Welcome to the 7 July 2025 release of Gptcmd! This release introduces new automation and data management features, including a macro system, message-level metadata, and support for audio attachments.

Important notes

Python 3.7 support has been removed

Gptcmd now requires Python 3.8.6 or later.

New features

• Gptcmd now supports user-defined command macros, making it easier and more efficient to enter complex or frequently-used commands.
  • Macros are defined in the [macros] table of the Gptcmd configuration file. Each key in [macros] creates a custom command; each value lists the underlying commands to run, one per line.
  • Positional arguments ({1}, {2}, {*}), default parameters ({1?default}), and built-in variables ({account}, {model}, {thread}) are supported.
  • For more information on the macro format, consult Gptcmd's default configuration.
• Messages can now store arbitrary key–value metadata. Use the meta and unmeta commands to add, view, or remove metadata on any message. This is useful for keeping personal notes or for interacting with third-party LLMProviders that make special parameters available.
• Gptcmd now supports audio: the audio command allows local or remote audio files to be attached to messages, similar to the existing image command.
  • Audio requires support from the active LLMProvider: OpenAI (via an audio-specific model) and Gemini (when configured as an "OpenAI-like" provider) are known to work.
• A grep command has been added to search the active thread using regular expressions. This is especially useful if you forget the index of a message and want to find it again!

Changes

• Gptcmd now loads non-default accounts on demand, improving start-up time for users with many configured accounts.
• Cost estimation has been updated for the o3 model family and added for the gpt-4o-audio-preview model.

Bug fixes

• Crash dumps should now be saved in a broader range of exceptional situations, improving data recovery.

Welcome to the 27 April 2025 release of Gptcmd! This release focuses on improving model handling and compatibility (especially for "OpenAI-like" provider configurations), increasing stability, and refining the user experience.

Important notes

Python 3.7 deprecation

Python 3.7 reached end-of-life on 27 June 2023. This is the final release with 3.7 compatibility. Future releases will require Python 3.8.6 or later.

New features

• Gptcmd can now disambiguate partial model names (for instance, resolving 4o to gpt-4o if appropriate), simplifying model selection.
• Ongoing API requests can now be cancelled by pressing Ctrl+c.
• Added cost estimation support for the gpt-4.1 model family, o3, and o4-mini.
• LLMProvider implementations can now opt out of model validation, improving support, for instance, for "OpenAI-like" providers that don't implement a model listing endpoint.
• The automatic creation of new named threads by the retry command is now configurable via the create_new_thread_on_retry configuration setting (set to ask by default). To restore the previous behaviour, set this option to always.
• To prevent data loss, Gptcmd now attempts to save messages and some application state to a JSON file in the current directory when an unexpected application failure occurs.

Changes

• On OpenAI accounts, gpt-4.1 is now used by default if available.

Bug fixes

• Improved handling of edge cases in OpenAI responses, enhancing stability especially when interacting with "OpenAI-like" providers.
• General improvements to stability and performance.

Welcome to the 7 March 2025 release of Gptcmd!

New features

• Added cost estimation support for gpt-4.5-preview.

Changes

• Enabled streaming on o1.

Welcome to the 9 February 2025 release of Gptcmd! This minor update keeps Gptcmd current with OpenAI's latest changes, particularly for the o-series models.

New features

• Gptcmd's cost estimator now supports the latest available o1, o3-mini, and gpt-4o models.
• Added support for system messages on o1 and o3 series models.

Changes

• Gptcmd now shows a simpler error message when an OpenAI account fails to initialize from configuration. (#1)
• Since the default model is now vision-capable and to reduce false positives, the "this model may not support vision" warning has been removed.
• Messages can now be moved backwards within a thread (for example, move -1 1 is equivalent to flip in Gptcmd 1.x).

Welcome to the second 29 November 2024 release of Gptcmd.

Bug fixes

• Gptcmd now once again correctly handles unquoted file paths on Windows, thanks @Grahamwp.

Welcome to the 29 November 2024 release of Gptcmd! This is a very substantial release that introduces multi-provider and multi-account support, adds the ability to attach images to messages for use with vision models, implements a new configuration system and message editor, and enhances cost estimation, streamed responses, and the general command-line experience.

