Thanks!
It is enabled, and I have access. Now I am starting to sound like Data from Star Trek. 
https://community.openai.com/discourse-ai/ai-bot/conversations?agent=forum-helper&llm=gpt-56-medium
What I am not sure about is how many users on that forum—or on other forums—would also have access to that page.
As a quick improvement, perhaps the search/Ask AI page could provide a link to the AI Bot conversations page, or otherwise make that option more discoverable. Something like Continue in AI Bot might make the relationship between the two interfaces clearer.
In doing some experimenting with Discourse AI:
The list of available agents can be a bit overwhelming. Having followed some of the earlier evolution of Discourse AI, I have at least some idea what they are for. Coming to this for the first time, though, I think I would find the choices difficult to evaluate.
I tried the same prompt with two agents:
- Forum Helper — because it is described as general purpose.
- Discover — because it was the default agent.
As you noted, the UI and workflow are much better suited to entering a substantial prompt.
The result from Discover was a single paragraph with a few links and, for this particular prompt, did not give me much of value.
The result from Forum Helper, however, was several paragraphs and was worth reading and following up on. It appeared to do considerably more investigation of the internet rather than simply returning a few related topics.
That makes me wonder whether the choice of agent should either be more obvious to the user or perhaps be selected automatically based on what the user is trying to do.
That brings me to another possible issue farther down the road.
When I ask ChatGPT to search https://community.openai.com/ from outside the forum, I have occasionally received a response indicating that the available number of requests for the day has been reached.
I do not yet know whether that limit originates with Discourse, an intermediary, or the external retrieval system, so I do not want to draw a conclusion from it.
The next time I encounter it, I will try to remember to capture the exact message and post it here. It may be useful when considering how external AI systems interact with a Discourse site’s public content.