Local llm serving setup — Atom
Local LLM serving setup audit: a deliverable delivered per API call as an independent atom service.
التثبيت
أضف هذا إلى إعدادات MCP في عميلك (Claude Desktop: claude_desktop_config.json، وCursor: .cursor/mcp.json، ويستخدم VS Code وغيره الصيغة نفسها «mcpServers»).
claude mcp add --transport http local-llm-serving-setup-atom 'https://atoms.pipavia.com/mcp/local-llm-serving-setup' --header 'Authorization: <This atom'\''s own API key, as '\''Bearer <key>'\''. Keys are per user, per atom (namespace local_llm_serving_setup); issue yours from the atom'\''s page after sign-up.>'
{
"mcpServers": {
"local-llm-serving-setup-atom": {
"type": "http",
"url": "https://atoms.pipavia.com/mcp/local-llm-serving-setup",
"headers": {
"Authorization": "<This atom's own API key, as 'Bearer <key>'. Keys are per user, per atom (namespace local_llm_serving_setup); issue yours from the atom's page after sign-up.>"
}
}
}
}
التوثيق
مأخوذ من ملف README للمشروع على GitHub ويُحدَّث أسبوعيًا. تفتح الروابط على GitHub.
atom-services
Conceived by Nrupal Akolkar · Built by Muse (Arjun/Wright).
One collective repo for all product atoms. An atom is a small, independent service that delivers one bounded judgment artifact — a VERDICT, a CHECK, a MAP, a STAMP, a REPORT — over API / MCP / agent / human endpoints, paid per unit delivered. Atoms never sell the underlying knowledge; they sell the artifact the knowledge produces.
The model in plain words
- Independent atoms, one home. Every atom is its own named service
unit in
services/<service-name>/, with its own endpoints, its own payment plan, and its own growth / retention / maintenance docs. Atoms are never merged into existing product repos — this repo is their only home. Shared rails live once at the repo root (rails/), not inside any product. - The rail is Conduit. Atoms are listed and discovered through
Conduit (the owner's connector marketplace, live since 2026-11-12)
and served as key-gated MCP/API endpoints in Conduit's shape. Where
Conduit lacks a piece atoms need, the gap is named in
rails/and inBUILD-LOG.md— never papered over. - One by one. Atoms release one by one (or bundle by bundle), in
the approved queue order — never the whole register at once.
Release #1 is
release-readiness-verdict. - Volume pricing doctrine. Cheap per unit, millions of units. The earning lever is refinement — narrower, sharper services get called more — not price height. Every price must clear the atom's measured production cost; until cost is measured, atoms carry a pricing unit only and no price.
- One Stripe account (2351485 ALBERTA INC., public name AIMLDS) settles everything. Settlement is aggregated — prepaid balance / packs drawn down per unit — never a card charge per call.
Layout
registry/ The authoritative register (register.yaml + ATOMS-REGISTER.md,
copied unchanged) + PUBLIC-LIST.md (generated).
rails/ serving.md · billing.md · docs-standard.md · contracts.md —
how every atom in this repo is served, metered, billed,
documented, and contracted. SPEC marks what is not built yet.
services/ One directory per PUBLIC atom (162), generated from the
register by tools/generate.py. Each unit: SERVICE.md,
contract.yaml, growth.md, retention.md, maintenance.md.
tools/ generate.py — regenerates registry/PUBLIC-LIST.md and all
service scaffolds from registry/register.yaml.
BUILD-LOG.md What was built, verified counts, test evidence, named gaps.
CHARTER.md The binding rules of this repo (read before changing anything).
The public cut
255 candidates were screened. Public = Tier A (48) + Tier B (114) =
162. Tier C (34, personal/internal), Tier D (40, held) and
FAIL-self-use (19, not services) get no service units — they exist
in registry/ only, marked non-public with reasons
(registry/PUBLIC-LIST.md).
Status
Scaffold + atom #1's deterministic core are built and tested locally.
Nothing in this repo is deployed or live. Atom #1's remaining road
to live on Conduit is listed in BUILD-LOG.md.
الأسئلة الشائعة
ما هو Local llm serving setup — Atom؟
Local LLM serving setup audit: a deliverable delivered per API call as an independent atom service.
كيف أثبّت Local llm serving setup — Atom؟
يعمل كخادم بعيد. في Claude Code: claude mcp add --transport http local-llm-serving-setup-atom 'https://atoms.pipavia.com/mcp/local-llm-serving-setup' --header 'Authorization: <This atom'\''s own API key, as '\''Bearer <key>'\''. Keys are per user, per atom (namespace local_llm_serving_setup); issue yours from the atom'\''s page after sign-up.>'. وفي العملاء الآخرين، الصق إعداد JSON من هذه الصفحة في إعدادات MCP الخاصة بهم.
هل يحتاج Local llm serving setup — Atom إلى مفتاح API؟
نعم. يتطلب Authorization، وتضبطها في إعداداته.
هل Local llm serving setup — Atom مفتوح المصدر؟
شيفرته علنية على GitHub، لكن المستودع لا يعلن ترخيصًا قياسيًا.
من يصون Local llm serving setup — Atom؟
ينشره nrupala. يدرجه AI.tn لكنه لا يصونه ولا يوصي به.
من سجل MCP الرسمي. الإدراج ليس توصية: راجع شيفرة الخادم وصلاحياته قبل منحه الوصول إلى بياناتك.