Follow new updates and improvements to Chasm.
June 20th, 2024
Updated the Model Information UI:
Add two left right button for user to scroll the models horizontally.
Added model suggestions button to every model page. Reorganized models and providers in alphabetical order for easier navigation.
Reorganized models and providers in alphabetical order for easier navigation.
Implemented a copy button feature for the Prompt and Workflow output sections.
Fixed the 500 error occurring when sending a string input to the "Input" field.
June 10th, 2024
Introduced the new model "Merovingian I".
Introduced the new model "Eridu".
Added a Beta tag to new and experimental models.
Fixed the issue where the LLM reset upon expanding and closing prompt node windows.
Fixed the issue where the modal to inform users to top up credits was not triggered when attempting to run a prompt with low credits, resulting in a console error.
Fixed the issue where the "Settings > Usage: Monthly Bill" section included dataset costs instead of just top-up costs.
Fixed the issue where the "Datasets: Chunk Overlap", "Chunk Size", and "Top K" sliders reset to zero when using backspace.
Fixed the issue where the "Datasets: Source Text" field displayed file content instead of chunks.
May 29th, 2024
Introduced the new model "Qwen 1.5 14B Chat"
Fixed the issue where the filter could not be cleared by clicking the 'x' icon in the Usage-History Module.
Fixed the issue where the Date filter in the Usage-History Module did not include the end date, omitting results that matched the end date.
Fixed the issue in the prompts editor where it did not automatically scroll down to follow the cursor when hitting enter, requiring users to scroll manually.
May 20th, 2024
Introduction of the Dataset and RAG Module, allowing users to process long context files through the new data retrieval feature
Introduced the new model "GPT 4o"
Addition of a Blog post section to share AI knowledge
Inclusion of Model's uptime to provide users with more detailed model information
Enhanced User Experience on the flow where system prompt should come before user prompt
Fixed the issue of incorrect names being displayed for Workflows inside the History -Usage table.
Fixed the issue where cursor focus gets displaced to the end when the user double clicks on an input text and attempts to replace.
Fixed the issue where pasting HTML formatted text twice resulted in the text format going missing, resulting in raw HTML being pasted.
May 8th, 2024
Introduced the new model βMythoMax L2 13Bβ.
May 2nd, 2024
Introduced the new model βMeta: Llama 3 70B Instructβ.
Resolved an issue with text in system prompt and user prompt breaking by character instead of by word.
April 19th, 2024
Implemented error handling for prompts involving deprecated models, guiding users to select live models.
Introduced Execution Summary feature on Usage history, displaying transaction details like Token in, Token out, and Inference time.
Resolved issue of screen expansion with long text input lacking spaces.
Fixed unexpected app width increase when returning code blocks in output.
April 10th, 2024
Introduced the new model "mixtral-8x-7b."
Implemented a value suggestion feature for input on the Prompt Playground and global input on the Workflow Playground.
Added a "Compare Models" feature on the Pricing page.
Enhanced the Select Model UI to provide more comprehensive model information.
Improved the filter design on the Usage tab.
Added a banner notification for LLMS timeout requests.
Resolved the Input and Global Input value bug to allow numbers.
March 26th, 2024
New Features:
Introduced Pricing Plans
Launched a new pricing landing page
Added top-up credits feature through different tiers of plans on the account usage page
Implemented display of estimated and actual charges in prompt and workflow pages
Enabled low credit notifications through toast messages and email
Improvements:
Improved visibility of the publish button on small screen sizes
Expanded tooltip coverage across the UI
Enhanced onboarding experience for new accounts with three sample workflows and their prompts
Enhanced user experience by allowing in-page prompt editor windows to close upon hitting the "ESC" button or clicking outside of the editor windows
Changes:
Upgraded GPT 3.5 models to GPT 3.5 Turbo 1106
Implemented a limit of 50 API key creations
Bug Fixes:
Resolved issue with displaying two prompt label tags ("Live" or "Draft") Instead of one
Fixed unsaved changes alert pop-up upon leaving a prompt without changes
Corrected issue where prompts were not searchable after the first page in the workflow editor
Fixed connected prompt in workflow using previous variables as the current variable upon prompt run
January 22nd, 2024
Added
- Save workflow shortcut key (Ctrl/Cmd + S) [#973]
Fixed
- Constant variables value is not populated when the nodes are connected to more than one prompt node [#978]
- User is unable to search for prompts in workflow module [#979]
- Side panel config is not in sync with in-page prompt editor in the workflow module [#982]
Changed
- Frequency of workflow module autosave to 2000ms [#981]
Removed
- Query validation for workflow autosave [#980]
- Permission for users to upload more than one file [#989]