PHPackages vielhuber/runpodhelper - PHPackages - PHPackages [Skip to content](#main-content)[PHPackages](/)[Directory](/)[Categories](/categories)[Trending](/trending)[Leaderboard](/leaderboard)[Changelog](/changelog)[Analyze](/analyze)[Collections](/collections)[Log in](/login)[Sign up](/register) 1. [Directory](/) 2. / 3. [Utility & Helpers](/categories/utility) 4. / 5. vielhuber/runpodhelper ActiveLibrary[Utility & Helpers](/categories/utility) vielhuber/runpodhelper ====================== Automates self-hosted llm inference 1.5.4(1mo ago)1121↓20%MITShellPHP >=8.1 Since Mar 12Pushed 1w agoCompare [ Source](https://github.com/vielhuber/runpodhelper)[ Packagist](https://packagist.org/packages/vielhuber/runpodhelper)[ RSS](/packages/vielhuber-runpodhelper/feed)WikiDiscussions main Synced 3w ago READMEChangelogDependencies (12)Versions (54)Used By (0) [![GitHub Tag](https://camo.githubusercontent.com/3074d5edeeed4741e213fe10941e3712a29c14e242d49377444b000556d25e6f/68747470733a2f2f696d672e736869656c64732e696f2f6769746875622f762f7461672f7669656c68756265722f72756e706f6468656c706572)](https://github.com/vielhuber/runpodhelper/tags)[![Code Style](https://camo.githubusercontent.com/1540f8ce219727155ab62506c77b818b720421c22c4cf0b18a5f160942132e2d/68747470733a2f2f696d672e736869656c64732e696f2f62616467652f636f64655f7374796c652d7073722d2d31322d6666363962342e737667)](https://www.php-fig.org/psr/psr-12/)[![License](https://camo.githubusercontent.com/21e456533aa4518421d521d559169b0219ada2f0d0f4d4936939e7c09cc65cda/68747470733a2f2f696d672e736869656c64732e696f2f6769746875622f6c6963656e73652f7669656c68756265722f72756e706f6468656c706572)](https://github.com/vielhuber/runpodhelper/blob/main/LICENSE.md)[![Last Commit](https://camo.githubusercontent.com/3ff0477ea8d75f44cf66115c0a16fb6e2c515cffe9eda7ec2dcd0095fa92285a/68747470733a2f2f696d672e736869656c64732e696f2f6769746875622f6c6173742d636f6d6d69742f7669656c68756265722f72756e706f6468656c706572)](https://github.com/vielhuber/runpodhelper/commits)[![PHP Version Support](https://camo.githubusercontent.com/a37bd357a2d387737abed18c84b3b09a64d15ad8c54124fe33b15c048479da60/68747470733a2f2f696d672e736869656c64732e696f2f7061636b61676973742f7068702d762f7669656c68756265722f72756e706f6468656c706572)](https://packagist.org/packages/vielhuber/runpodhelper)[![Packagist Downloads](https://camo.githubusercontent.com/2d71845869d523e40e13cb39a7943f271784414c84a6be55c73056e3eff14cfa/68747470733a2f2f696d672e736869656c64732e696f2f7061636b61676973742f64742f7669656c68756265722f72756e706f6468656c706572)](https://packagist.org/packages/vielhuber/runpodhelper) ⛈ runpodhelper ⛈ ================ [](#-runpodhelper-) runpodhelper automates the full lifecycle of self-hosted llm inference on runpod gpu cloud. it provisions pods via the runpod graphql api, installs lm studio or llama.cpp, downloads gguf models from huggingface, and serves them behind a cloudflare tunnel. usage ----- [](#usage) ``` ./vendor/bin/runpod.sh create --config pods.yaml ./vendor/bin/runpod.sh delete --all ./vendor/bin/runpod.sh status ./vendor/bin/runpod.sh create \ --gpu "4x RTX PRO 6000" \ --hdd 250 \ --model "unsloth/MiniMax-M2.7-GGUF-UD-Q4_K_XL" \ --image "runpod/pytorch:1.0.2-cu1281-torch280-ubuntu2404" \ --type "llamacpp" \ --api-key "your-static-api-key" \ --context-length 262144 \ --parallel 2 \ --datacenter "EUR-IS-2" \ --auto-destroy 3600 ./vendor/bin/runpod.sh delete --id 001 ./vendor/bin/runpod.sh test quality --runs 5 ./vendor/bin/runpod.sh test quantity --runs 80 ./vendor/bin/runpod.sh scale --start \ --gpu "RTX 5090" \ --hdd 50 \ --model "unsloth/Qwen3.5-27B-GGUF-UD-Q4_K_XL" \ --image "runpod/pytorch:1.0.2-cu1281-torch280-ubuntu2404" \ --type "lmstudio" \ --api-key "your-static-api-key" \ --context-length 65536 \ --parallel 2 \ --datacenter "EUR-IS-2" \ --auto-destroy 3600 \ --pod-count 3 ./vendor/bin/runpod.sh scale --start \ --gpu "L40S" \ --hdd 60 \ --model "unsloth/Qwen3.5-35B-A3B-GGUF" \ --image "runpod/pytorch:1.0.2-cu1281-torch280-ubuntu2404" \ --type "llamacpp" \ --api-key "your-static-api-key" \ --context-length 131072 \ --parallel 1 \ --pod-count 1 ./vendor/bin/runpod.sh scale --start \ --gpu "2x RTX PRO 6000" \ --hdd 250 \ --model "unsloth/MiniMax-M2.7-GGUF-UD-Q4_K_XL" \ --image "runpod/pytorch:1.0.2-cu1281-torch280-ubuntu2404" \ --type "llamacpp" \ --api-key "your-static-api-key" \ --context-length 131072 \ --parallel 1 \ --pod-count 1 ./vendor/bin/runpod.sh scale --start \ --gpu "4x RTX PRO 6000" \ --hdd 250 \ --model "unsloth/MiniMax-M2.7-GGUF-UD-Q4_K_XL" \ --image "runpod/pytorch:1.0.2-cu1281-torch280-ubuntu2404" \ --type "llamacpp" \ --api-key "your-static-api-key" \ --context-length 262144 \ --parallel 2 \ --pod-count 1 ./vendor/bin/runpod.sh scale --stop ./vendor/bin/runpod.sh scale --pod-count 20 ./vendor/bin/runpod.sh scale --refresh --context-length 65536 --parallel 2 ./vendor/bin/runpod.sh scale --refresh ``` rules ----- [](#rules) - `gpu-vram ≈ model-size + context-length * model-factor` - `token-budget-per-session ≈ parallel * context-length` - `workers-per-pod ≈ workers-count / pod-count` - `running-workers-per-pod ≈ parallel` - `concurrent-workers ≈ parallel * pod-count` - `concurrent-workers ≈ 0.2 * parallel * workers-count` - `pod-count ≈ 0.2 * workers-count` ### RTX 5090 + Qwen3.5-27B [](#rtx-5090--qwen35-27b) - `gpu-vram ≈ 32 GB` - `model-size ≈ 17.6 GB` - `model-factor ≈ 0.00022` - `=> max-context-length ≈ 65536` - `=> max-parallel ≈ 2` (at context-length 65536) ### L40S + Qwen3.5-27B [](#l40s--qwen35-27b) - `gpu-vram ≈ 48 GB` - `model-size ≈ 17.6 GB` - `model-factor ≈ 0.00022` - `=> max-context-length ≈ 138240` - `=> max-parallel ≈ 4` (at context-length 65536) ### RTX PRO 6000 + Qwen3.5-122B-A10B (MoE) [](#rtx-pro-6000--qwen35-122b-a10b-moe) - `gpu-vram ≈ 96 GB` - `model-size ≈ 66 GB` - `model-factor ≈ 0.00013` (MoE, 10B active params) - `=> max-context-length ≈ 131072` (128K model limit) - `=> max-parallel ≈ 1` (at context-length 131072, ~83 GB total) - `=> max-parallel ≈ 2` (at context-length 98304, ~92 GB total) ### RTX PRO 6000 + Qwen3.5-27B [](#rtx-pro-6000--qwen35-27b) - `gpu-vram ≈ 96 GB` - `model-size ≈ 17.6 GB` - `model-factor ≈ 0.00022` - `=> max-context-length ≈ 356352` - `=> max-parallel ≈ 10` (at context-length 65536) ### RTX PRO 6000 + gemma-4-26B-A4B (MoE) [](#rtx-pro-6000--gemma-4-26b-a4b-moe) - `gpu-vram ≈ 96 GB` - `model-size ≈ 27.9 GB` - `model-factor ≈ 0.00013` (MoE, 30 layers) - `=> max-context-length ≈ 256K (model