A natural language interface for computers
Better local model support, bug fixes, faster load, new features. Official release + blog post coming soon.
Fixed a potential recurring computer
import that occurs when teaching new skills to the 01
, along with several great fixes from the community:
Full Changelog: https://github.com/OpenInterpreter/open-interpreter/compare/v0.2.3...v0.2.4
interpreter --local
is really good now thanks to @tyfiero. Try it!
Full Changelog: https://github.com/KillianLucas/open-interpreter/compare/v0.2.2...v0.2.3
Quick fixes for computer.mail
and computer.sms
.
The New Computer Update Part II introduces our first local vision model, a suite of native Mac integrations, 5x launch speed, and dozens of other requested features.
The new calendar
, contacts
, browser
, mail
, and sms
Computer API modules let Open Interpreter control your Mac's native applications.
For example, Open Interpreter can execute the following code to get a contact, then send a text and an email with an attachment.
The browser
lets Open Interpreter quickly browse the web by querying a web-enabled language model.
We have developed a model called point
which is capable of locating visual controls precisely. It was designed to run locally on consumer hardware.
We leverage existing open-source models to "semantically search the screen" for text and icons. Language models can then call on this composite model to 'point' at text or icons on the screen:
While this model is capable of understanding simple interfaces with discrete icons, we intend to explore more general solutions in the next few weeks.
…
This is the most significant upgrade to Open Interpreter since 0.1.0
. Almost every component has been rewritten to support our project's objective—building a standard interface between language models and computers.
Note: This update is not backwards compatible. If you use Open Interpreter in Python, please read our migration guide.
In 2023, Open Interpreter started building the world's first real-time code execution environment for language models.
Now, we're creating an API for language models to use in that environment, starting with basic I/O controls like display
, mouse
, and keyboard
:
computer.display.view() # Returns a screenshot to vision models
computer.mouse.click("On-screen text") # Locates text on-screen and clicks it
computer.mouse.move(icon="Description of icon") # Locates icon on-screen and hovers over it
computer.clipboard.view() # Returns the contents of the user's clipboard
# Full reference: https://docs.openinterpreter.com/computer-api/reference
We are also launching a free preview of the hosted tools that power computer.mouse
at api.openinterpreter.com.
OS
ModeYou can instruct Open Interpreter to use the Computer API to control your computer graphically:
interpreter --os
Even local vision models running via .llamafile
, LM-Studio
, or Jan.ai
are supported.
To support the incoming L
anguage M
odel C
omputer architecture, the new messaging format that powers Open Interpreter extends OpenAI's messages format to include additional information, and a new role called computer
:
[
{
"role": "assistant",
"type": "code",
"format": "python",
"content": "plot = create_plot_from_data('data')\ndisplay_as_image(plot)\ndisplay_as_html(plot)"
},
{
"role": "computer",
"type": "image",
"format": "base64.png",
"content": "base64"
},
{
"role": "assistant",
"type": "message",
"content": "Plot generated successfully."
}
]
Read about LMC Messages
here.
In addition to these major updates, 0.2.0
comes with a suite of fixes and enhancement from our growing open-source community:
Full Changelog: https://github.com/KillianLucas/open-interpreter/compare/v0.1.17...v0.2.0
This is only the beginning. Happy 2024. ●
●
The New Computer Update is the most significant update to Open Interpreter since v0.1.0
.
Ahead of its release, we're soft-launching this preview to get feedback on the update.
pip install --upgrade open-interpreter
interpreter --os
gpt-4-1106-preview
→ gpt-4
We decided to make gpt-4
no longer default to gpt-4-1106-preview
, which we've found performs worse on code interpretation tasks.
Important Note: This was not published from the repository. The last release, v0.1.16
, was simply patched to not switch to gpt-4-1106-preview
. The updates in the repo (and the commit associated with this release) are therefore out of sync with the binaries below.
We'll be back in sync by the next release of 0.2.0
, the New Computer update.
Critical fixes and cool new magic commands in this one.
%% {insert shell code}
for custom shell code running and %info
added by @CyanideByteconfig
thanks to @hargunmujral in https://github.com/KillianLucas/open-interpreter/pull/786
gpt-4
, we give a great error, and let them switch to gpt-3.5
gpt-4
vision
now works via the Python packageREADME
thanks to @iongpt!terminal_interface
, another for core
shell
output processingThank you to everyone that made it happen!
Cannot go without thanking the extraordinary @ericrallen for this release as well, who practically single-handedly brought our issues below 100 (from nearly 300 just a few weeks ago). Have a safe trip to Japan Eric!
Full Changelog: https://github.com/KillianLucas/open-interpreter/compare/v0.1.15...v0.1.16
Quick fixes to resolve some common issues:
end_of_message
wasn't always being emitted)Full Changelog: https://github.com/KillianLucas/open-interpreter/compare/v0.1.14...v0.1.15