# mostlygeek/llama-swap

Reliable model swapping for any local OpenAI/Anthropic compatible server - llama.cpp, vllm, etc

Repository: https://github.com/mostlygeek/llama-swap
Canonical: https://ross.abutalabs.com/products/llama-swap
Language: Go
License: MIT
License Family: permissive
Topics: golang, llama, llamacpp, localllama, localllm, openai, openai-api, vllm
Last push: 2026-08-26T14:40:01+00:00

## Health v2 (maintenance only)
Score: 85/100 (v2, computed 2026-09-03T02:39:23.370411+00:00)
- activity 99, release rhythm 87, longevity 49
- inputs: {"age_days": 698, "days_push": 7, "days_rel": 10, "gap_med": 2.0, "n_releases_24m": 193}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 5480, forks 433 (observed 2026-08-28T04:09:19.676934+00:00)

## What it is
A Go-based proxy service that lets you run multiple local AI inference servers (llama.cpp, vllm, ComfyUI, etc.) and hot-swap between models on demand behind a single OpenAI/Anthropic-compatible API endpoint. It ships as a single dependency-free binary configured by one file, with a built-in web UI for managing running models.

## Use cases
- run multiple local LLMs and switch between them on demand
- expose one OpenAI-compatible API for several local models
- automatically unload and load models to fit limited GPU memory
- proxy requests to llama.cpp or vllm behind a unified endpoint
- serve chat, embeddings, image, and audio models from one local server
- manage local AI model servers with a web UI

## When to choose
- you run several local models on limited hardware and need automatic swapping
- you want a single OpenAI/Anthropic-compatible endpoint for heterogeneous inference servers
- you want a zero-dependency, single-binary local AI gateway
- you use llama.cpp, vllm, ComfyUI, or stable-diffusion.cpp together

## When to avoid
- you need multi-model concurrent serving on abundant GPU memory rather than swapping
- you run cloud-hosted inference and don't need local model management
- you need a production-scale load balancer across many replicas

## Facets
- artifact type: service
- maturity: active
- function: llm-inference, http-server, proxy, api-framework
- domain: large-language-models, artificial-intelligence, self-hosted, developer-tools
- platform: windows, go, self-hosted
- tags: model-swapping, openai-api, anthropic-api, llama-cpp, vllm, local-llm, reverse-proxy, single-binary, linux, macos, docker

## Member repositories
- mostlygeek/llama-swap (main) score 85

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:09:19.676934+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 2026-08-29T17:56:23.254921+00:00, confidence not recorded.
  - readme: https://github.com/mostlygeek/llama-swap (fetched 2026-08-28T04:09:19.676934+00:00, sha a2bc405ccb0b)
- Data as of 2026-08-30T08:39:29.467469+00:00.