ocr-and-documents

Name: ocr-and-documents
Author: adityahimaone

72stars5forks

How to install ocr-and-documents?

adityahimaone/hermes-agent-rtk-caveman/ocr-and-documents

$npx -y skills add adityahimaone/hermes-agent-rtk-caveman --skill ocr-and-documents

Installs into the current project.

Summary

Extract text from PDFs and scanned documents. Use web_extract for remote URLs, pymupdf for local text-based PDFs, marker-pdf for OCR/scanned docs. For DOCX use python-docx, for PPTX see the powerpoint skill.

Files · 1

View on GitHub

SKILL.md

1	# PDF & Document Extraction
2
3	For DOCX: use `python-docx` (parses actual document structure, far better than OCR).
4	For PPTX: see the `powerpoint` skill (uses `python-pptx` with full slide/notes support).
5	This skill covers PDFs and scanned documents.
6
7	## Step 1: Remote URL Available?
8
9	If the document has a URL, always try `web_extract` first:
10
11	```
12	web_extract(urls=["https://arxiv.org/pdf/2402.03300"])
13	web_extract(urls=["https://example.com/report.pdf"])
14	```
15
16	This handles PDF-to-markdown conversion via Firecrawl with no local dependencies.
17
18	Only use local extraction when: the file is local, web_extract fails, or you need batch processing.
19
20	## Step 2: Choose Local Extractor
21
22	\| Feature \| pymupdf (~25MB) \| marker-pdf (~3-5GB) \|
23	\|---------\|-----------------\|---------------------\|
24	\| Text-based PDF \| ✅ \| ✅ \|
25	\| Scanned PDF (OCR) \| ❌ \| ✅ (90+ languages) \|
26	\| Tables \| ✅ (basic) \| ✅ (high accuracy) \|
27	\| Equations / LaTeX \| ❌ \| ✅ \|
28	\| Code blocks \| ❌ \| ✅ \|
29	\| Forms \| ❌ \| ✅ \|
30	\| Headers/footers removal \| ❌ \| ✅ \|
31	\| Reading order detection \| ❌ \| ✅ \|
32	\| Images extraction \| ✅ (embedded) \| ✅ (with context) \|
33	\| Images → text (OCR) \| ❌ \| ✅ \|
34	\| EPUB \| ✅ \| ✅ \|
35	\| Markdown output \| ✅ (via pymupdf4llm) \| ✅ (native, higher quality) \|
36	\| Install size \| ~25MB \| ~3-5GB (PyTorch + models) \|
37	\| Speed \| Instant \| ~1-14s/page (CPU), ~0.2s/page (GPU) \|
38
39	Decision: Use pymupdf unless you need OCR, equations, forms, or complex layout analysis.
40
41	If the user needs marker capabilities but the system lacks ~5GB free disk:
42	> "This document needs OCR/advanced extraction (marker-pdf), which requires ~5GB for PyTorch and models. Your system has [X]GB free. Options: free up space, provide a URL so I can use web_extract, or I can try pymupdf which works for text-based PDFs but not scanned documents or equations."
43
44	---
45
46	## pymupdf (lightweight)
47
48	```bash
49	pip install pymupdf pymupdf4llm
50	```
51
52	Via helper script:
53	```bash
54	python scripts/extract_pymupdf.py document.pdf # Plain text
