Composum AI 1.2.1
Release summary
The release 1.2.1 introduces speech recognition / dictation of prompts as a new feature: there is now a new button "Dictate" in the AEM Content Creation Dialog and the Side Panel AI that records while you're pressing the button and uses the OpenAI audio transcriptions service to transcribe this into text and add it at the cursur position to the prompt.
There is also a new feature in the RAG / AI supported search: the most important one is that we allow a preprocessing step where the AI suggests search keywords for doing the actual search, instead of just feeding the plaintext query of the user to the JCR search engine. This likely works better if the user actually asks a question.
Also, we switched from gpt-3.5-turbo to gpt-4o-mini as a default model, since since 3.5 is obsolete and more expensive, too.
Downloads
AEM
- Download and install the package , configure an OpenAI API Key at "GPT Chat Completion Service" .
- Or add the Composum AI for AEM all package on Maven Central to your deployment.
- There is an example configuration package you could optionally use or extend from the source
Composum Pages
- Try it live on cloud.composum.com
- As Sling package: https://central.sonatype.com/artifact/com.composum.ai/composum-ai-integration-composum-package/1.2.1
- Prepackaged in the Composum Starter and docker image composum/featurelauncher-composum , both created in the Composum Launcher project.
Dependencies
The AEM variant runs on AEM >= 6.5 and AEMaaCS.
The Composum variant is an extension of Composum Pages and needs at least
version 1.5.1.
What's Changed
- Switch from gpt-3.5-turbo to gpt-4o-mini since 3.5 is obsolete and more expensive, too by @stoerr in #108
- Speech recognition / dictation by @stoerr in #114
- Feature/rag improvements by @stoerr in #115
- some bugfixes and documentation improvements by @stoerr in #123
- Some version bumps by @dependabot
Full Changelog: composum-ai-integration-1.1.6...composum-ai-integration-1.2.1