Important notes

Python 3.7 deprecation

Python 3.7 reached end-of-life on 27 June 2023. While Gptcmd currently maintains best-effort compatibility with Python 3.7, this support is deprecated and will be removed in the next release. Updating to Python 3.8.6 or later is strongly recommended.

JSON file compatibility

JSON files created with older versions of Gptcmd can be loaded with this version, but any files saved with Gptcmd version 2.0.0 or later will be incompatible with previous versions due to changes in the JSON format. Users attempting to load the newer JSON format using a previous release will be instructed to update Gptcmd to version 2.0.0.

New features

• An image command has been added, which allows images to be attached to messages by file or URL. Consult the readme for instructions on using Gptcmd with vision models.
• Gptcmd can now be used with additional providers besides OpenAI:
  • Azure AI.
  • OpenAI-compatible APIs, such as OpenRouter and Ollama.
  • Anthropic Claude.
  • A custom provider of your own design.
• Cost estimation has been completely rewritten:
  • The new cost estimator supports nearly all OpenAI models.
  • Gptcmd now provides cost estimates for streamed responses that complete successfully.
  • Incomplete cost estimates (estimates for sessions where not every response has a cost estimate available) can optionally be enabled.
  • Gptcmd now takes the discount on cached prompt tokens into account when calculating estimated OpenAI costs.
• Gptcmd can now display prompt and sampled token usage on streamed responses.
• Gptcmd now has a configuration system for setting application options and specifying credentials for large language model provider accounts. Consult the readme for more information about the configuration format.
  • An account command has been added to switch between configured large language model provider accounts.
• The flip command has been replaced with a move command that allows for arbitrary message reordering.
• The slice command has been replaced with a copy command that appends copies of a message range to a specified thread.
• Gptcmd now supports the use of an external text editor for some operations:
  • With no arguments, the user, assistant, system, and say commands now open an external editor for message composition.
  • An edit command has been added, which opens the selected message in an external editor so its content can be modified.
• Command feedback across the application has been significantly improved.

Changes

• Gptcmd now displays token usage by request, not by session.
• The n API parameter, which controls the number of responses generated by OpenAI models, is no longer supported.
• The default temperature setting of 0.6 has been removed, so no temperature value is sent with API requests unless explicitly set. This means that the default temperature for OpenAI requests is now effectively set to 1.
• Gptcmd now works with API parameters and the selected model on a per-account (not per-thread) basis and no longer saves these parameters to JSON files.
• By default, Gptcmd now streams responses when possible.
• Gptcmd now uses gpt-4o by default in OpenAI sessions.
• The retry command now deletes from the end up to the last assistant message, not the entire span of messages after the last user message, before resending. In conversations of alternating user and assistant messages, this change has no effect. However, this greatly simplifys the use of retry with models that allow the generated assistant response to be constrained with a custom prefix.

Bug fixes

• Gptcmd is now much more stable when streaming responses.
• Gptcmd's command parsing has been improved, including better handling of quoted file paths and more predictable behaviour with message ranges containing a single negative index.
• General improvements to stability and performance have been introduced as part of a larger refactoring effort.

Welcome to the 22 July 2024 release of Gptcmd!

New features

• gpt-4-turbo is now used by default if available.

Bug fixes

• Fixed role validation in the name command (thanks @Grahamwp).

Welcome to the 19 February 2024 release of Gptcmd!

New features

• In place of "OK", Gptcmd now reports more descriptive messages to confirm the success of various actions across the application.
• A new command-line flag (-m or --model) has been added to switch to a specific OpenAI model on launch. For example, gptcmd -m gpt-4-turbo-preview will use the new GPT-4 model.
• Guidance on API parameter usage in the readme has been updated, especially concerning limiting output length when running GPT at high temperatures.

Bug fixes

• Fixed a crash in message thread deserialization.
• When loading a thread from disk with non-default API parameters, Gptcmd no longer always warns to save changes even when none have been made.
• Attempting to run retry in a thread containing only assistant messages no longer causes Gptcmd to crash.

Welcome to the 7 November 2023 release of Gptcmd!

This release makes Gptcmd compatible with the v1 OpenAI SDK, which will enable new functionality in future updates.

Welcome to the 28 October 2023 release of Gptcmd!

Bug fixes

• Improved validation of message roles (user/assistant/system) across the application. Thanks @RobGallo for reporting!