llm-workflow-engine

Power CLI and Workflow manager for LLMs (core package)

MIT License

Stars
3.6K

Bot releases are hidden (Show)

llm-workflow-engine - 0.6.3

Published by thehunmonkgroup over 1 year ago

  • Sat Mar 18 2023: clean up template display/workflows
  • Sat Mar 18 2023: extract description separate from overrides, fixes #238
llm-workflow-engine - 0.6.2

Published by thehunmonkgroup over 1 year ago

  • Sat Mar 18 2023: /templates command improvements
  • Sat Mar 18 2023: fix secondary invocations with browser backend, fixes #236
llm-workflow-engine - 0.6.1

Published by thehunmonkgroup over 1 year ago

🚒Breaking Changes🚒

The --config-dir and --data-dir arguments have changed how they interpret locations:

  • Both now point to the root chatgpt-wrapper directory instead of a profile directory
  • Config and data are still stored under profiles/[profile] subdirectories inside these directories
  • Installations that use the default locations instead of providing CLI arguments for the locations are unaffected
  • See the output of chatgpt config with no other arguments to see these updates reflected in the File configuration section

Full list of changes

  • Fri Mar 17 2023: find version in version.py
  • Fri Mar 17 2023: doc for template front matter
  • Fri Mar 17 2023: refactor config/data dir implementation, support non-profile specific templates/plugins dirs BREAKING CHANGE
  • Thu Mar 16 2023: pretty up templates list output
  • Thu Mar 16 2023: add descriptions to example templates
  • Thu Mar 16 2023: better formatting of template front matter, use description key from front matter in /templates list
  • Thu Mar 16 2023: enable debug logging for test scripts
  • Thu Mar 16 2023: check for running event loop, use if found
  • Thu Mar 16 2023: clarify how to use the sample config
llm-workflow-engine - 0.6.0

Published by thehunmonkgroup over 1 year ago

  • Thu Mar 16 2023: fix crash after initial user creation on api backend
  • Wed Mar 15 2023: Basic plugin functionality (alpha, subject to change)
  • Wed Mar 15 2023: improvements to model handling
  • Tue Mar 14 2023: set new backend model after user edit
  • Tue Mar 14 2023: add set_model method to API backend, error handling/logging for API requests
  • Tue Mar 14 2023: bump openai version requirement
  • Tue Mar 14 2023: Minor bug fix: model option was not used in the wrapper (default option was hardcoded)
  • Tue Mar 14 2023: added gpt4 model option
  • Tue Mar 14 2023: move signal handling to base shell class, fixes #226
  • Tue Mar 14 2023: repl_history file use platform agnostic temp dir, fixes #227
  • Mon Mar 13 2023: Convert commands from underscore to dash
  • Mon Mar 13 2023: don't start gen_title thread if title already exists
  • Mon Mar 13 2023: only add check_same_thread for sqlite connections
  • Sun Feb 26 2023: added flask to requirements
  • Sun Feb 26 2023: improvement to docker (speed up in debugging and adding api port)
llm-workflow-engine - 0.5.5

Published by thehunmonkgroup over 1 year ago

  • Mon Mar 13 2023: fix threading error with SQLite connections
  • Mon Mar 13 2023: updates to example config
  • Sun Mar 12 2023: add note about adding EDITOR env var in Windows
  • Sun Mar 12 2023: try to get windows editor from env first
  • Sun Mar 12 2023: add install notes for windows users
llm-workflow-engine - 0.5.4

Published by thehunmonkgroup over 1 year ago

  • Sun Mar 12 2023: launch backend after check for config CLI arg
  • Sun Mar 12 2023: fix ask/ask_stream signatures to support custom titles
  • Sun Mar 12 2023: add prompt-engineer example
  • Sun Mar 12 2023: add 'Backend configuration' section to config output
  • Sun Mar 12 2023: temp workaround for issue #224
  • Sat Mar 11 2023: allow overriding system message in template front matter
  • Sat Mar 11 2023: add support for frontmatter in templates
llm-workflow-engine - 0.5.3

Published by thehunmonkgroup over 1 year ago

  • Sat Mar 11 2023: add some example templates and API scripts
  • Sat Mar 11 2023: allow passing custom title to ask/ask_stream in api backend
  • Sat Mar 11 2023: init defaults for templates
  • Sat Mar 11 2023: try to discover env editor on osx
  • Sat Mar 11 2023: template_copy/template_delete commands
  • Sat Mar 11 2023: kill special sauce for linux editor filetype, no longer needed
  • Fri Mar 10 2023: ensure self.templates is a list
  • Fri Mar 10 2023: add link to new video walkthrough
  • Fri Mar 10 2023: fix markdown filetype for vim syntax highlighting
llm-workflow-engine - 0.5.2

Published by thehunmonkgroup over 1 year ago

BREAKING CHANGE:

The autocreation of the templates directory was broken, and creating the directory in the wrong location.

