# Local AI Install Kit: Ollama setup check, a model picker by hardware, and private document Q&A

> Run models on your own computer with Ollama: install steps for macOS, Windows and Linux, a setup check that reads RAM, disk, GPU, the server, installed models, network exposure and cloud models, a model picker that turns your hardware into a memory budget, and docqa.py, private document Q&A that answers with sources and refuses any server or model that is not on this machine.

- URL: https://paperchasewebb.ai/bots/marketplace/items/local-ai-install-kit-ollama-private-document-qa/
- Type: Prompt pack
- Category: Solo business
- By: PaperChaseWebb
- Works with: Claude Code, Codex, Grok Bot, Cursor (named in the listing); Plain .md and Python 3.8 or newer scripts, standard library only; unit tested on Python 3.8 to 3.13
- Price: $39 USD
- Difficulty: Beginner
- Setup time: 45 min
- Time to result: A checked install and first document answers in an evening
- Format: Markdown (.md) file, local-ai-install-kit-ollama-private-document-qa.md
- Version: 1.0, updated 2026-10-31
- License: Use for your own work and for clients. Do not resell or republish the pack itself.

## What you end up with

A local model that fits your machine, a setup that is checked rather than guessed, and answers about your own files with numbered sources, with the files never leaving the computer.

## What you get

- 10 sections in one .md file
- Install steps for macOS, Windows and Linux, and how to keep the server private
- A model picker by hardware with a sizing rule you can check
- local_ai_check.py: setup check with fixes for every line
- docqa.py: private document Q&A with sources
- AGENTS.md for an agent that sets this up or runs it
- Troubleshooting table
- 28 unit tests

## Who it is for

People who want a model running on their own machine for private documents and routine work, set up right the first time and checked.

## Tools needed

- Ollama: runs the models on your computer
- Python 3.8 or newer: runs local_ai_check.py and docqa.py; standard library only
- You need: A Mac, Windows or Linux computer, 8 GB RAM or more (16 GB suggested)
- You need: 10 GB or more free disk
- You need: Python 3.8 or newer

Cost to run: The pack is the only purchase. The scripts use no hosted model tokens.

## Sample prompt: Set it up (Ollama)

```
Run python3 local_ai_check.py check and fix every FAIL line, asking me before you pull any model.
Then run python3 local_ai_check.py pick and tell me which model to start with and why.
```

From the pack.

## Free preview of local-ai-install-kit-ollama-private-document-qa.md

````markdown
# Local AI Install Kit: run models on your own machine with Ollama

A PaperChaseWebb pack, version 1.0, 2026-10-31. For macOS, Windows and Linux. Plain Markdown plus two small Python scripts (3.8 or newer, nothing to install besides Ollama itself).

This kit gets a model running on your own computer, picks a model that fits your hardware, and lets you ask questions about your own documents without any of them leaving the machine. The setup check tells you what is missing before you waste an afternoon. The document Q&A script refuses to talk to any server that is not this machine, and refuses Ollama's cloud models, so private files stay private by design and not by promise.

## What is in the kit

| Section | What it is for |
|---|---|
| 01 When local makes sense | What a local model is good at, where it falls short, and what it costs you in time and disk |
| 02 Install Ollama | macOS, Windows and Linux steps, first run, where models live, and how to keep the server private |
| 03 Pick a model | Memory budget by hardware, a sizing rule you can check yourself, and a starter list by machine |
| 04 Setup check | `local_ai_check.py`: RAM, disk, GPU, Ollama install, server, models, network exposure, cloud models |
| 05 Private document Q&A | `docqa.py`: index a folder, ask questions, get answers with numbered sources. Local only |
| 06 Use it from your tools | The local API, the OpenAI compatible endpoint, and agents on the same machine |
| 07 AGENTS.md | Rules for an AI agent that sets this up or runs it for you |
| 08 Troubleshooting | The problems people hit most, and the fix for each |
| 09 How we tested it | What was run, on which Python versions, and what was not tested |
| 10 The scripts | Both scripts and their unit tests, ready to copy |

## What you get

- **A setup check** that reads your RAM, free disk, GPU and Ollama status and prints PASS, FAIL or NOT_CHECKED with the fix for each line. It warns when the server listens on your whole network and when a cloud model is installed.
- **A model picker** that turns your hardware into a memory budget and marks each model as fits, tight, slow or too big.
- **Private document Q&A** over a folder of text, Markdown, CSV, JSON and HTML files, with sources on every answer and an index that only re-reads files that changed.
- **An AGENTS.md** so an agent that installs or runs this for you knows what it may do alone and where it must stop and ask.
````

## Test record

- Date: Oct 31, 2026
- What was run: 28 unit tests on Python 3.8, 3.9, 3.11, 3.12 and 3.13 with warnings as errors, from the source files and from the files extracted out of the pack, against a fake Ollama server; local_ai_check.py check against a real Ollama 0.40.13 server on Linux with no GPU.
- Result: All 28 tests passed on every version from both copies; the real server check reported RESULT PASS with warnings for RAM, no GPU and no models yet.
- Not run: Answer quality with real models (the test machine could not download models). macOS and Windows runs.

Copied from this item's dated test notes. What was not run stays in.

## Worth knowing

- Local models are smaller than the large hosted models and do worse on hard questions; the kit says where local fits and where it does not.
- Model sizes are planning numbers from a rule, not measurements.
- Tested 2026-10-31: 28 unit tests on Python 3.8, 3.9, 3.11, 3.12 and 3.13 with warnings as errors, from the source files and from the files extracted out of the pack, against a fake Ollama server; local_ai_check.py check against a real Ollama 0.40.13 server on Linux with no GPU. Result: All 28 tests passed on every version from both copies; the real server check reported RESULT PASS with warnings for RAM, no GPU and no models yet. Not run: Answer quality with real models (the test machine could not download models). macOS and Windows runs.

## Questions

**Do my documents leave my computer?** No. docqa.py refuses any server that is not this machine unless you pass a flag for your own server, and refuses Ollama cloud models.

**Do I need a GPU?** No. Without one, models run on the CPU, more slowly; the picker marks large models as slow.

## How to buy

$39 one time at https://paperchasewebb.ai/bots/marketplace/items/local-ai-install-kit-ollama-private-document-qa/. Stripe checkout, then an instant download page; the download link is also in your emailed receipt.

Seller: PaperChaseWebb Inc., mgmt@paperchasewebb.com