limit)` - `=> max-parallel ≈ 8` (at context-length 65536) ### RTX PRO 6000 + gemma-4-31B [](#rtx-pro-6000--gemma-4-31b) - `gpu-vram ≈ 96 GB` - `model-size ≈ 27.5 GB` - `model-factor ≈ 0.00025` (dense, 60 layers) - `=> max-context-length ≈ 256K (model limit)` - `=> max-parallel ≈ 4` (at context-length 65536) ### RTX PRO 6000 + Qwen3.6-35B-A3B (MoE) [](#rtx-pro-6000--qwen36-35b-a3b-moe) - `gpu-vram ≈ 96 GB` - `model-size ≈ 22 GB` (UD-Q4\_K\_XL) - `model-factor ≈ 0.00013` (MoE, 3B active params) - `=> max-context-length ≈ 256K (model limit)` - `=> max-parallel ≈ 2` (at context-length 131072, ~56 GB total) - `=> max-parallel ≈ 4` (at context-length 65536, ~56 GB total) installation ------------ [](#installation) - install library - `composer require vielhuber/runpodhelper` - `./vendor/bin/runpod.sh init` - setup cloudflare - Create a domain `custom.xyz` - Profile > API Tokens > Create Token - Permissions: - `Zone / DNS / Edit` - `Zone / Single Redirect / Edit` - `Account / Cloudflare Tunnel / Edit` - Account Resources - `Include > Your account` - Zone Resource - Include / Specific zone / `custom.xyz` - Set `CLOUDFLARE_DOMAIN`/`CLOUDFLARE_API_KEY` in `.env` - Each pod gets a subdomain based on its config ID: - `001.custom.xyz` - `002.custom.xyz` - … - edit config - `vi ./.env` - `vi ./models.yaml` mcp server ---------- [](#mcp-server) ``` { "mcpServers": { "runpodhelper": { "command": "/usr/bin/php", "args": ["/path/to/project/runpodhelper/bin/mcp-server.php"] } } } ``` recommended models ------------------ [](#recommended-models) NameHDDModelContext lengthParalleltok/sNotesNVIDIA GeForce RTX 509050 GBQwen3.5-27B-GGUF-UD-Q4\_K\_XL655362~43best current MCP/tool-use baselineNVIDIA L40S50 GBQwen3.5-27B-GGUF-UD-Q4\_K\_XL655364~252x parallel slots vs. RTX 5090NVIDIA RTX PRO 600050 GBQwen3.5-27B-GGUF-UD-Q4\_K\_XL6553610~20max parallel slots, single podNVIDIA RTX PRO 600050 GBgemma-4-26B-A4B-it-GGUF-UD-Q8\_K\_XL655368~65MoE: 3.8B active params, best parallelism on 96 GBNVIDIA RTX PRO 600050 GBgemma-4-31B-it-GGUF-UD-Q6\_K\_XL655364~18dense, best reliability on 96 GBNVIDIA RTX PRO 600080 GBQwen3.5-122B-A10B-GGUF-UD-Q4\_K\_XL1310721~?MoE: 10B active params, #1 intelligence indexNVIDIA RTX PRO 600050 GBQwen3.6-35B-A3B-GGUF-UD-Q4\_K\_XL1310722~?MoE: 3B active params, high throughput at 128K ctxNVIDIA A4050 GBQwen3.5-27B-GGUF-UD-Q4\_K\_XL655362~20discontinued/unavailable as of 2026-03manual deployment ----------------- [](#manual-deployment) - > Pods > Deploy - Pod template > Edit - Expose HTTP ports (comma separated): `1234` - Container Disk: `100 GB` - Copy: SSH over exposed TCP - `ssh root@xxxxxxxxxx -p xxxxx` ``` curl -fsSL https://lmstudio.ai/install.sh | bash export PATH="/root/.lmstudio/bin:$PATH" # this is unreliable #lms get -y qwen/qwen3-coder-next mkdir -p ~/.lmstudio/models/unsloth/MiniMax-M2.1-GGUF cd ~/.lmstudio/models/unsloth/MiniMax-M2.1-GGUF wget -c https://huggingface.co/unsloth/MiniMax-M2.1-GGUF/resolve/main/MiniMax-M2.1-UD-TQ1_0.gguf mkdir -p ~/.lmstudio/models/lmstudio-community/Qwen3.5-35B-A3B-GGUF cd ~/.lmstudio/models/lmstudio-community/Qwen3.5-35B-A3B-GGUF wget -c https://huggingface.co/lmstudio-community/Qwen3.5-35B-A3B-GGUF/resolve/main/Qwen3.5-35B-A3B-Q4_K_M.gguf lms server