🛰️-primal-terrain

[elfin atlas]

Daily Recap — 2025-11-22

[carmine panther]

Lots of progress today. We have 'finalized' the Location Grid and the New Location flow. We have started on the "Location Overview" page and have the ability to edit the newly created locations!

[carmine panther]

We are removing most of the "toolbox" tools that we have been using for a while now.

We will be migrating to use a "detail panel" that is fixed width/allows for better user controls/interaction.

Someday we will introduce the ability to "collapse" (full screen the map)

We will keep the concept of a "map toolbox shortcut" for certain things like "Sun Controls" or "Weather Controls" etc. but for the bulk of what the user will leverage in terms of tooling will be moved into this alternative UI/UX.

[lime stag]

Ooooh

[carmine panther]

We opted out of "Two Sections" on LEFT side from the UI Mock and went with a single one called "Explorer." This our stopping point for today! 🙂

[stuck kernel]

whats this btw

[carmine panther]

It's the software that I've been streaming the creation of.

I'm calling it primal terrain.

What you're seeing in that screenshot above is very out of date. 😉

Holistically the software is first built for me. I am however building it at an Enterprise grade level.

I have a strong need for a software that can be used for site planning/homestead planning.

I manage a few spaces, and I've come to terms that it's extremely hard to keep track of all my paper notes across locations and want to centralize and streamline the process.

As the seasons evolve so do the plans.

This software initially will not have any task-related activity however the future state would certainly allow for things like that.

I during lunch came up with my next stream plan/actionables and am currently scheduled for 6:30 CST tonight / about five and a half hours from now to stream.
I'm hoping now that I'm done with all my wood carvings and gift making I should be back on a more frequent schedule, but it would be the best way for me to provide a demo.

[carmine panther]

#1443335677709254746 message

This openai Discord project post is also my most up-to-date on the project.

I have three themes supported and the ability to log in and impersonate users with different roles for those themes, so The styling and look may vary.

[stuck kernel]

oh that sounds amazing

are the people in this server collaborating to this project?

[carmine panther]

I've considered open source and there may become aspects that I do open up for open source, like the Toolbox which is a very extensible component framework within my app already/one of the first things built for this project.

The reason I am not is because I initially want to make it fully operational from a local host context as this product is something I'm going to use almost daily.

I have a strong need for this project and a need to build it quickly, I think opening it up will slow me down tbh, and have a very strong vision of what makes this project "live." (aiming 100th stream day to be live live, but I work on this project most days even if i don't stream)

With that said I have a very strong opinion on its development at this point.

I'm not against bringing in some additional developers within my network for collaborating/or people here if parties here are interested enough over time-- the route to do it would be to not only have "production" github, but a dev, uat, prod sort of infrastructure, which again would slow me down at this point.

Attached is the product of "shade map" whos developer I've considered to bring in/we talk sometimes~ I've contributed to his library for shadow simulation.

As you can see he is not just using "building data" for his shadow simulation, but also tree data which I think is based on Lidar and then converted to DEIM. ^^

At some point I wish to work out a deal with him on my eventual $40 monthly cost if I went enterprise/update his UI/UX. (mine yellow ss))

---

This community though is encouraged to be very collaborative. We learn from eachother. During streams, yes I'm working on primal-terrain, but when people are in discussions~ we often derail into whatever someone is working on or new technologies or just discussions .

I imagine this community will have two sides, and a merging of the two over time. One of the domains I own is forged-alliance which I imagine is a merging of tech + folky/homesteady/diy people. (shared as a "word definition" vs a concept of a new community)

In this community I will always help facilitate/encourage others to share real life projects/not only the tech side. That side of this discord will EXPLODE come spring...

Since we are in Winter there is extra emphasis on tech projects. I've learned LOTS from the people that have joined this stream and have had a lot of growth/ideations come out from the "hive."

Since this software is being developed from the perspective of enterprise grade whenever I do stream, I do my best to explain why / what I am doing~ sharing tips and tricks based on the current audience skill level/set.

We are a small enough that we can do this and I do hope as we grow the people in this community keep offering ideas to eachother. (ive seen and been encouraged by organic convos I've seen thus far)

---

Another major driving reason why this isn't open source though at this point is I believe I have a very clear path once this project is completed to go to market with it vs just for my use.

I want to have a free version of this application available to everyone if I do so. With things like shadow simulation or "multi-user-tenancy" or "public facing domain" to be pay walled.

I ended up finding an older model really cheap on marketplace for a Moasure (attached) which I will soon make an integration for. (it arrives Friday!)

The integration developed for Moasure will involve translating X/Y/Z sort of data into a way that we can "anchor it" to GeoJSON/Exact location data. This type of feature I could see being pay walled/I am even considering making an api outside this app for that if that sort of thing doesn't already exist.

I will be mapping out several of the properties I manage as "test data" using Moasure and to start planning future outdoor projects on stream.

for me Winter is when I make HUGE progress digitally and over time the people here will see a wide variety of things. Like mushroom log drilling and growing, etc.

