local-vault

Name: local-vault
Author: genli-ai

25installs5forks

How to install local-vault?

genli-ai/market-research-skills/local-vault

$npx -y skills add genli-ai/market-research-skills --skill local-vault

Installs into the current project.

Summary

Build and query a local Markdown knowledge base ("vault"). TWO functions — (1) CONVERT raw files (PDF, Word/docx, PowerPoint/pptx, Excel/xlsx, csv/tsv, images, html, md/txt, json/yaml/code, audio/video) into clean Markdown with retrieval-friendly frontmatter; local-first (pandoc

Files · 1

View on GitHub

SKILL.md

1	# local-vault
2
3	Turn a folder of raw files into a Markdown vault that an LLM can grep, and
4	then answer questions over that vault responsibly.
5
6	Mental model: `SOURCE` = raw files (source of truth). `VAULT` = one `.md` per
7	source file, carrying retrieval frontmatter (abstract / tags / synonyms) + a
8	`source` backlink. The vault is the layer the LLM reads; the raw files are where
9	the user goes to verify.
10
11	There are two distinct jobs — figure out which the user wants:
12
13	- A. Convert / sync — they dropped files in and want them in the vault → run
14	the pipeline (`scripts/sync.py`).
15	- B. Retrieve / answer — they want answers from an existing vault → follow
16	the Retrieval & feedback protocol below. Do not run the pipeline for this.
17
18	---
19
20	## A. Convert / sync
21
22	### One-time setup (do this for the user if not already done)
23
24	1. Python deps (user-level, no venv):
25	```
26	python3 -m pip install --user requests python-dotenv pypdf pymupdf4llm openpyxl python-pptx
27	```
28	2. pandoc (for docx/rtf/odt/epub): `brew install pandoc` (macOS) / distro pkg.
29	3. ffmpeg (only for audio/video transcription): `brew install ffmpeg` (macOS) /
30	distro pkg. The whisper engine is auto-selected by platform — `mlx-whisper`
31	on Apple Silicon (GPU), `faster-whisper` elsewhere (cross-platform CPU/CUDA) — and
32	auto-installed after the user consents at the first-run prompt (no manual pip
33	needed). On that first run with audio/video present, the tool shows the model-size
34	options (tiny ~75 MB / small ~480 MB / turbo ~1.6 GB / large-v3 ~3 GB) and lets the
35	user pick or skip; the choice is saved to `.env` (`KB_WHISPER_MODEL`) so it never
36	re-asks. Fully local — no token/quota; the model downloads once, then offline.
37	4. `claude` CLI on PATH — the pipeline shells out to `claude -p` for frontmatter
38	enrichment and PPT-image OCR. If absent, those steps are skipped (not fatal).
39	5. Configure paths — two ways:
40	- Guided (recommended for the user): just run `python3 scripts/sync.py` in a
41	terminal. On first run (when paths aren't configured yet) it launches an
42	interactive wizard: it asks for the raw-files folder + the vault folder
43	(+ optional MinerU token), creates them, writes `scripts/.env`, and prints how
44	to use the tool. Then they re-run to convert.
45	- Manual: copy `scripts/.env.example` → `scripts/.env` and set
46	`KB_SOURCE_DIR` (raw files) and `KB_TARGET_DIR` (the Markdown vault), both
47	absolute. `MINERU_TOKEN` is optional (only for legacy .doc/.ppt, .html, scanned
48	PDFs, images — get one at https://mineru.net).
49	- When you (Claude) run the setup for the user, prefer the manual path: ask
50	them for the two folders, then write `scripts/.env` directly (the wizard only
51	fires on an interactive TTY, which a `claude -p` subprocess is not).
52
53	### Run it
54
55	```
56	python3 scripts/sync.py
57	```
58
59	On macOS, the first run (wizard or any normal run) also drops a clickable
60	`sync.command` into the knowledge-base root — the parent of the SOURCE folder,
61	with the absolute path to `sync.py` baked in (tool and data live apart — under
62	`/plugin install` the script sits in `~/.claude/plugins/cache/…`, far from the data
63	folders, so a relative launcher can't work). After that the daily loop is: drop
64	files into SOURCE → double-click `sync.command` → read the `.md` in VAULT. The
65	launcher is idempotent; a stale auto-generated copy left in the SOURCE folder by an
66	older version is removed automatically (a user-written one is never touched). If a
67	different `sync.command` already exists at the root, an interactive terminal
68	prompts update / skip; non-interactively, our own out-of-date launcher self-heals
69	silently while a user-customized one is left alone.
70
71	First terminal run with no config → the setup wizard (above). Once `.env` exists:
72
73	- Incremental: only files in SOURCE without a matching `.md` in VAULT are
74	processed. To force a re-convert, delete that `.md` first, then re-run.
75	- No MinerU token needed for the local paths (xlsx/csv/docx/pptx/md/txt/code +
76	digital PDF). Token is validated lazily, only

