0.5.1

• Update the prompt which generates each subtopic to discourage the model from sometimes making false promises in the generated text. This will invalidate the disk cache for each subtopic, but not the list of subtopics.
• [breaking] Update the CLI to replace the obsolete fire package with click. If previously specifying -c False to skip confirmations, now -nc can be specified instead. Other command-line arguments should remain unaffected.
• Update dependency versions.

0.4.0

• [breaking] In command-line mode, change the default value of -c from False to True, thereby requiring confirmations by default. To not require confirmations when using as a command-line application, use -c False.
• Add a devcontainer definition, also documenting it in the readme.

0.3.0

• Fix an issue whereby the topic name was not getting vocalized sometimes in the resulting audio file. The punctuation around the topic name has been updated to be more friendly to the TTS. This change however invalidates the cached audio file on disk for the first subsection of a topic.

0.2.2

• Retry if the generated list of subtopics is detected as being invalid. An observed example is when an invalid numbering of subtopics was generated.

0.2.1

• Update the prompt which lists subtopics to avoid including fluff or filler content. To some extent, this should prevent any irrelevant subtopics, making each generation slightly smaller. This however invalidates any prior work cache on disk.

0.1.5

• As necessary, make more than one attempt to get the voice. This is useful because the first attempt sometimes obtains an invalid voice.

0.1.4

Initial release.