Skip to main content

Computer vision, OCR, and input automation toolkit

Project description

CHIVEL

The CHIVEL logo.

CHIVEL, previously known as CHISL (Computer-Human Interaction Scripting Language) is a Python extension meant for controlling your device. It provides simple interfaces for finding things on the screen, controlling the keyboard/mouse, manipulating images, and more. This was originally its own scripting language, but was later revamped into a Python module, so that the powerful features of Python could be used in addition to the computer vision that CHIVEL provides. The project was also renamed from CHISL to CHIVEL, as CHISL was already taken on PyPI.

CHIVEL is a Python automation toolkit for:

  • Screen capture on multi-display setups
  • OCR text lookup
  • Template matching
  • Mouse and keyboard automation
  • Record and replay workflows
  • Lightweight image processing and drawing

It is designed for practical desktop automation and test scripting.

Install

pip install chivel

Quick Start

import chivel as cv

# Capture the primary display and find text.
img = cv.capture()
hits = cv.find_text(img, "Settings")

if hits:
	img.draw_matches(hits, color=cv.Color(255, 0, 0), thickness=2)
	cv.save(img, "hits.png")

Core Data Types

  • Point(x, y)
  • Rect(x, y, width, height)
  • Color(r, g, b, a=255)
  • Match(rect, label=None)
  • Image(width, height, channels=3)
  • Recording(version, recorded_at, stop_key, events)

Display and Capture APIs

DISPLAY_COUNT

Number of detected displays.

print(cv.DISPLAY_COUNT)

set_primary_display(display_index)

Sets the default display used by APIs that allow omitted display index.

cv.set_primary_display(1)

get_primary_display()

Returns the current default display index.

idx = cv.get_primary_display()
print(idx)

display_get_rect(display_index=None)

Returns global screen-space bounds for a display.

rect = cv.display_get_rect()      # uses primary display
rect0 = cv.display_get_rect(0)    # explicit display

capture(display_index=None, rect=None)

Captures a display and returns an Image.

full = cv.capture()  # primary display
region = cv.capture(0, cv.Rect(100, 100, 400, 300))

File IO APIs

load(path)

Loads a supported file path.

  • Images return Image
  • Text-like files (.txt, .md, .log, .json, .csv) return string
  • Directories return a list of loaded entries
img = cv.load("button.png")
text = cv.load("notes.txt")
items = cv.load("assets")

save(image, path)

Saves Image to disk.

cv.save(img, "out.png")

show(image, name="image", blocking=True)

Shows an image in an OpenCV window.

cv.show(img, name="Preview", blocking=True)
cv.show(img, name="Live", blocking=False)

Image and OCR Search APIs

find_image(source, search, threshold=0.8)

Template match. Returns list of Match.

screen = cv.capture()
needle = cv.load("icon.png")
matches = cv.find_image(screen, needle, threshold=0.9)

find_text(source, search, threshold=0.0)

OCR search. Returns list of Match.

  • search can be plain string (case-insensitive substring)
  • search can be compiled regex pattern
import re

screen = cv.capture()
m1 = cv.find_text(screen, "File")
m2 = cv.find_text(screen, re.compile(r"File|Edit|View", re.IGNORECASE))

find_any(source, search, threshold=0.8)

Searches in order and returns first non-empty result.

hits = cv.find_any(screen, ["Save", cv.load("save_icon.png")], threshold=0.85)

find_all(source, search, threshold=0.8)

All search terms must match; returns combined matches or empty list.

hits = cv.find_all(screen, ["Username", "Password"])

expect_any(*search, interval=1.0, timeout=-1.0, display_index=0, threshold=0.8)

Polling wait until any search term matches, or timeout.

hits = cv.expect_any("Done", "Success", timeout=15.0, interval=0.2)

expect_all(*search, interval=1.0, timeout=-1.0, display_index=0, threshold=0.8)

Polling wait until all search terms match, or timeout.

hits = cv.expect_all("Name", "Email", timeout=10.0)

wait(seconds)

Simple sleep helper.

cv.wait(0.5)

Mouse APIs

mouse_move(pos, display_index=None, relative=False)

Moves mouse.

  • relative=False: pos is display-local coordinate
  • relative=True: pos is delta from current position
  • pos can be Point, Rect, or (x, y)
  • when pos is Rect, CHIVEL moves to rect center
cv.mouse_move((200, 150))
cv.mouse_move((20, -10), relative=True)
cv.mouse_move((100, 100), display_index=0)
cv.mouse_move(cv.Rect(300, 200, 120, 40))  # moves to center

Rect.center()

Returns center point of a Rect.

rect = cv.Rect(100, 50, 200, 80)
pt = rect.center()
print(pt.x, pt.y)

mouse_click(button=BUTTON_LEFT, count=1, delay=None)

Clicks with configurable hold delay.

