“Near me” is only useful when the assistant knows where to start. A local model's training data cannot tell it your current location or whether a nearby place is open today.
OGAD (Off Grid AI Desktop) can request your Mac's location and use it as the starting point for a nearby task. With a local model and Web Use, you can ask for places that meet a clear need, then inspect the current pages behind the result.
What would you like to do with Off Grid AI?
Have a feature or use case you would like us to support? Tell us what you want to do and which device you use.
Write to support@offgridmobileai.co, join our Slack community, or talk to us on Reddit.
Local model inference and fresh place information are separate parts of this workflow. Web listings need internet. The search or map service can receive the location used in the query, even when the AI model runs on your Mac.
Ask for the place you need, not a generic list
Give the assistant a useful selection rule:
Find three parks near my current location. Use current web sources and include a link for each. Tell me which details you could confirm.
Or provide the starting point yourself:
Find public libraries near this neighborhood: [name]. Check their official pages for opening hours and link the sources.
A specific request gives the assistant something to compare. If you need step-free access, an opening time, or a particular facility, name it and ask for a source. A listing that does not mention a facility should not become an assumed yes.
This is a way to gather current options. Check the source page before leaving, especially for opening hours that can change.
Let the Mac provide the starting location
When you ask for a current-location task, OGAD has a native tool that requests coordinates from macOS. If macOS asks for permission, allow it when you want the app to use your location for that request.
If Location Services are off or the request fails, provide an address or neighborhood. You do not need to keep retrying location permission just to compare places around a starting point you already know.
A location estimate can have limited accuracy. For a large campus, station, or city center, an explicit starting address may be more useful than the device's coordinates.
The Mac location workflow is included in OGAD 0.0.51.
Get your first useful shortlist
- Install OGAD and prepare a local, tool-capable chat model.
- For the stable 0.0.51 Web Use workflow, activate Pro and turn Assistant on in chat.
- Ask for a small number of nearby options and state the condition that matters to you.
- Allow the macOS location request, or supply an address or neighborhood instead.
- Let Web Use inspect current pages. Read the result and open its sources to check the details you will rely on.
Web Use runs in OGAD's built-in browser. Keep the first task to finding and comparing public information. A request to make a booking or send a message is a separate action with a different result.
If the assistant cannot get your location, it should ask for a starting point rather than silently use a guessed place. If a page fails to load, a partial result is not evidence that no suitable places exist.
Keep local processing and web access clear
Select a local model to keep the model's reasoning on your Mac. That does not make a web lookup offline. Queries and requests still go to the sites used for the lookup, and the location can be included to find nearby results.
You can avoid sharing device coordinates by naming a broader area in the request. That trades a less exact starting point for a query based on the place you chose to provide.
After downloading a model, you can ask offline questions about information you already have. Finding current opening hours or newly changed listings remains an online task.
Turn the shortlist into a decision
Suppose you need a quiet place to read for an hour before a meeting. Ask for a small set of libraries in the area you choose, with today's opening hours and official source links. If step-free access matters, include that requirement explicitly.
A useful comparison has one row per place and separates confirmed facts from missing details. Ask:
For each option, show the address, the opening hours you could verify and the source page. Put “not confirmed” beside any access detail the page does not establish. Do not make a booking or contact the venue.
Open the most promising source before leaving. Check the date and whether the hours apply to the day you need. If access information is missing, use the venue's own contact route to confirm it rather than allowing a model inference to decide the trip.
Use a broader starting point when exact location is unnecessary
For a task such as finding weekend parks across a city district, you can provide the district name yourself. That may be enough to produce a useful list without requesting device coordinates. For a narrow search around a large station, a named entrance can be more useful than a rough location estimate.
If the results look far away, check the starting point before changing models. If all options lack a required facility, narrow the source search or state which condition can change. The assistant helps gather evidence for your choice; you can revise the task without treating the first list as the only possible answer.
Turn a vague search into a practical choice
Download OGAD, choose one nearby task, and ask for a short list with sources. Let the Mac supply the starting point or give one yourself, then use the confirmed details to choose where to go.
Top comments (0)