New resources for feedback loop

[overmode]

Description

Smaller version of #9368, only focused on API and resources
-> AgentMessageFeedbackResource handles sequelize basic db operations
-> front/lib.api.assistants/feedback manages extra logic around feedback creation (esp. around configurationID for global models)
-> Added a few non-attributes to model to allow sequelize joins

Risk

Break feedbacks

Deploy Plan

Deploy front, no migration needed.

[github-actions]

	Warnings
⚠️	Files in **/lib/models/ have been modified and the PR has the migration-ack label. Don't forget to run the migration from prodbox.

Generated by 🚫 dangerJS against abd155a

[overmode]

Did a EXPLAIN ANALYZE for this one
Note that global agents put aside, the assistant with most feedbacks has only 9 feedbacks.

I targeted claude-3-sonnet which has 637 feedbacks across workspaces (removed the workspace filter below, to target them all and get a big number)

The query is

EXPLAIN ANALYZE
SELECT 
"agent_message_feedback"."createdAt", 
"agent_message_feedback"."updatedAt", 
"agent_message_feedback"."agentConfigurationId", 
"agent_message_feedback"."agentConfigurationVersion", 
"agent_message_feedback"."thumbDirection", 
"agent_message_feedback"."content", 
"agent_message_feedback"."isConversationShared", 
"agent_message_feedback"."id", 
"agent_message_feedback"."workspaceId", 
"agent_message_feedback"."agentMessageId", 
"agent_message_feedback"."userId", 
"agentMessage"."id" AS "agentMessage.id", 
"agentMessage->message"."id" AS "agentMessage.message.id", 
"agentMessage->message"."sId" AS "agentMessage.message.sId", 
"agentMessage->message->conversation"."id" AS "agentMessage.message.conversation.id", 
"agentMessage->message->conversation"."sId" AS "agentMessage.message.conversation.sId", 
"user"."id" AS "user.id", 
"user"."name" AS "user.name", 
"user"."imageUrl" AS "user.imageUrl", 
"user"."email" AS "user.email" 
FROM 
"agent_message_feedbacks" AS "agent_message_feedback" 
LEFT OUTER JOIN "agent_messages" AS "agentMessage" ON "agent_message_feedback"."agentMessageId" = "agentMessage"."id" 
LEFT OUTER JOIN "messages" AS "agentMessage->message" ON "agentMessage"."id" = "agentMessage->message"."agentMessageId" 
LEFT OUTER JOIN "conversations" AS "agentMessage->message->conversation" ON "agentMessage->message"."conversationId" = "agentMessage->message->conversation"."id" 
LEFT OUTER JOIN "users" AS "user" ON "agent_message_feedback"."userId" = "user"."id" 
WHERE  "agent_message_feedback"."workspaceId" = 1 AND "agent_message_feedback"."agentConfigurationId" = 'claude-3-sonnet' 
ORDER BY "agent_message_feedback"."agentConfigurationVersion" DESC, "agent_message_feedback"."createdAt" DESC 
LIMIT 50

The output is

-> Index Scan using conversations_pkey on conversations "agentMessage->message->conversation" (cost=0.43..0.47 rows=1 width=19) (actual time=0.003..0.003 rows=1 loops=636)
Index Cond: (id = "agentMessage->message"."conversationId")
-> Memoize (cost=0.30..5.07 rows=1 width=160) (actual time=0.001..0.001 rows=1 loops=636)
Cache Key: agent_message_feedback."userId"
Cache Mode: logical
Hits: 559 Misses: 77 Evictions: 0 Overflows: 0 Memory Usage: 24kB
-> Index Scan using users_pkey on users "user" (cost=0.29..5.06 rows=1 width=160) (actual time=0.004..0.004 rows=1 loops=77)
Index Cond: (id = agent_message_feedback."userId")
Planning Time: 0.872 ms
Execution Time: 8.725 ms