55	python scripts/extract_pymupdf.py document.pdf --markdown # Markdown
56	python scripts/extract_pymupdf.py document.pdf --tables # Tables
57	python scripts/extract_pymupdf.py document.pdf --images out/ # Extract images
58	python scripts/extract_pymupdf.py document.pdf --metadata # Title, author, pages
59	python scripts/extract_pymupdf.py document.pdf --pages 0-4 # Specific pages
60	```
61
62	Inline:
63	```bash
64	python3 -c "
65	import pymupdf
66	doc = pymupdf.open('document.pdf')
67	for page in doc:
68	print(page.get_text())
69	"
70	```
71
72	---
73
74	## marker-pdf (high-quality OCR)
75
76	```bash
77	# Check disk space first
78	python scripts/extract_marker.py --check
79
80	pip install marker-pdf
81	```
82
83	Via helper script:
84	```bash
85	python scripts/extract_marker.py document.pdf # Markdown
86	python scripts/extract_marker.py document.pdf --json # JSON with metadata
87	python scripts/extract_marker.py document.pdf --output_dir out/ # Save images
88	python scripts/extract_marker.py scanned.pdf # Scanned PDF (OCR)
89	python scripts/extract_marker.py document.pdf --use_llm # LLM-boosted accuracy
90	```
91
92	CLI (installed with marker-pdf):
93	```bash
94	marker_single document.pdf --output_dir ./output
95	marker /path/to/folder --workers 4 # Batch
96	```
97
98	---
99
100	## Arxiv Papers
101
102	```
103	# Abstract only (fast)
104	web_extract(urls=["https://arxiv.org/abs/2402.03300"])
105
106	# Full paper
107	web_extract(urls=["https://arxiv.org/pdf/2402.03300"])
108
109	# Search
110	web_search(query="arxiv GRPO reinforcement learning 2026")
111	```
112
113	## Split, Merge & Search
114
115	pymupdf handles these natively — use `execute_code` or inline Python:
116
117	```python
118	# Split: extract pages 1-5 to a new PDF
119	import pymupdf
120	doc = pymupdf.open("report.pdf")
121	new = pymupdf.open()
122	for i in range(5):
123	new.insert_pdf(doc, from_page=i, to_page=i)
124	new.save("pages_1-5.pdf")
125	```
126
127	```python
128	# Merge multiple PDFs
129	import pymupdf
130	result = pymupdf.open()
131	for path in ["a.pdf", "b.pdf", "c.pdf"]:
132	result.insert_pdf(pymupdf.open(path))
133	result.save("merged.pdf")
134	```
135
136	```python
137	# Search for text across all pages
138	import pymupdf
139	doc = pymupdf.open("report.pdf")
140	for i, page in enumerate(doc):
141	results = page.search_for("revenue")
142	if results:
143	print(f"Page {i+1}: {len(results)} match(es)")
144	print(page.get_text("text"))
145	```
146
147	No extra dependencies needed — pymupdf covers split, merge, search, and text extraction in one package.
148
149	---
150
151	## Notes
152
153	- `web_extract` is always first choice for URLs
154	- pymupdf is the safe default — instant, no models, works everywhere
155	- marker-pdf is for OCR, scanned docs,

