LocalAIModelManager 1.0.10

🤖 Local AI Model Manager

Find every AI model on your computer — and clean up the ones you don't need.

If you use local AI tools like Ollama, LM Studio, or Hugging Face, your disk fills up fast. A 20 GB model here, a few more there — before you know it, hundreds of gigabytes are gone, and you have no idea what's where or what's safe to delete.

Every tool stores its models in different places, with different naming, in different ways depending on your operating system. Tracking it all manually is nearly impossible — you'd have to know where each tool hides its files on Windows, macOS, and Linux.

Local AI Model Manager solves this. One command scans your entire system and shows every model in a single clean table. You see the name, size, format, type, parameters, location, and status of each model. Then you decide what to keep and what to delete.

Who is this for?

• 🧑‍💻 Developers juggling multiple AI tools and losing track of disk space
• 🎯 Power users who want to see exactly what's stored where
• 🧹 Anyone whose SSD is full of models they downloaded once and forgot about
• 💰 People who paid for storage and want it back

Works on Windows, macOS, and Linux. No setup, no config — just install and run.

---

📦 Install

Requirement: .NET 10 SDK must be installed.

dotnet tool install --global localaimodelmanager

That's it — the command localaimodelmanager is now available in your terminal.

Update & Uninstall

dotnet tool update --global localaimodelmanager      # update
dotnet tool uninstall --global localaimodelmanager   # remove

---

🚀 Quick Start (30 seconds)

localaimodelmanager            # show all models on your system
localaimodelmanager summary    # see where your disk space is going

That's all most people need. The tool finds everything automatically — no setup, no config.

---

✨ Why use this?

Without this tool	With this tool
❌ No idea how many models are on your disk	✅ One table shows every model
❌ Every tool stores files in a different place	✅ All locations scanned automatically
❌ Can't tell which models are duplicates	✅ Duplicates identified and grouped
❌ Broken downloads waste space silently	✅ Broken models flagged for easy cleanup
❌ Deleting feels risky — what if it's in use?	✅ Running models are protected from deletion
❌ No idea what's taking the most space	✅ Summary view shows biggest/oldest models

---

📋 Commands

summary — Where is your space going? ⭐

The fastest way to understand your disk usage. Shows a storage breakdown instead of a long list.

localaimodelmanager summary
localaimodelmanager summary --engine Ollama    # scope to one tool

The summary shows three things:

1. By engine — how many models and how much space each tool uses
2. Top 5 by size — the biggest models on your system
3. Oldest 5 — models you haven't touched in months (shown with "days ago")

---

list — Show all models

Lists every AI model in a clean table — with format, type, parameters, size, date, status, and location.

localaimodelmanager list

Filters

Option	What it does	Example
--engine <name>	Show only one tool's models	--engine Ollama
--format <name>	Show only one file format	--format GGUF
--name <keyword>	Search by model name	--name llama
--type <type>	Show only one model type	--type LLM
--sort <key>	Sort the results	--sort size
--deletable	Show only deletable models	--deletable

Model types: LLM · Embedding · Vision · Multimodal · Audio · Image

Sort keys: engine (default) · name · size · date

Examples

localaimodelmanager list --engine Ollama              # only Ollama models
localaimodelmanager list --format GGUF --sort size    # GGUF files, biggest first
localaimodelmanager list --name llama --type LLM      # find Llama LLMs
localaimodelmanager list --sort date                  # oldest first (find stale models)

---

delete — Remove models you don't need

Deletes models matching your filters. Always asks for confirmation before removing anything.

localaimodelmanager delete --engine Ollama --model "gemma4:e4b"

Options

Option	What it does
--engine <name>	Delete only under this tool
--model <name>	Delete models whose name matches
--format <name>	Delete only this format
--all	Confirm deletion without further filters

⚠️ You must provide at least one filter. The tool will never delete everything by accident.

Examples

localaimodelmanager delete --engine Ollama --model "gemma4:e4b"   # one specific model
localaimodelmanager delete --engine Ollama --all                  # all Ollama models (be careful)

After confirming, the tool reports how many models were removed and how much space was recovered.

---

help — Show usage

localaimodelmanager help

---

⚡ Quick Reference

Command	What it does
localaimodelmanager	List all models (same as list)
localaimodelmanager summary	Storage breakdown by engine
localaimodelmanager list --engine <name>	Filter by tool
localaimodelmanager list --sort size	Biggest models first
localaimodelmanager list --sort date	Oldest models first
localaimodelmanager delete --engine <name> --model "<name>"	Delete one model

---

🧠 What it detects

The tool automatically finds models from all these tools and formats — no setup required:

AI tools: Ollama · LM Studio · Microsoft Foundry Local · Hugging Face · GPT4All · Jan · llama.cpp · MLX (Apple Silicon) · AI Toolkit

Model formats: GGUF · ONNX · SafeTensors · PyTorch · TensorFlow / Keras

🔍 The tool also detects broken downloads — if a download was interrupted, it's marked as Broken so you know it's safe to delete and reclaim the space.

---

📊 Understanding the table

Column	Meaning
ID	Row number (for reference)
Engine	Which tool or format the model belongs to
Model	The model name
Format	File format (GGUF, ONNX, SafeTensors, etc.)
Model Type	LLM, Embedding, Vision, Multimodal, Audio, Image
Parameters	Model size in parameters (7B, 13B, 70B, etc.)
Size	Actual disk space used
Modified	Last modified date
Status	Idle, InUse, Broken, Orphaned
Location	Full path to the model on your disk

On macOS, the same file can sometimes be reached through two paths. When that happens, the alias is shown in grey below the main path — it's the same physical file, not a second copy, so the size is only counted once.

---

🖥️ Cross-platform

The same tool runs identically on Windows, macOS, and Linux. It detects CPU, GPU, NPU, cores, and memory on every platform.

---

📸 All Screenshots

1. System info and full model list — your system details, then every model found.

2. Models grouped by engine — each model with format, type, parameters, size, date, status, and path.

3. Summary view — storage usage broken down by engine, with the biggest and oldest models.

4. List with filters — narrow down by engine, format, name, type, or sort.

5. Delete confirmation — review before anything is removed.

6. Space recovered — how much was freed after deletion.

7. Broken downloads flagged — interrupted downloads marked for easy cleanup.

8. Running on Windows — same commands, same output.

---

❓ FAQ

Does this delete anything automatically? No. The delete command always asks for confirmation first. list and summary only read — they never change anything.

Will it find models in custom folders? The tool scans the default locations each AI tool uses (Ollama, LM Studio, Hugging Face, GPT4All, Jan, and more). For most users, that's everything — no setup needed.

Is it safe? The tool only deletes what you tell it to, after confirmation. Models that are currently running (loaded by Ollama or Foundry) are protected — the tool won't let you delete a model that's in use.

Does it need admin rights? No. It runs as your normal user account.

Why are some models shown as Broken? That usually means a download didn't finish (common with large Hugging Face models). The weights are incomplete — safe to delete to reclaim space.

What does Status mean?

• Idle — model is on disk but not currently loaded
• InUse — model is currently running (protected from deletion)
• Broken — download was interrupted; incomplete and unusable
• Orphaned — cache entry with no manifest

---

💡 Tips

• Start with summary to see where your space is going, then use list to dig in.
• Sort by date (list --sort date) to find models you haven't touched in months.
• Filter by engine before deleting — safer than deleting across all tools at once.
• Broken = free space — any model showing Broken is reclaimable disk space from a failed download.

---

👨‍💻 Author

Venu Thomas — from Kerala, India 🇮🇳

Copyright © 2026 Venu Thomas. All rights reserved.