feat: add ability to ask LLM questions about a report #75
Conversation
| protected readonly models: Model[] = [ | ||
| {id: 'gemini-2.5-flash', name: 'Gemini 2.5 Flash'}, | ||
| {id: 'gemini-2.5-pro', name: 'Gemini 2.5 Pro'}, | ||
| {id: 'gemini-2.5-flash-lite', name: 'Gemini 2.5 Flash Lite'}, |
There was a problem hiding this comment.
In a follow-up, we could look into adding other models here. For now, this is a starting point.
| [class.hidden]="!isAiAssistantVisible()" | ||
| [reportGroupId]="reportGroupId()" | ||
| (close)="isAiAssistantVisible.set(false)" | ||
| /> |
There was a problem hiding this comment.
This is the new code. The rest is apparently prettier formatting
Often times I find the AI Summary to be super helpful, but there are also cases where the summary is lacking critical information, or giving me too verbose information. Every user might look for different things here. To address this need, and leverage the full power of AI for e.g. quantifying findings (instead of manual chore work, clicking through every app), this PR introduces an AI assistant chat. Users can ask questions about a report and get answers! This is an extension of the AI Summary and can be extremely helpful.
There was a problem hiding this comment.
Looks great, thanks @devversion 👍 I've left a couple of comments, but we can address them in a followup PR (I will proceed with the merge).
| <span class="material-symbols-outlined">smart_toy</span> | ||
| </button> | ||
|
|
||
| <app-ai-assistant |
There was a problem hiding this comment.
We can use @defer here :) The amount of code is quite small atm, so it's not super important, but having a @defer would allow us to have more code without affecting IPL time.
| </button> | ||
|
|
||
| <app-ai-assistant | ||
| [class.hidden]="!isAiAssistantVisible()" |
There was a problem hiding this comment.
Also a quick question: this assistant uses Gemini models atm, I'm wondering if we should have a check whether Gemini key is provided on the server and don't bring up the UI if not? For ex. via an API call like /api/assistant/config, which can return a JSON like {configured: true} and we can extend it later with more options in the future if needed.
Often times I find the AI Summary to be super helpful, but there are also cases where the summary is lacking critical information, or giving me too verbose information.
Every user might look for different things here. To address this need, and leverage the full power of AI for e.g. quantifying findings (instead of manual chore work, clicking through every app), this PR introduces an AI assistant chat.
Users can ask questions about a report and get answers! This is an extension of the AI Summary and can be extremely helpful.