People who liked this also viewed

6 picks

Type

ocr-and-documents

72stars5forks

View on GitHub

How to install ocr-and-documents?

adityahimaone/hermes-agent-rtk-caveman/ocr-and-documents

$npx -y skills add adityahimaone/hermes-agent-rtk-caveman --skill ocr-and-documents

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# PDF & Document Extraction
2
3	For DOCX: use `python-docx` (parses actual document structure, far better than OCR).
4	For PPTX: see the `powerpoint` skill (uses `python-pptx` with full slide/notes support).
5	This skill covers PDFs and scanned documents.
6
7	## Step 1: Remote URL Available?
8
9	If the document has a URL, always try `web_extract` first:
10
11	```
12	web_extract(urls=["https://arxiv.org/pdf/2402.03300"])
13	web_extract(urls=["https://example.com/report.pdf"])
14	```
15
16	This handles PDF-to-markdown conversion via Firecrawl with no local dependencies.
17
18	Only use local extraction when: the file is local, web_extract fails, or you need batch processing.
19
20	## Step 2: Choose Local Extractor
21
22	\| Feature \| pymupdf (~25MB) \| marker-pdf (~3-5GB) \|
23	\|---------\|-----------------\|---------------------\|
24	\| Text-based PDF \| ✅ \| ✅ \|
25	\| Scanned PDF (OCR) \| ❌ \| ✅ (90+ languages) \|
26	\| Tables \| ✅ (basic) \| ✅ (high accuracy) \|
27	\| Equations / LaTeX \| ❌ \| ✅ \|
28	\| Code blocks \| ❌ \| ✅ \|
29	\| Forms \| ❌ \| ✅ \|
30	\| Headers/footers removal \| ❌ \| ✅ \|
31	\| Reading order detection \| ❌ \| ✅ \|
32	\| Images extraction \| ✅ (embedded) \| ✅ (with context) \|
33	\| Images → text (OCR) \| ❌ \| ✅ \|
34	\| EPUB \| ✅ \| ✅ \|
35	\| Markdown output \| ✅ (via pymupdf4llm) \| ✅ (native, higher quality) \|
36	\| Install size \| ~25MB \| ~3-5GB (PyTorch + models) \|
37	\| Speed \| Instant \| ~1-14s/page (CPU), ~0.2s/page (GPU) \|
38
39	Decision: Use pymupdf unless you need OCR, equations, forms, or complex layout analysis.
40
41	If the user needs marker capabilities but the system lacks ~5GB free disk:
42	> "This document needs OCR/advanced extraction (marker-pdf), which requires ~5GB for PyTorch and models. Your system has [X]GB free. Options: free up space, provide a URL so I can use web_extract, or I can try pymupdf which works for text-based PDFs but not scanned documents or equations."
43
44	---
45
46	## pymupdf (lightweight)
47
48	```bash
49	pip install pymupdf pymupdf4llm
50	```
51
52	Via helper script:
53	```bash
54	python scripts/extract_pymupdf.py document.pdf # Plain text
55	python scripts/extract_pymupdf.py document.pdf --markdown # Markdown
56	python scripts/extract_pymupdf.py document.pdf --tables # Tables
57	python scripts/extract_pymupdf.py document.pdf --images out/ # Extract images
58	python scripts/extract_pymupdf.py document.pdf --metadata # Title, author, pages
59	python scripts/extract_pymupdf.py document.pdf --pages 0-4 # Specific pages
60	```
61
62	Inline:
63	```bash
64	python3 -c "
65	import pymupdf
66	doc = pymupdf.open('document.pdf')
67	for page in doc:
68	print(page.get_text())
69	"
70	```
71
72	---
73
74	## marker-pdf (high-quality OCR)
75
76	```bash
77	# Check disk space first
78	python scripts/extract_marker.py --check
79
80	pip install marker-pdf
81	```
82
83	Via helper script:
84	```bash
85	python scripts/extract_marker.py document.pdf # Markdown
86	python scripts/extract_marker.py document.pdf --json # JSON with metadata
87	python scripts/extract_marker.py document.pdf --output_dir out/ # Save images
88	python scripts/extract_marker.py scanned.pdf # Scanned PDF (OCR)
89	python scripts/extract_marker.py document.pdf --use_llm # LLM-boosted accuracy
90	```
91
92	CLI (installed with marker-pdf):
93	```bash
94	marker_single document.pdf --output_dir ./output
95	marker /path/to/folder --workers 4 # Batch
96	```
97
98	---
99
100	## Arxiv Papers
101
102	```
103	# Abstract only (fast)
104	web_extract(urls=["https://arxiv.org/abs/2402.03300"])
105
106	# Full paper
107	web_extract(urls=["https://arxiv.org/pdf/2402.03300"])
108
109	# Search
110	web_search(query="arxiv GRPO reinforcement learning 2026")
111	```
112
113	## Split, Merge & Search
114
115	pymupdf handles these natively — use `execute_code` or inline Python:
116
117	```python
118	# Split: extract pages 1-5 to a new PDF
119	import pymupdf
120	doc = pymupdf.open("report.pdf")
121	new = pymupdf.open()
122	for i in range(5):
123	new.insert_pdf(doc, from_page=i, to_page=i)
124	new.save("pages_1-5.pdf")
125	```
126
127	```python
128	# Merge multiple PDFs
129	import pymupdf
130	result = pymupdf.open()
131	for path in ["a.pdf", "b.pdf", "c.pdf"]:
132	result.insert_pdf(pymupdf.open(path))
133	result.save("merged.pdf")
134	```
135
136	```python
137	# Search for text across all pages
138	import pymupdf
139	doc = pymupdf.open("report.pdf")
140	for i, page in enumerate(doc):
141	results = page.search_for("revenue")
142	if results:
143	print(f"Page {i+1}: {len(results)} match(es)")
144	print(page.get_text("text"))
145	```
146
147	No extra dependencies needed — pymupdf covers split, merge, search, and text extraction in one package.
148
149	---
150
151	## Notes
152
153	- `web_extract` is always first choice for URLs
154	- pymupdf is the safe default — instant, no models, works everywhere
155	- marker-pdf is for OCR, scanned docs,

People who liked this also viewed

6 picks

Type