  • delay=None defaults to 0.0 for one click, 0.1 for multiple clicks
cv.mouse_click()
cv.mouse_click(cv.BUTTON_RIGHT)
cv.mouse_click(count=2)  # uses default 0.1 delay

mouse_down(button=BUTTON_LEFT)

cv.mouse_down()

mouse_up(button=BUTTON_LEFT)

cv.mouse_up()

mouse_scroll(vertical, horizontal=0)

cv.mouse_scroll(1, 0)
cv.mouse_scroll(0, -1)

mouse_get_location()

Returns (Point, display_index).

pt, display_idx = cv.mouse_get_location()
print(pt.x, pt.y, display_idx)

mouse_get_display()

Returns current display index.

print(cv.mouse_get_display())

Keyboard APIs

type(text, delay=0.01)

Types text character-by-character, with a delay between each character.

cv.type("hello")
cv.type("slow", delay=0.05)

key_click(keys, count=1, delay=None)

Key tap or chord tap.

  • keys can be one key or a sequence
  • chord behavior is all downs, then all ups
  • delay=None defaults to 0.0 for single-key, 0.1 for chord
cv.key_click(cv.KEY_A)
cv.key_click([cv.KEY_CTRL, cv.KEY_C])

key_down(keys)

Press one or more keys.

cv.key_down([cv.KEY_CTRL, cv.KEY_SHIFT])

key_up(keys)

Release one or more keys.

cv.key_up([cv.KEY_CTRL, cv.KEY_SHIFT])

wait_for(keys_or_buttons, delay=0.01, timeout=-1)

Waits for one of the requested keys or mouse buttons and returns the code. If timeout is set (>0), returns None if no input is detected in that time.

# Wait for Enter, Escape, or left mouse button, with a 5 second timeout
k = cv.wait_for([cv.KEY_ENTER, cv.KEY_ESCAPE, cv.BUTTON_LEFT], timeout=5.0)
if k is None:
	print("Timed out!")
else:
	print("Pressed:", k)

get_clipboard()

Returns the current clipboard text as a string. Returns an empty string if the clipboard is empty or unavailable.

text = cv.get_clipboard()
print("Clipboard:", text)

set_clipboard(text)

Sets the clipboard text to the given string.

cv.set_clipboard("Hello from CHIVEL!")

check_for(keys_or_buttons)

Checks if any key or mouse button in the sequence is currently pressed. Returns the first code found, or None if none are pressed.

pressed = cv.check_for([cv.KEY_CTRL, cv.KEY_SHIFT, cv.BUTTON_LEFT])
if pressed:
	print(f"Pressed: {pressed}")
else:
	print("None pressed")

Input Event Hooks

Subscribe to input events by key/button code.

def on_e_down(code):
	print("E down", code)

def on_left_click(code):
	print("Left click", code)

# Subscribe
cv.on_key_down[cv.KEY_E] += on_e_down
cv.on_mouse_click[cv.BUTTON_LEFT] += on_left_click

# Trigger by using normal input APIs
cv.key_down(cv.KEY_E)
cv.mouse_click(cv.BUTTON_LEFT)

# Unsubscribe
cv.on_key_down[cv.KEY_E] -= on_e_down
cv.on_mouse_click[cv.BUTTON_LEFT] -= on_left_click

For mouse move, you can subscribe globally without a key/index:

def on_any_move(pt, display_index):
	print("move", pt.x, pt.y, display_index)

cv.on_mouse_move += on_any_move
cv.mouse_move((100, 100))
cv.on_mouse_move -= on_any_move

You can also use explicit methods:

cv.on_key_down.on(cv.KEY_E, on_e_down)
cv.on_key_down.off(cv.KEY_E, on_e_down)

cv.on_mouse_move.on(on_any_move)
cv.on_mouse_move.off(on_any_move)

pause(prompt="Press Enter to continue...")

Blocks execution until Enter is pressed in the terminal.

cv.pause()
cv.pause("Press Enter when setup is complete...")

Recording and Playback APIs

record(output_dir=None, simplify=0, simplify_threshold=None, stop_key=KEY_ESCAPE, step_key=None, step_size=(50, 50))

Records input events and returns Recording.

  • output_dir is a folder. If provided, recording.json and step images are saved there.
  • simplify uses SIMPLIFY_* bit flags.
  • simplify_threshold caps max gap between events.
  • step_key enables step mode (captures step images, suppresses move events).

Simplify Flags

The simplify pipeline runs in this order:

  1. SIMPLIFY_MOVE
  2. SIMPLIFY_MOUSE
  3. SIMPLIFY_KEY
  4. simplify_threshold gap capping
  5. SIMPLIFY_TIME

Flag details:

  • SIMPLIFY_NONE: No simplify pass is applied.
  • SIMPLIFY_MOVE: Consecutive mouse_move events are collapsed; only the last move in each run is kept.
  • SIMPLIFY_MOUSE: Removes mouse_down/mouse_up events that have no button value.
  • SIMPLIFY_KEY: Removes key_unknown events.
  • SIMPLIFY_TIME: Adds dt to each event (time offset from the first event). It does not remove time.
  • SIMPLIFY_ALL: Combines all flags above.

