All notable changes to this project will be documented in this file.
The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.
- Fixed issue where the OPENAI_BASE_URL was no longer being set correctly in the 0.10.0 update.
- Streaming is disabled for all completion requests to prevent issues
mark models
subcommand will list all available LLM models available
- OpenAI reasoning models now work with Mark
- LLM is now a dependency to expand model availability to Mark and delegate low-level model interaction.
- Added model specification test
- Import direct import of
openai
(referencing LLMs dependencies directly)
- Import of
image_to_data_url
has been removed from LangChain.
- Slightly improved errors raise when scraping pages on low memory hardware
- Bump langchain from 0.2.14 to 0.2.15
- Bump httpx from 0.27.0 to 0.27.2
- Bump openai from 1.42.0 to 1.43.0
- Bump ipython from 8.26.0 to 8.27.0
- Bump langchain-community from 0.2.4 to 0.2.12
- Bump openai from 1.41.1 to 1.42.0
- Bump openai from 1.14.2 to 1.41.1
- Bump langchain from 0.2.1 to 0.2.14
- Bump markdownify from 0.12.1 to 0.13.1
- Bump pyyaml from 6.0.1 to 6.0.2
- Bump langchain-community from 0.2.1 to 0.2.4
- Bump ipython from 8.21.0 to 8.26.0
- Bump flake8 from 7.1.0 to 7.1.1
- Bump pytest from 6.2.5 to 8.3.
- Support for
--model
option to allow for selecting a specific OpenAI model
- Support for
--version
option in the CLI
- Aliases for cli options
--system
(-s
) and--generate-images
(-i
)
- Gracefully handle timeouts when fetching urls
- Updated the scraping logic to render pages as clean markdown which exposes the LLM to urls on the page.
- Gracefully handle broken links in markdown files.
- Cleaner OpenAI error handling for common issues
- Response log for image generation
- Ability to override the OpenAI endpoint with OPENAI_API_BASE_URL env var
- Adding experimental support for DALL-E image generation
- Requests are now logged to
~/.mark/logs/
- USER_AGENT warning
- Included additional files in the project for
templates/default_system_prompt.md
.
- Updated default system prompt and refactored into the templates directory.
- Returned a pretty error if no
OPENAI_API_KEY
is found.
- Fixed stdin use case.
- Added local file references to page links.
- Utilized LangChain image utilities for local image encoding.
- Bumped version.
- Added new dependencies:
langchain ^0.2.1
andlangchain-community ^0.2.1
.
- Updated model.
- Fixed issue handling malformed image tags.
- Fixed pathing issue with images.
- Added parsing support for images in markdown text.
- Added new dependencies:
beautifulsoup4 ^4.12.3
andmarkdown ^3.6
.
- Initial setup with dependencies:
python ^3.8
,PyYAML 5.4.1
,ipython 8.21.0
,openai 1.14.2
. - Replace
typer
withclick
for CLI tool. - Setup CLI interface with entry point
agent_gpt.__main__:cli
. - Added development dependencies:
pytest ^6.2.5
.