start --port 1234 --bind 0.0.0.0 ``` alternative: use runpodctl -------------------------- [](#alternative-use-runpodctl) - `ssh-keygen -t ed25519 -C "name@tld.com"` - `wget https://github.com/Run-Pod/runpodctl/releases/download/v1.14.3/runpodctl-linux-amd64 -O runpodctl` - `chmod +x runpodctl` - `mv runpodctl /usr/bin/runpodctl` - `runpodctl config --apiKey ` - `runpodctl version` more commands ------------- [](#more-commands) - `curl http://localhost:1234/v1/models` - `lms --help` - `lms status` - `lms server stop` - Copy: HTTP services > URL ``` curl https://xxxxxxxxx-1234.proxy.runpod.net/v1/responses \ -X POST \ -H "Content-Type: application/json" \ -H "Authorization: Bearer your-static-api-key" \ -d '{ "model": "xxxxxxxxxxxxx", "messages": [ {"role": "user", "content": [{"type": "input_text", "text": "hi"}]} ], "temperature": 1.0, "stream": true }' ``` ### Health Score 46 — FairBetter than 92% of packages Maintenance96 Actively maintained with recent releases Popularity14 Limited adoption so far Community6 Small or concentrated contributor base Maturity56 Maturing project, gaining track record Bus Factor1 Top contributor holds 100% of commits — single point of failure How is this calculated?**Maintenance (25%)** — Last commit recency, latest release date, and issue-to-star ratio. Uses a 2-year decay window. **Popularity (30%)** — Total and monthly downloads, GitHub stars, and forks. Logarithmic scaling prevents top-heavy scores. **Community (15%)** — Contributors, dependents, forks, watchers, and maintainers. Measures real ecosystem engagement. **Maturity (30%)** — Project age, version count, PHP version support, and release stability. ### Release Activity Cadence Every ~1 days Total 53 Last Release 35d ago ### Community Maintainers ![](https://avatars.githubusercontent.com/u/3183737?v=4)[David Vielhuber](/maintainers/vielhuber)[@vielhuber](https://github.com/vielhuber) --- Top Contributors [![vielhuber](https://avatars.githubusercontent.com/u/3183737?v=4)](https://github.com/vielhuber "vielhuber (58 commits)") ### Embed Badge ![Health badge](/badges/vielhuber-runpodhelper/health.svg) ``` [![Health](https://phpackages.com/badges/vielhuber-runpodhelper/health.svg)](https://phpackages.com/packages/vielhuber-runpodhelper) ``` ### Alternatives [laravel/framework The Laravel Framework. 34.8k532.1M19.4k](/packages/laravel-framework)[tempest/framework The PHP framework that gets out of your way. 2.2k31.1k12](/packages/tempest-framework)[matomo/matomo Matomo is the leading Free/Libre open analytics platform 21.6k38.2k](/packages/matomo-matomo)[pressbooks/pressbooks Pressbooks is an open source book publishing tool built on a WordPress multisite platform. Pressbooks outputs books in multiple formats, including PDF, EPUB, web, and a variety of XML flavours, using a theming/templating system, driven by CSS. 45344.0k1](/packages/pressbooks-pressbooks)[aedart/athenaeum Athenaeum is a mono repository; a collection of various PHP packages 245.2k](/packages/aedart-athenaeum)[lion/bundle Lion-framework configuration and initialization package 122.3k3](/packages/lion-bundle) PHPackages © 2026 [Directory](/)[Categories](/categories)[Trending](/trending)[Changelog](/changelog)[Analyze](/analyze)