How simplify_threshold works:

  • If set (for example 0.5), any gap between consecutive events larger than that value is clamped.
  • This makes long pauses shorter in playback while preserving event order.
  • This applies even if no SIMPLIFY_* flags are set.
# Normal recording
rec = cv.record(output_dir="out_record", stop_key=cv.KEY_ESCAPE)

# Step recording (F8)
rec2 = cv.record(
	output_dir="out_step_record",
	step_key=cv.KEY_F8,
	step_size=(80, 80),
	simplify=cv.SIMPLIFY_ALL,
	simplify_threshold=0.5,
)

play(recording, speed=1.0)

Replays Recording, path, or dict payload.

cv.play(rec)
cv.play("out_record", speed=1.5)

Recording.save(output_dir)

rec.save("session1")

Recording.load(input_path)

Loads from directory or explicit recording.json path.

loaded = cv.Recording.load("session1")

Image Class Method Examples

img = cv.Image(640, 360)
img.draw_text("CHIVEL", cv.Point(20, 40), color=cv.Color(255, 255, 0), font_size=1, thickness=2)
img.draw_rect(cv.Rect(10, 60, 120, 80), color=cv.Color(255, 0, 0), thickness=2)
img.draw_line(cv.Point(10, 10), cv.Point(200, 200), color=cv.Color(0, 255, 0), thickness=2)
img.draw_ellipse(cv.Point(300, 120), 40, color=cv.Color(0, 0, 255), thickness=2)

copy = img.clone()
copy.crop(cv.Rect(0, 0, 320, 180))
copy.grayscale()
copy.blur(5)
copy.sharpen(0.8)
copy.threshold(120)
copy.invert()
copy.normalize()
copy.edge(80, 180)
copy.emboss()
copy.flip(cv.FLIP_HORIZONTAL)
copy.rotate(5)
copy.scale(1.2)
copy.resize(cv.Point(400, 250))
copy.contrast(1.2)
copy.brightness(15)
copy.convert(cv.COLOR_SPACE_BGR)

mask = cv.Image(copy.get_size().x, copy.get_size().y, channels=1)
copy.mask(mask)
channels = img.split()
img.merge(channels)
img.draw_image(copy, cv.Point(20, 20), alpha=0.5)

cv.save(img, "demo.png")

Constants

CHIVEL exports useful constants for mouse buttons, keyboard virtual keys, color spaces, and simplify flags.

Common examples:

  • BUTTON_LEFT, BUTTON_RIGHT, BUTTON_MIDDLE
  • KEY_ENTER, KEY_ESCAPE, KEY_CTRL, KEY_ALT, KEY_F1 through KEY_F12
  • SIMPLIFY_MOVE, SIMPLIFY_MOUSE, SIMPLIFY_KEY, SIMPLIFY_TIME, SIMPLIFY_ALL
  • FLIP_VERTICAL, FLIP_HORIZONTAL, FLIP_BOTH
  • COLOR_SPACE_BGR, COLOR_SPACE_BGRA, COLOR_SPACE_RGB, COLOR_SPACE_RGBA, COLOR_SPACE_GRAY, COLOR_SPACE_HSV

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

chivel-0.8.1.tar.gz (23.0 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

chivel-0.8.1-py3-none-any.whl (19.5 kB view details)

Uploaded Python 3

File details

Details for the file chivel-0.8.1.tar.gz.

File metadata

  • Download URL: chivel-0.8.1.tar.gz
  • Upload date:
  • Size: 23.0 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.14.3

File hashes

Hashes for chivel-0.8.1.tar.gz
Algorithm Hash digest
SHA256 cf67400995bbca712070727aa6b031d5ad7e91b2f0fe4feaa902bd39e5e3cdcf
MD5 cb02cb4a526ba172a47acef4200f26cd
BLAKE2b-256 a7f8fef351332c3042186a6f3807a23526d110ef70c43bb9ef49194a113c20de

See more details on using hashes here.

File details

Details for the file chivel-0.8.1-py3-none-any.whl.

File metadata

  • Download URL: chivel-0.8.1-py3-none-any.whl
  • Upload date:
  • Size: 19.5 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.14.3

File hashes

Hashes for chivel-0.8.1-py3-none-any.whl
Algorithm Hash digest
SHA256 2ffa17358d44079a9eccb15892daa07dc9520f4732b550b954a945caeb242fba
MD5 f321f9a10dcad3e556852d32bc4cef18
BLAKE2b-256 4fbe78c47e0b876850b7bc7fd48e573b24c952fb0abce2b089864e26521205f8

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page