fix(chat): concatenate all text parts in conversation item rendering

[omjeem]

Problem

message.parts.find(p => p.type === "text") returns the first text chunk, which is empty in multi-step agent turns. After a tool call, the AI SDK produces a step-start + empty text part for the tool-use step, followed by another step-start + the real response text. .find() grabs the empty one → editor gets "" → blank message.

Solution

Replace .find() with .filter() + .map().join() to concatenate all non-empty text parts into a single string for the TipTap editor.

Changes

• apps/webapp/app/components/conversation/conversation-item.client.tsx
• Replace .find() with .filter() to collect all non-empty text parts
• Update useEditor initial content and useEffect to use the combined string

SDK_Text_Fix.mov

Fix #900