[stuck kernel]

are there people here currently working on some open source stuff

would love to contribute

[carmine panther]

@stuck kernel

@paper sedge may be a good canidate.
https://github.com/shadowquillapp/shadowquillapp

[stuck kernel]

this runs locally?

[carmine panther]

Yup. https://shadowquill.org/
it takes a small amount of setup. The vision of the product is alturistic from what I've gathered and observed. I'm not fully sure what is 'needed' for complete release or what features could be worked on etc.

[stuck kernel]

I got a 512gb laptop my guy

I dont have space for this I think

Left with 50gigs

do you know whos working on this

[carmine panther]

@shoomy is working on it.

[elfin atlas]

Daily Recap — 2025-12-26

[paper sedge]

Hey yeah

So it needs like 10gb

Around 9gb for ollama + gemm3:4b and the ShadowQuill app itself is around 600mb

[carmine panther]

@paper sedge - Critical is interested in helping out on open source projects. I've been following your git and have seen mostly version sort of changes, so not sure if there is any dev you have left on that product--

If you have interest in talking about your product/some "dream future states" it would be great if you could DM / connect with Critical in finding ways he might be able to help. 🙂

Real life positional data generated using a Moasure device.

Will attempt to add to map in 10 mins on stream!

[carmine panther]

I was able to quickly convert. However I've been running into a challenge on how I originally was fundamentally going to use this device, to quickly and accurately, map items/trackables.

Part of this is Z is relative vs "explicit" to actual elevatation data. I have some experimental ideas in my head and will need to marinate on this prior to fully committing to a path.

[carmine panther]

Current lead is to leverage "Three.Js" to give a 3d model of Moasure measurements.
https://discord.com/channels/1303190929733320735/1442668096463831111

[carmine panther]

Moasure Data in ++ Three.JS integration complete. Tomorrow stream focus on cleanup/UI/UX support for this behavior...

[carmine panther]

UI/UX changes for 3d Viewers.

Procedural model / design pattern for creating trackable models in place for future scale.

[carmine panther]

Big. I've secured access to satellite data. 😉

[stuck kernel]

REALLY?

Will it help you to map out ur stuff better

[carmine panther]

Yes! Having the ability to display the Satellite data is a HUGE win. There are challenges with this API. Its RIDICULOUSLY complicated lololol

Leave it to .gov to make things silly difficult 😛

The challenge with this is that it really "pins" the "need" for a dedicated server. The tile data for Wisconsin **ONLY **takes 2TB of space lol.

I may however in the now design it to "tile on need" ; essentially only download "terrain tiles" of a users locations spatial bounds.

[carmine panther]

||
Current chaellenge I'm having and running into with using a Ollama localized model that my computer can run and host without bogging it down and a model that has access to tooling/APIs is that it doesn't have the capability to perform geometry calculations.

Sample prompt of :

given [lat,lng] create a 30x30' geoJSON poylgon

Is resulting in close but incorrect geometries.

[stuck kernel]

Interesting

[carmine panther]

Ended up getting the largest model my comp can local host which is

Gpt-osss:20 which to solved issues. It also allowed for calling or Apis to be consistent.

😉

Those real small ollama models are kind of ehh if you want any form of tooling or need 'calculative' sorts of outputs.

[carmine panther]

---

Purchasing a separate computer (currently I am only on laptop) to act as my LLM server and tiling server. This won't have immediate impact on the application, but it will allow for me to eventually make everything publicly accessible 🙂

[stuck kernel]

Wild idea: run it on a raspberry pi

[carmine panther]

Great idea!

Issue is the model that resulted in proper geometry calculations requires more RAM than I believe a pi could do.

I also need something to store multiple terabytes of mapping data.

At least eight terabytes for Wisconsin alone.

Fortunately I found specs on Facebook at a price extremely reasonable/same price I sold my PC at 2 years ago and this would be considered a considerable upgrade to that one.

You raise a great point though. I've seen USB sort of LLMS for sale recently, The olama model that is very accessible is like 1 GB in size It is really good at conversations. Just not geometry. 😭

I also like the idea of being able to run many agents concurrently/scaling over time as I understand more.

[stuck kernel]

Gemma 3n on a mobile

[carmine panther]

We were able to toolchain for the first time. Thanks to Robert we determined how to maximize the amount of kvCache available for our local model.

Previously we were not getting toolchaining as our configured token limits were 'flushing' and losing context.

With qwen2.5:32b-instruct model
I am able to run 16k context leaving myself 2gb of ram haha

It will be very nice when I eventually am able to host the application on a dedicated machine with more ram. 🙂

[quartz scroll]

How big is the LLM?

https://www.zdnet.com/article/claude-code-alternative-free-local-open-source-goose/

[carmine panther]

I'm leveraging right now:
qwen2.5:32b-instruct

My current issue in tool chaining locally is with the Qwen2 model I can only have a 16k context length with my current amount of ram on this laptop. (seems to break after 2-3 tool chains)

I want to move to this model once I have more ram available and hopefully I can do like 64k context length...
https://huggingface.co/Qwen/Qwen3-30B-A3B-Thinking-2507

I was using Codex 5.3 today to actually do natural language seeding. It was NAILING the requirements/this was my output. I was having it create like 50 trackables at a time.