People who liked this also viewed

6 picks

Type

local-vault

25installs5forks

View on GitHub

How to install local-vault?

genli-ai/market-research-skills/local-vault

$npx -y skills add genli-ai/market-research-skills --skill local-vault

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# local-vault
2
3	Turn a folder of raw files into a Markdown vault that an LLM can grep, and
4	then answer questions over that vault responsibly.
5
6	Mental model: `SOURCE` = raw files (source of truth). `VAULT` = one `.md` per
7	source file, carrying retrieval frontmatter (abstract / tags / synonyms) + a
8	`source` backlink. The vault is the layer the LLM reads; the raw files are where
9	the user goes to verify.
10
11	There are two distinct jobs — figure out which the user wants:
12
13	- A. Convert / sync — they dropped files in and want them in the vault → run
14	the pipeline (`scripts/sync.py`).
15	- B. Retrieve / answer — they want answers from an existing vault → follow
16	the Retrieval & feedback protocol below. Do not run the pipeline for this.
17
18	---
19
20	## A. Convert / sync
21
22	### One-time setup (do this for the user if not already done)
23
24	1. Python deps (user-level, no venv):
25	```
26	python3 -m pip install --user requests python-dotenv pypdf pymupdf4llm openpyxl python-pptx
27	```
28	2. pandoc (for docx/rtf/odt/epub): `brew install pandoc` (macOS) / distro pkg.
29	3. ffmpeg (only for audio/video transcription): `brew install ffmpeg` (macOS) /
30	distro pkg. The whisper engine is auto-selected by platform — `mlx-whisper`
31	on Apple Silicon (GPU), `faster-whisper` elsewhere (cross-platform CPU/CUDA) — and
32	auto-installed after the user consents at the first-run prompt (no manual pip
33	needed). On that first run with audio/video present, the tool shows the model-size
34	options (tiny ~75 MB / small ~480 MB / turbo ~1.6 GB / large-v3 ~3 GB) and lets the
35	user pick or skip; the choice is saved to `.env` (`KB_WHISPER_MODEL`) so it never
36	re-asks. Fully local — no token/quota; the model downloads once, then offline.
37	4. `claude` CLI on PATH — the pipeline shells out to `claude -p` for frontmatter
38	enrichment and PPT-image OCR. If absent, those steps are skipped (not fatal).
39	5. Configure paths — two ways:
40	- Guided (recommended for the user): just run `python3 scripts/sync.py` in a
41	terminal. On first run (when paths aren't configured yet) it launches an
42	interactive wizard: it asks for the raw-files folder + the vault folder
43	(+ optional MinerU token), creates them, writes `scripts/.env`, and prints how
44	to use the tool. Then they re-run to convert.
45	- Manual: copy `scripts/.env.example` → `scripts/.env` and set
46	`KB_SOURCE_DIR` (raw files) and `KB_TARGET_DIR` (the Markdown vault), both
47	absolute. `MINERU_TOKEN` is optional (only for legacy .doc/.ppt, .html, scanned
48	PDFs, images — get one at https://mineru.net).
49	- When you (Claude) run the setup for the user, prefer the manual path: ask
50	them for the two folders, then write `scripts/.env` directly (the wizard only
51	fires on an interactive TTY, which a `claude -p` subprocess is not).
52
53	### Run it
54
55	```
56	python3 scripts/sync.py
57	```
58
59	On macOS, the first run (wizard or any normal run) also drops a clickable
60	`sync.command` into the knowledge-base root — the parent of the SOURCE folder,
61	with the absolute path to `sync.py` baked in (tool and data live apart — under
62	`/plugin install` the script sits in `~/.claude/plugins/cache/…`, far from the data
63	folders, so a relative launcher can't work). After that the daily loop is: drop
64	files into SOURCE → double-click `sync.command` → read the `.md` in VAULT. The
65	launcher is idempotent; a stale auto-generated copy left in the SOURCE folder by an
66	older version is removed automatically (a user-written one is never touched). If a
67	different `sync.command` already exists at the root, an interactive terminal
68	prompts update / skip; non-interactively, our own out-of-date launcher self-heals
69	silently while a user-customized one is left alone.
70
71	First terminal run with no config → the setup wizard (above). Once `.env` exists:
72
73	- Incremental: only files in SOURCE without a matching `.md` in VAULT are
74	processed. To force a re-convert, delete that `.md` first, then re-run.
75	- No MinerU token needed for the local paths (xlsx/csv/docx/pptx/md/txt/code +
76	digital PDF). Token is validated lazily, only

People who liked this also viewed

6 picks

Type