Description

Description
TUI messages get lost or displayed out of order during interactive chat sessions. When sending a new message, the previous message's response appears instead, making the conversation incoherent. No visual indicator (spinner, "typing...", sequence number) shows which message the agent is currently responding to. The agent also attempts tool calls (e.g., sessions.create) that fail silently, and responses to earlier prompts surface long after newer messages have been sent.

Environment

Device:        Brev cloud instance / Linux automation runners
OS:            Ubuntu 22.04 / Ubuntu 24.04
Architecture:  x86_64
Node.js:       v22.22.2
npm:           10.9.7
Docker:        29.4.1
OpenShell CLI: openshell 0.0.36
NemoClaw:      v0.0.28
OpenClaw:      2026.4.9 (0512059)

1. Onboard a sandbox with Cloud API provider
2. Connect to sandbox: nemoclaw my-assistant connect
3. Start TUI: openclaw tui --session test-session
4. Send message A and wait briefly
5. Before agent fully responds, send message B
6. Observe chat area — message B's response may show message A's answer, or messages appear out of order
7. Send message C — agent may respond with generic "Once you tell me your goal, I'll do my best to assist you!" regardless of input

1. Messages are sent and displayed sequentially in order
2. Each agent reply clearly corresponds to the user's input
3. A visual indicator (spinner, "typing...", or sequence number) shows which message the agent is processing
4. Tool calls either succeed or report failure visibly
5. The conversation remains coherent across multiple turns

1. Messages appear to get queued rather than sent sequentially
2. When sending a new message, the previous message's response appears instead
3. No visual indicator shows which message the agent is responding to
4. Agent frequently responds with generic prompts like "Once you tell me your goal, I'll do my best to assist you!" regardless of what was asked
5. Agent attempts tool calls (e.g., sessions.create) that fail silently
6. Responses to earlier prompts surface long after newer messages have been sent
7. The TUI provides no way to distinguish which reply corresponds to which input — chat experience effectively unusable

VDR4 finding: BR-5
Related test case: T5882257 (openclaw tui — reconnect to existing session)

Bug Details

[NVB#6122196]