To fix this in the default case, and restore access to existing templates:

  1. Change directories to the chatgpt-wrapper/profiles configuration directory
  2. mv defaultdefault/* default/ && rmdir defaultdefault
  • Fri Mar 10 2023: HOTFIX for broken templates directory location
  • Fri Mar 10 2023: indicator for current conversation in /history list
  • Fri Mar 10 2023: tweak /chat help
  • Fri Mar 10 2023: set new conversation in API backend on user login
  • Fri Mar 10 2023: add default_user_id arg to init of API backend
  • Fri Mar 10 2023: add tests for chatgpt-api Python module
  • Fri Mar 10 2023: output user id in users list
  • Fri Mar 10 2023: add utility scripts for commit log and pypi release
llm-workflow-engine - 0.5.1

Published by thehunmonkgroup over 1 year ago

  • Add completions for many more commands
  • Show/set system message (initial context message for all conversations)
  • System message aliases
  • Template management system. See below for details (alpha, subject to change)
  • Set 'markdown' filetype for editor invocations (supports syntax highlighting)
  • Add built template variables, see below for details
  • Native editor module (removes vipe dependency)
llm-workflow-engine - 0.5.0

Published by thehunmonkgroup over 1 year ago

🚒Breaking Changes🚒

  • The return values for the public methods of the ChatGPT/AsyncChatGPT classes have changed, they are now tuple with the following values:

    • success: Boolean, True if the operation succeeded, False if the operation failed.
    • data: Object, the data the command generated.
    • message: Human-readable message about the outcome of the operation.
  • Introduced the concept of multiple 'backends' -- see below for the currently supported ones

  • Added the 'chatgpt-api' backend, communicates via the official OpenAI REST endpoint for ChatGPT

    • Basic multi-user support (admin party at CLI)
    • Data stored in a database (SQLite by default, any configurable in SQLAlchemy allowed)
    • Allows full model customiztion
    • Numerous new shell commands and enhancements
llm-workflow-engine - 0.4.3

Published by thehunmonkgroup over 1 year ago

🚒Breaking Changes🚒

  • ChatGPT/AsyncChatGPT classes have changed how they receive configuration values, be sure to investigate the new function signatues for their init() and create() methods.

What is new?

llm-workflow-engine - 0.4.2

Published by thehunmonkgroup over 1 year ago

  • Fix broken ChatGPT sync class
  • Removed nest_asyncio dependency
  • Convert CLI to use AsyncChatGPT class
  • Initial implementation of stop generating text response
llm-workflow-engine - 0.4.1

Published by thehunmonkgroup over 1 year ago

REVERT BREAKING CHANGE: Asyncio module requirement removed from usage of ChatGPT class, it is now a sync wrapper around the async class

llm-workflow-engine - 0.4.0

Published by thehunmonkgroup over 1 year ago

🚒Breaking Changes🚒

  • Command leader changed from '!' to '/'
  • Asyncio module is now required to use ChatGPT class directly (refer to Python usage)

What is new?

New commands

  • Added '/quit' command
  • Added '/delete' support for history IDs/UUIDs
  • Added '/chat' command
  • Added '/switch' command
  • Added '/title' command
  • Added limit/offset support for '/history'

New features

  • Migrated to async Playwright
  • Initial API in Flask (see How to use the API)
  • Added tab completion for commands
  • Added '/tmp' volume for saving Playwright session
  • Added CI and CodeQL workflows
  • Added simple developer debug module
  • Improved session refreshing (/session now works!)
  • Migrated to Prompt Toolkit
llm-workflow-engine - 0.3.17

Published by thehunmonkgroup over 1 year ago

  • Added debug mode (visible browser window).
  • @thehunmonkgroup fixed chat naming.
  • @thehunmonkgroup added !delete command to remove/hide conversations.
  • @thehunmonkgroup added --model flag to select model ('default' or 'legacy-paid' or 'legacy-free').
  • @thehunmonkgroup added !editor command to open the current prompt in an editor and send the edited prompt to ChatGPT.
  • @thehunmonkgroup added !history command to show the list of the last 20 conversations.
  • @NatLee added docker support.
llm-workflow-engine - 0.3.16

Published by thehunmonkgroup over 1 year ago

  • Ability to open multiple sessions in parallel.
  • Code now works with ChatGPT Plus subscription.
llm-workflow-engine - 0.3.15

Published by thehunmonkgroup over 1 year ago

Updated model to text-davinci-002-render-sha (turbo model).

llm-workflow-engine - 0.3.14

Published by thehunmonkgroup over 1 year ago

llm-workflow-engine - 0.3.13

Published by thehunmonkgroup over 1 year ago

llm-workflow-engine - 0.3.12

Published by thehunmonkgroup over 1 year ago

Badges
Extracted from project README
Test status CodeQL status Star History Chart
Related Projects