LLM-driven extraction from raw HTML and website screenshots, preserving spatial context with optional validation.
Project description
# extracthero
Now you can extract information from any data with almost zero compromise.
## Features
- **Multi-modal input**: ingest raw HTML strings, screenshot image files, or paired HTML+image inputs
- **Spatial context preservation**: maintain element positions to extract data that depends on layout or visual proximity
- **LLM-powered extraction**: plug in your favorite LLM (e.g. OpenAI, Anthropic Claude) for flexible schema-driven parsing
- **Optional validation**: enable built-in rules or custom validators to enforce types, ranges and required fields
- **Extensible pipeline**: hook into pre- and post-processing steps
## Installation
```bash
pip install extracthero
Quickstart
from extracthero import Extractor, ExtractConfig
# 1. Initialize with your LLM backend and schema
config = ExtractConfig(
llm_backend="openai",
prompt_template="Extract product titles and prices from the page",
validation_enabled=True
)
extractor = Extractor(config)
# 2a. Extract from raw HTML
html = "<html>…</html>"
result_html = extractor.extract_from_html(html)
print(result_html)
# 2b. Extract from a screenshot
result_img = extractor.extract_from_screenshot("page.png")
print(result_img)
# 2c. Extract using both HTML + screenshot (preserve layout)
result_combo = extractor.extract_from_both(html, "page.png")
print(result_combo)
Configuration
| Option | Type | Default | Description |
|---|---|---|---|
llm_backend |
str |
"openai" |
Which LLM to use (“openai”, “anthropic”, etc.) |
prompt_template |
str |
None |
Template guiding the LLM’s extraction instructions |
validation_enabled |
bool |
False |
Turn on built-in schema and rule validation |
ocr_engine |
str |
"tesseract" |
OCR engine for screenshot text extraction |
spatial_threshold |
float |
0.5 |
Minimum layout-overlap ratio to consider two elements related |
from extracthero import ExtractConfig
config = ExtractConfig(
llm_backend="anthropic",
validation_enabled=False,
spatial_threshold=0.7
)
Validation Logic
When validation_enabled=True, extracted fields are checked against:
- Type rules (e.g. string, number, date)
- Range checks (e.g. price ≥ 0, date within the last year)
- Presence (required vs. optional fields)
Customize or extend:
from extracthero.validation import Validator, FieldRule
# custom rule: price must be < 1 000 000
class PriceRule(FieldRule):
def validate(self, value):
return isinstance(value, (int, float)) and value < 1_000_000
config.custom_validators = {
"price": PriceRule()
}
CLI Interface
# Extract from HTML file
extracthero html input.html --schema product_schema.json --output out.json
# Extract from screenshot
extracthero image page.png --prompt "Get titles" --output out.json
# Combined mode
extracthero combo input.html page.png --output out.json
Use extracthero --help for full options and flags.
Examples
- E-commerce product scraper
- News article metadata extraction
- Invoice data capture with layout
See the examples/ folder for ready-to-run demos.
Contributing
- Fork the repo
- Create a feature branch (
git checkout -b feat/my-feature) - Implement & test
- Submit a pull request
Please follow our code style guide and write tests for new features.
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
extracthero-0.0.3.tar.gz
(14.8 kB
view details)
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file extracthero-0.0.3.tar.gz.
File metadata
- Download URL: extracthero-0.0.3.tar.gz
- Upload date:
- Size: 14.8 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.1.0 CPython/3.9.23
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
32c86e39792ab3798e592b39c7399c2e9f8338e349f1610bc62c30c7dea0d2a5
|
|
| MD5 |
7de9142c0348d5e9f6a8300729ef3d12
|
|
| BLAKE2b-256 |
5b529b84cf248b1282bf90113b2564c4473d871823d7fbd3860ef120ff9bc826
|
File details
Details for the file extracthero-0.0.3-py3-none-any.whl.
File metadata
- Download URL: extracthero-0.0.3-py3-none-any.whl
- Upload date:
- Size: 16.7 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.1.0 CPython/3.9.23
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
0b9160c40971a32c8f6f3ff3dcf61078568ff03712489d958269dfe1f8263717
|
|
| MD5 |
8316162be3b2aa046608e13b489e1710
|
|
| BLAKE2b-256 |
6062bc7f6ff1f8856a8f8ee121d1109d20fc38f4933c40ea54091b96f78ed537
|