It really makes me consider just saying F "local" and pay for cloud chat... Which I may do someday...

Next steps with the tool (tomorrows stream) is to work on satellite. I think I need to improve that feature now to more accurately place my plans...

I've also been making a bunch of like small enhancements in the app over the last few days as I cook/just better behavior and polishing several features that were just shelled.

Plenty to do, but the above SS is the actual future property plans, just not "pixel perfect" in terms of placements etc...

It's really nice to be able to say things for record creation like:

Create 50 llillac bushes in the X Layer, in a new zone called Y, spaced 7' apart starting at the NW corner of the property-- going E to W based on the "Tax Parcel" boundary. Offset from this boundary 10' south

was absolutely nailed on first time...

[proven hornet]

Hi all, this sounds too sick. Love this!

[carmine panther]

Hell yeah brother! Starting next week I should be able to get back to the groove of things and pump some nice updates/we stream each Tuesday! I did shell out the api path to get satellite data--

but grabbing it and then hosting is next up.

You a homesteader or whats the big draw to you?

[carmine panther]

We now have satellite data.

We have a script that takes a spatial area and has the ability to download the scenes from the USGS and then host it on our Docker server that is currently exposing our map tiles. We then wired this into our show satellite view icon.

I'm going to look at if I can get higher quality satellite imagery.

My next focus in the app is to really nail down the slash map UI.

I believe I need to actually undergo a structural change of zones, layers, and trackables to just have trackables

To allow expanding on that functionality, ultimately simplifying the UI.

Over time I can introduce a concept of grouping which solves the reason behind layer in zones in the first place without introducing confusion.

I am only storing a small subset of satellite data.

Until I dedicate the hosting of this data, the image here is as much as I have right now stored / hosted on development machine.

[carmine panther]

I am considering a stream tomorrow in addition to the normal Tuesday stream.

Tuesdays stream we will be showcasing the "Atlas" browser and how it can be leveraged effectively from a design lens.

I have also underwent in last 2 weeks an EXTREMELY large refactor recently that has redefined the entire map view in terrms of interaction and experience. (over 50k lines effected)

I feel VERY positive about the changes I've made and have been introducing new functionality around being able to draw polygons, polylines, and points within the map

and we are really doubling down on "finalizing" this page and its functionality.

I'm starting to see the end of the tunnel in terms of being able to have a hosted version of this Primal Terrain application is out and available online.

I could see it happening as early as two to three months from today!

I'll make sure to post some screenshots after stream on Tuesday in terms of the new UI look!

[carmine panther]

It's the small things. (he he ignore the wispr flow button on left)

[carmine panther]

Update.

[carmine panther]

GitNexus is an interesting way to visualize your repository and understand some of its dependencies. The problems it resolves or addresses in the github feel very valid.

It's been facinating to look at my project through a different lens.

I can see this being extremely helpful if you want to minify the work your LLMs have to use in terms of understanding dependencies etc.

[carmine panther]

• Migrated from Docker self hosting satellite and map tiling. No longer need to self host.
• Migrated Agent to use Anthropic from self hosted Ollama. (this may change/become multiple providers)
• Introduce "Vegetation" (canopy data) to shadow simulation.

[carmine panther]

We have a staging environment set up in Vercel. I am hoping to get to a stable version of the application and then open it up for a small user group testing for feedback.

I have plenty of work I wish to do before I release the production version of this website
But I'm garnering feedback at this point and getting all my devops in place so that I have a release strategy, etc

---

Current next steps are to work through some of the speed insights as I have a lot of cumulative layout shifts as a result of how I am doing styled components. It will be interesting to improve the experience over time

** send a dm if u want to be part of the user testing group!

[carmine panther]

I made massive improvements along the board. It feels much smoother.

I just finished adding search embeddings/pg vectorings to trackables and introducing additional tooling to allow for that.

essentially "natural language" database search.

I will be fine-tuning my scoring as I still do trig search, and word matching in addition as part of the scoring to determine results

Next I will be taking a pass at the agent orchestration layer as it is overly complicated and by streamlining it, it should create a better user experience
Once that is good I will then work on
"Draft mode"
Which once this is complete, I will open up for user testing

[carmine panther]

I will send out emails to anyone that is interested in ALPHA testing access. DM interest.

Alpha release coming soon for Primal terrain and we'll be hosted at our primal-terrain.com domain.

Data that gets created during this time period may not be preserved, but users can export their data at any time and before I do I will give a week's heads up

Actual release date is scheduled for 06-21

Beta access will be granted earlier to test groups that also may end up using the platform and would not result in a wipe of data.

This Alpha test I am just looking for any form of feedback. I know the product isn't perfect, but I've landed on most of what I wish to see in terms of map (minus bugs and some clarity fixes) and the unpolished assistant that is my next primary focus in terms of ui and actual function.

[carmine panther]

ALPHA - Invites have been sent out!

[carmine panther]

Alpha testing has been a great success. We are now working on applying some of the feedback end users have had and working on our launch!