hor-light-rag SKILL.md skillA
hor-light-rag from openreachtech/hora-skills-ort-renchan is an agent skill: a SKILL.md folder an AI agent loads when a task calls for it.
mdr add openreachtech/hora-skills-ort-renchan/hor-light-rag@git:20261016.5f0404cPins git:20261016.5f0404c in mdr.lock. Pin the fingerprint instead.
- Audit grade
- A, 17 of 17 checks passed
- Kind
- SKILL.md skill
- Repository
- openreachtech/hora-skills-ort-renchan
- Path
kit/skills/hor-light-rag/SKILL.md- Versions recorded
- 2
- Latest version
- git:20261016.5f0404c, committed Oct 16, 2026
- First version recorded
- Sep 5, 2026
- Size
- 13,173 bytes
- Repository stars
- 32
- License
- Apache-2.0
Add lightweight retrieval-augmented generation (RAG) for AI-agent documents in a renchan backend without a vector database — a vector-first / LLM-fallback pipeline that ranks stored Document rows with an offline local embedder, plus a MySQL/MariaDB BOOLEAN-mode n-gram fulltext builder for keyword search. Use whenever the user asks to add or tune document retrieval for an agent, generate or store embeddings, or change how background knowledge is assembled into the prompt.
Collected from public GitHub. Every version is saved exactly as it was and never changes. Install it at an exact version with the mdr tool, and every file is checked before it reaches your agent. The main file is checked against the SHA-256 fingerprint recorded here, and the others against the fingerprints GitHub keeps for its source commit. The automatic checks below grade the latest version, and the same checks always give the same file the same grade.
What the file says
# Light RAG (vector-first / LLM-fallback document retrieval, no vector DB) This backend does RAG "lightly" — no external vector database, no heavy embedding service. Documents live as `Document` rows; a **local, offline embedder** produces small vectors stored in `DocumentEmbedding`; retrieval ranks them with **in-app cosine** and falls back to an **LLM-driven id selection** when vectors find nothing. A separate n-gram fulltext builder exists for generic MySQL/MariaDB keyword search. ## Grand principle: size the retrieval to the corpus — local vectors first, LLM fallback, no vector DB For an agent's own document set (tens–hundreds of docs), a vector database is overkill. Rank a small set of **offline-precomputed local vectors** with cosine in the app; only when that yields nothing, ask the model to pick document ids from a metadata catalog. - **Why a local, dependency-free embedder**: `LocalTextEmbedder` hashes word tokens + CJK n-grams into a fixed 256-dim L2-normalized vector — no network, no API cost, works offline, and (being normalized) cosine is just a dot product. It matches surface forms well; true semantic paraphrase is its weakness, so …
Read the whole file at its exact version.
How to install
mdr add openreachtech/hora-skills-ort-renchan/hor-light-rag@sha256:e714f0279d44c8d5The line at the top of this page pins the label git:20261016.5f0404c, which is what most people want. This one pins the sha256 fingerprint of the contents instead, so it survives a label being moved. Either way the exact version is written to mdr.lock, and mdr install downloads and checks those contents again, so it installs them exactly or stops with an error.
[](https://markdownregistry.com/a/art_fr5xh4ubjntl7pwf)
3 badge image fetches in 30 days. What the badge shows.
Versions
The registry has recorded 2 versions of this file, between Sep 5, 2026 and Oct 16, 2026, about one every 40 days.
| version | committed | commit | size | audit | |
|---|---|---|---|---|---|
| git:20261016.5f0404c latest | 2026-10-16 | 5f0404c | 13,173 B | A | view · diff |
| git:20260905.5a04969 | 2026-09-05 | 5a04969 | 12,760 B | A | view |
Audit of the latest version
- pass: Frontmatter block present
- pass: Frontmatter declares a name
- pass: Frontmatter declares a description
- pass: Size between 200 bytes and 200 KB (13173 bytes)
- pass: No zero-width or bidi control characters
- pass: No instruction hidden inside an HTML comment
- pass: No link to an exfiltration or paste host
- pass: No credential-shaped string
- pass: No instruction to send local credentials anywhere
- pass: No text hidden with inline styles
- pass: No prompt-injection phrasing
- pass: No curl or wget piped into a shell
- pass: No recursive delete of root, home or parent
- pass: No instruction to read or print local credentials
- pass: No base64 blob over 200 characters
- pass: No link to a raw IP address
- pass: No script tag
Source
openreachtech/hora-skills-ort-renchan · 32 stars · license Apache-2.0 · pushed 2026-10-16 · branch main
API
GET https://markdownregistry.com/api/v1/artifacts/art_fr5xh4ubjntl7pwf GET https://markdownregistry.com/api/v1/resolve?ref=openreachtech/hora-skills-ort-renchan/hor-light-rag GET https://markdownregistry.com/api/v1/blob/e714f0279d44c8d548ae186399a489abd9057d5474816c8854445935c1921b58
Your AI assistant does the legwork. You hear about the deals worth your word. Hand yours the standing instructions at modelranch.com and it joins the network that reads files like this one.
Guides: the SKILL.md format and its required fields · what to check before you install a skill · how to pin a skill to an exact version · where to find agent skills · keeping a team on the same skills · which agent file does what.