javier/investbrain
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix:refined follow up prompt

Browse Source
This commit is contained in:
hackerESQ
2024-11-03 16:13:24 -06:00
parent fb96792821
commit 56ceb92c2b
1 changed files with 9 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files
resources/views/livewire/ai-chat-window.blade.php
+9 -8
View File
@@ -121,7 +121,7 @@ new class extends Component {
"properties" => [
"text" => [
"type" => "string",
"description" => "The prompt question (no more than 5 words)"
"description" => "The suggested prompt question (no more than 5 words)"
],
"value" => [
"type" => "string",
@@ -140,20 +140,21 @@ new class extends Component {
],
'messages' => [
['role' => 'system', 'content' => "
Your role is to assist investors in asking thoughtful questions to their investment advisors.
Your role is to assist investors in asking thoughtful questions of their investment advisors.
When you help investors ask good questions, you should ensure the questions are based on the
provided context. Be sure to keep the questions short!
When you help investors ask good questions, you should ensure the you questions you recommend
are based on the provided context. Be sure to keep the questions short!
The questions you recommend might be based on natural follow up from the given context, requests
to clarify undefined terms, common decision frameworks, possible risks or benefits, or commonly
understood investing concepts.
to further refine a previous response, clarify undefined terms, common decision frameworks,
possible risks or benefits, or commonly understood investing concepts that may require additional
explanation.
Your response should only include valid JSON.
"],
['role' => 'user', 'content' => "
Generate a list of ". rand(1,5) ." follow up questions a savvy investor might ask their advisor
based on the following conversation:
Generate between 1 and 5 (no more than 5) follow up questions a savvy investor might ask their
advisor based on the following conversation:
\n\n
".json_encode(array_slice($this->messages, -4))
],