ModelScout (modelscout-llm)
Find local AI models that fit your hardware, with accurate memory and generation speed estimates.
Downloading a 10GB–50GB model file only to discover that it overflows your VRAM or crawls at 1 token per second is a frustrating experience. ModelScout inspects your machine and tells you exactly what open-source models will run smoothly — before you spend hours downloading weights.
Whether you're running on Apple Silicon (M1–M4), NVIDIA RTX, AMD Radeon / ROCm, Intel Arc, or CPU-only, ModelScout calculates real-world weights, KV cache requirements, and memory-bandwidth token generation speeds.
⚡ Quick Start with uv (Zero Installation)
You don't even need to clone a repository or set up a virtual environment. If you have uv installed, you can run ModelScout instantly via uvx:
# Scan your hardware and get the best matching models
uvx modelscout-llm@latest
# Launch the interactive local Web Dashboard (http://localhost:1234)
uvx modelscout-llm@latest web
# Inspect your detected GPU/CPU specs and AI capability score
uvx modelscout-llm@latest hardware
📦 Install via pip
You can also install ModelScout globally or into any Python 3.11+ environment:
pip install modelscout-llm
Once installed, the modelscout command is available everywhere:
# Run the scanner
modelscout
# Launch the Web UI
modelscout web
# Target a specific workload
modelscout --profile coding
modelscout --profile reasoning
modelscout --profile vision
# Plan hardware requirements for a model
modelscout plan "llama 3 70b"
# Generate ready-to-run Python code snippet
modelscout snippet "llama 3" --runner ollama
# Start an interactive chat session
modelscout run "llama-3.2-1b"
🛠️ Key Capabilities
- Native Hardware Probing: Automatically detects Apple Silicon unified memory & bandwidth, NVIDIA NVML/CUDA, AMD ROCm, Intel Arc, and CPU AVX/NEON instruction sets.
- Architecture-Aware Memory Engine: Accurately models weights + GQA/MQA KV cache footprints + activation buffers + framework overhead.
- Bandwidth-Bound Speed Estimates: Derives honest tokens/second ranges based on your system's actual memory bus bandwidth (GB/s).
- Interactive Web Interface: Complete browser dashboard running locally on port
1234, featuring hardware autocompletion, real-time filtering, and side-by-side model comparison. - Hardware Simulation: Test potential upgrades before buying hardware (e.g.
modelscout --gpu "2x RTX 4090"ormodelscout upgrade). - Flexible Formats: Export clean GitHub-flavored Markdown (
-m) or strict JSON (--json) for scripting and automation.
🔗 Links
- GitHub Repository: https://github.com/jagan-jijo/modelscout
- Author: Jagan Jijo (Portfolio)
- License: MIT
1 maintainer
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file modelscout_llm-0.1.1.tar.gz.
File metadata
- Download URL: modelscout_llm-0.1.1.tar.gz
- Upload date:
- Size: 178.9 kB
- Tags: Source
- Uploaded using Trusted Publishing? Yes
- Uploaded via:
twine/7.0.0 CPython/3.13.14
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
dfc3bc263d98233a4c75744d3774c20e8a31ed4e0c8e2bd13467e2c92c3c9b9e
|
|
| MD5 |
5707f9a13eb6ca2b22160e7f665045e5
|
|
| BLAKE2b-256 |
7d91617382fe02bff8254bd4697d618f082a9dfe6e1128a5d044bac09c83f21c
|
Provenance
The following attestation bundles were made for modelscout_llm-0.1.1.tar.gz:
Publisher:
publish.yml on jagan-jijo/modelscout
-
Statement:
-
Statement type:
https://in-toto.io/Statement/v1 -
Predicate type:
https://docs.pypi.org/attestations/publish/v1 -
Subject name:
modelscout_llm-0.1.1.tar.gz -
Subject digest:
dfc3bc263d98233a4c75744d3774c20e8a31ed4e0c8e2bd13467e2c92c3c9b9e - Sigstore transparency entry: 2799996515
- Sigstore integration time:
-
Permalink:
jagan-jijo/modelscout@5b4ab448922569bf1111202a548ee6725ae268f2 -
Branch / Tag:
refs/heads/main - Owner: https://github.com/jagan-jijo
-
Access:
public
-
Token Issuer:
https://token.actions.githubusercontent.com -
Runner Environment:
github-hosted -
Publication workflow:
publish.yml@5b4ab448922569bf1111202a548ee6725ae268f2 -
Trigger Event:
workflow_dispatch
-
Statement type:
File details
Details for the file modelscout_llm-0.1.1-py3-none-any.whl.
File metadata
- Download URL: modelscout_llm-0.1.1-py3-none-any.whl
- Upload date:
- Size: 237.1 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? Yes
- Uploaded via:
twine/7.0.0 CPython/3.13.14
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
8dac3b779a7b44c59d400df8f04ca89f9ea24ab36c5dd24b87ad54709fd449a6
|
|
| MD5 |
32faf374dec9706ec25cb2c64a84a6ef
|
|
| BLAKE2b-256 |
a15f9a861b6830b924312dbcd633e947ccb0ea828e22a28738b7bb4608756aad
|
Provenance
The following attestation bundles were made for modelscout_llm-0.1.1-py3-none-any.whl:
Publisher:
publish.yml on jagan-jijo/modelscout
-
Statement:
-
Statement type:
https://in-toto.io/Statement/v1 -
Predicate type:
https://docs.pypi.org/attestations/publish/v1 -
Subject name:
modelscout_llm-0.1.1-py3-none-any.whl -
Subject digest:
8dac3b779a7b44c59d400df8f04ca89f9ea24ab36c5dd24b87ad54709fd449a6 - Sigstore transparency entry: 2799996570
- Sigstore integration time:
-
Permalink:
jagan-jijo/modelscout@5b4ab448922569bf1111202a548ee6725ae268f2 -
Branch / Tag:
refs/heads/main - Owner: https://github.com/jagan-jijo
-
Access:
public
-
Token Issuer:
https://token.actions.githubusercontent.com -
Runner Environment:
github-hosted -
Publication workflow:
publish.yml@5b4ab448922569bf1111202a548ee6725ae268f2 -
Trigger Event:
workflow_dispatch
-
Statement type:
