avifenesh/memra

Inference Engines

Rust + CUDA inference engine for NVIDIA RTX PRO 6000 Blackwell and RTX 5090. Serves safetensors and GGUF over an OpenAI-compatible API, with per-device tuned defaults and speculative decode gated byte-identical to plain decode. Hosted instance: inference.tiyuvta.ai

GitHub →Website →

6.2

GitHub Metrics

Stars

327

Forks

Open Issues

Watchers

Contributors

Weekly Commits

250

Language

Rust

License

MIT

Last Commit

Aug 28, 2026

Created

Jul 5, 2026

Latest Release

v0.117.0

Release Date

Aug 28, 2026

Synced: Aug 28, 2026

Quality Scores

Documentation Qualityw: 20%

6.1

Has docs site (https://inference.tiyuvta.ai). Description: 265 chars. Stars signal: 327. Contributors: 7. Score: 6.1/10

Community Healthw: 20%

4.6

Stars: 327. Contributors: 7. Watchers: 25. Forks: 38. Issue ratio: 0.6%. Score: 4.6/10

Maintenance Velocityw: 15%

9.7

Last commit: 0d ago. Weekly commits: 250. Latest release: v0.117.0. Score: 9.7/10

API Design & DXw: 20%

7.9

Stars/issues ratio: 164. Typed language: Rust. Has documentation site. Permissive license: MIT. Popularity signal: 327 stars. Score: 7.9/10

Production Readinessw: 15%

3.5

Battle-tested: 327 stars. Peer review: 7 contributors. Versioned: v0.117.0. Licensed: MIT. Age: 0.1 years. Maintenance: last commit 0d ago. Score: 3.5/10

Ecosystem Integrationw: 10%

5.3

Fork interest: 38. Major ecosystem: Rust. Integration-friendly: MIT. Adoption: 327 stars. Has web presence. Score: 5.3/10