# jamesob/local-llm

Everything I know about running LLMs locally

Repository: https://github.com/jamesob/local-llm
Canonical: https://ross.abutalabs.com/products/local-llm
Language: Shell
License Family: other
Last push: 2026-07-10T16:46:17+00:00

## Health v2 (maintenance only)
Score: 54/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 91, release rhythm 35, longevity 4
- inputs: {"age_days": 61, "days_push": 54, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, young, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1800, forks 106 (observed 2026-08-28T04:05:37.929279+00:00)

## What it is
A personal guide and configuration repository for running state-of-the-art LLMs and speech-to-text locally on consumer/prosumer hardware. It documents hardware choices (multi-GPU EPYC rig with PCIe switches), kernel/BIOS tuning, and ready-to-run Docker serving configs.

## Use cases
- build a multi-GPU rig for running large LLMs locally
- run speech-to-text models on my own hardware
- configure vLLM to serve a 594B model with tensor parallelism
- benchmark GPU peer-to-peer bandwidth and latency
- learn how to power-limit multiple GPUs on a home circuit
- set up PCIe switch bifurcation and ACS settings for GPU P2P

## When to choose
- you want battle-tested, opinionated hardware and configuration notes for a serious local LLM setup
- you need ready-to-run Docker/vLLM serving configs for very large models
- you're troubleshooting NCCL, IOMMU, or PCIe peer-to-peer issues in a multi-GPU box

## When to avoid
- you want a turnkey installer or maintained software package - this is a guide plus configs, not a tool
- you're running small models on a single consumer GPU
- you need something with a license or formal support - the repo has no license

## Facets
- artifact type: learning-resource
- maturity: active
- function: llm-inference, speech-recognition, gpu-computing, benchmarking, developer-tools
- domain: large-language-models, hardware, self-hosted, artificial-intelligence, speech-processing
- platform: self-hosted
- tags: local-llm, hardware-guide, vllm, tensor-parallelism, pcie-p2p, speech-to-text, docker-compose, shell-scripts, linux, docker, gpu

## Member repositories
- jamesob/local-llm (main) score 54

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:37.929279+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T03:22:14.663238+00:00, confidence not recorded.
  - readme: https://github.com/jamesob/local-llm (fetched 2026-08-28T04:05:37.929279+00:00, sha 21deac757315)
- Data as of 2026-08-30T08:39:29.467469+00:00.