather than retrying blindly. |
| Poll cannot connect | The gateway may have stopped or may not be reachable. | Confirm OGAD is still running before you create a new request. |
| Your polling deadline passes | Your client has stopped waiting. The server task is not cancelled. | Keep the ID if you plan to check the same job later while the gateway stays running. |
Why duplicate submissions are the main risk
The most common mistake is treating a slow poll as a failed submission and sending the request again. That starts a second model run on the same hardware, which slows both jobs and can leave you with two results to reconcile. Deduplicate on your own side by tying each user action to one stored request_id, and only resubmit after a job reports failed or after you have confirmed that the original ID is no longer available.
Restarts, storage and record limits
The article states that request records are held in memory and are lost when the application restarts. After a restart, an old ID may return that the record is unavailable, and that does not tell you whether the model ever produced output. The article also reports an in-memory store bounded to 500 records. The article does not name the exact release it describes, so treat both points as documented behaviour of that article and confirm them for your installed version before building guarantees on them.
Rank #2
- Built for Local AI Development: AMD Ryzen AI Halo is designed for local AI development and inference, featuring 128GB unified memory and support for up to 200B parameter models to build and run intensive AI workloads locally.
- 128GB Unified Memory: Features 128GB LPDDR5x unified memory at 8000 MT/s with 256 GB/s memory bandwidth, providing a shared memory pool across the CPU, GPU, and NPU to support larger AI models.
- AMD Ryzen AI Max+ 395 Processor: Features 16 cores, 32 threads, and Zen 5 architecture, paired with AMD Radeon 8060S integrated graphics featuring 40 RDNA 3.5 compute units and an AMD XDNA 2 NPU with up to 50 TOPS.
- Linux AI Developer Platform: Purpose-built for Linux-based AI development with full AMD ROCm software support and preloaded tools, models, and workflows optimized for local AI development.
- Compact, Connected Design: Includes a 2TB M.2 SSD, 10GbE LAN, Wi-Fi 7, Bluetooth 5.4, USB-C connectivity, and HDMI 2.1b.
- Persist the request body, the
request_id, and the completed output in your own database or file store. - On startup, reconcile unfinished job records: poll any IDs you still hold, and resubmit only those whose records the gateway no longer has and whose output you never saved.
- Do not rely on the gateway as the only copy of a result you need to keep.
Synchronous versus asynchronous calls
A synchronous call is simpler to write and suits short requests. The asynchronous pattern suits requests your client cannot afford to wait on in one connection. The comparison below uses the axes that matter when you choose.
Free tools Windows power users keep installed
One-click scans. No signup required.
| Consideration | Synchronous request | Asynchronous request with polling |
|---|---|---|
| How long the client connection stays open | For the whole model run. The Off Grid AI article does not give a versioned timeout guarantee for this path. | Only for each short submit and poll call. |
| Progress or pending state | Not available until the response returns, unless your app adds its own signal. | Available through the queued and running states. |
| Duplicate work after a client timeout | Retrying the call usually starts a new run. | Polling the same ID does not start a new run. |
| Recovery across an app restart | Depends on your own saving of results. | The gateway record is lost on restart, so your own storage is essential. |
Verify against your installed release
Endpoint support, request and response schemas, the port and the record limit can all vary by release. The OGAD repository points to a live API reference at /docs and an OpenAPI document at /openapi.json on the running gateway. Use those to confirm that your endpoint accepts async=true, what it returns, and how your version handles job records before you write production code against them.
Rank #3
- EVOLUTION AMD RYZEN AI MAX+ 395 MINI PC - GMKtec EVO-X2 is the next evolution in AI mini PC Ryzen Strix Halo series. Thanks to AMD Simultaneous Multithreading (SMT) the core-count is effectively doubled, to 32 threads. Ryzen AI Max+ 395 has 64 MB of L3 cache and can boost up to 5.1 GHz, depending on the workload. The Ryzen AI Max+ 395 is currently rated as the "most powerful x86 APU" on the market for AI computing.
- AI NPU with XDNA 2 ARCHITECTURE - Powered by 16 “Zen 5” CPU cores, 50+ peak AI TOPS XDNA 2 NPU and a truly massive integrated GPU driven by 40 AMD RDNA 3.5 CUs, the Ryzen AI MAX+ 395 is a transformative upgrade and delivers a significant performance boost over the competition. The Ryzen AI Max+ 395 excels in consumer AI workloads like the llama.cpp-powered application: LM Studio. Shaping up to be the must-have app for client LLM workloads, LM Studio allows users to locally run the latest language model without any technical knowledge required and unleash their creativity and productivity.
- AMD RADEON 8090S iGPU GAMING PC - The AMD Radeon RX 8060S offers all 40 CUs with up to 2.9 GHz graphics clock and uses the new RDNA 3.5 architecture. The powerful iGPU is positioned between an RTX 4060 and 4070 laptop GPU and therefore enables gaming in FHD at maximum details in most demanding games. The 8060S can also utilize the full 64GB pool, which is perfect for running LLMs such as Deepseek 32B, which runs comfortably on this machine.
- EIGHT CHANNEL LPDDR5X - LPDDR5X is a new ground breaking memory small form factor installed on-board. With blazing speeds up to to 8000MT/s, it runs 1.5x faster than the DDR5 SODIMMs; 90% better performance over DDR5 SODIMMs in video conferencing and photo editing; 30% better performance in productivity apps; 4% better performance in digital content workloads.
- QUAD SCREEN 8K DISPLAY SUPPORT - EVO-X2 AI Mini PC support 4-screen 4K/8K output via HDMI 2.1 (8K@60Hz), DisplayPort 1.4 (4K@60Hz), and dual USB 4 40Gbps Transfer speed (supporting PD3.0/DP1.4/DATA). Ideal for gaming, video editing, and multitasking, it provides expansive and crisp multi-display support.
Off Grid AI’s article also notes that long-running tools such as image generation can use asynchronous handling with a poll URL. Check that each endpoint you use supports this pattern before assuming the same polling code works for all of them.
The sample’s five-minute polling window is a client decision. It is not a server-side maximum runtime, so a job that takes longer than your deadline may still finish. Keep polling for as long as your application can afford, and keep the ID until you know the job’s outcome.
Quick Recap
Rank #4
- 【Leading AI Mini Workstation】MINISFORUM AI MS-S1 Max Workstation comes with AMD Ryzen AI Max+ 395 processor, which uses AMD's latest generation Zen 5 architecture. It has 16 Cores and 32 Threads, the boost clock is up to 5.1GHz. The overall processor performance is up to 126 TOPS, and the NPU performance reaches up to 50 TOPS. AMD Ryzen AI enables improved productivity, advanced collaboration, and improved efficiency.
- 【AMD Radeon 8060S Graphics 】The MS-S1 Max Mini PC equipped with AMD Radeon 8060S Graphics which built on the new generation of RDNA 3.5 architecture AMD graphics, it brings ultra-high frame rate experiences and advanced content creation features anywhere and delivers staggering performance. It can handle all your computing and multimedia tasks efficiently.
- 【Five 8K Video Output】This MS-S1 Max Workstation comes with five video outputs, 1x HDMI (8K@60Hz), 2x USB4(40Gbps,Alt DP2.0,PD out 15W) and 2x USB4 V2(80Gbps,Alt DP2.0,PD out 15W) Outputs, which support multiple monitors display at the same time and provide a larger and wider filed of view and improve your work efficiency. It is used in fields that require high-performance computing and graphics processing, including digital signage and securities trading, as well as work that uses CAD, such as engineering design, scientific calculations, animation production, and post-production for movies and television
- 【 Fast and Stable Wire & Wireless Speed】It comes with Two 10G Lan Ports for wired connection and and Wi-Fi 7 / BT5.4 for wireless connection, which increased the network speed greatly and expand its functions and improved performance of computer to a large extent and allows you to use more networks such as software routers (OpenWRT / DD-WRT / Tomato etc.), firewalls, NAT, network isolation etc.
- 【Large Storage & Flexible Expandability】This Workstation equipped with 64GB LPDDR5-8000MHz + 2TB M.2 2280 PCIe4.0 SSD. There is another PCIe4.0 SSD slot available for up to 8TB, these SSD slots are compatible with RAID0 and RAID1, you can store movies, videos, photos, important files easily. What’s more, it also comes with 1x standard PCIex16 slot(PCIe4.0x4) inside.
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
Recommended Free Tools
