fix: generate response button disappear on tool call

[louis-menlo]

Describe Your Changes

This PR fixes 2 issues:

1. The generate button doesn't show as expected when there's an incomplete tool call message.
   https://github.com/user-attachments/assets/c49973eb-2862-4833-82d1-8fc6baae727d

2. Gemma3 requires no consecutive user or assistant messages. This is to fix common cases such as regenerating and send a new message.
   https://github.com/user-attachments/assets/3cbc869a-aab0-414e-ac48-5eed5c96cc6d

Fixes Issues

• Closes #
• Closes bug: No way to regenerate or delete assistant response after stopping during tool use generation #5909

Self Checklist

• Added relevant comments, esp in complex areas
• Updated docs (for bug fixes / features)
• Created issues for follow-up changes or refactoring needed

---

Important

Fixes generate button visibility and prevents consecutive user messages in chat functionality.

• Behavior:
  • Fixes generate button visibility in useChat.ts and $threadId.tsx when there's an incomplete tool call message.
  • Ensures no consecutive user messages in CompletionMessagesBuilder in messages.ts.
• Functions:
  • Updates addUserMessage() in messages.ts to prevent consecutive user messages.
  • Modifies generateAIResponse() in $threadId.tsx to handle tool call messages correctly.
• Tests:
  • Adds test for non-consecutive user messages in messages.test.ts.
• Misc:
  • Updates error message in LoadModelErrorDialog.tsx to "Something went wrong".

^{This description was created by for 10763fd. You can customize this summary. It will automatically update as commits are pushed.}

[ellipsis-dev]

Important

Looks good to me! 👍

Reviewed everything up to 4da174a in 2 minutes and 34 seconds. Click for details.

• Reviewed 101 lines of code in 4 files
• Skipped 0 files when reviewing.
• Skipped posting 5 draft comments. View those below.
• Modify your settings and rules to customize what types of comments Ellipsis leaves. And don't forget to react with 👍 or 👎 to teach Ellipsis.

1. web-app/src/containers/dialogs/LoadModelErrorDialog.tsx:55

• Draft comment:
  Changed error text from 'Failed to load model' to 'Something went wrong'. Consider whether a more specific message would aid debugging.
• Reason this comment was not posted:
  Confidence changes required: 0% <= threshold 50% None

2. web-app/src/hooks/useChat.ts:247

• Draft comment:
  Conditionally adding the user message using the 'troubleshooting' flag prevents duplicate messages during regeneration. Confirm that this aligns with the intended behavior in regeneration flows.
• Reason this comment was not posted:
  Confidence changes required: 33% <= threshold 50% None

3. web-app/src/lib/messages.ts:41

• Draft comment:
  Popping the previous user message to avoid consecutive user messages meets the Gemma3 requirement for user messages. Verify if a similar safeguard is needed for consecutive assistant messages.
• Reason this comment was not posted:
  Decided after close inspection that this draft comment was likely wrong and/or not actionable: usefulness confidence = 10% vs. threshold = 50% The comment makes a speculative suggestion about assistant messages without evidence. It asks to "verify" something, which violates our rules. It references an external requirement ("Gemma3") that isn't visible in the context. The actual code change is clear and focused on user messages only. Maybe there really is a Gemma3 requirement about assistant messages that we don't know about. Maybe consecutive assistant messages could cause issues. Without clear evidence of the requirement or proof that consecutive assistant messages are problematic, we should not keep speculative comments or requests to verify things. The comment should be deleted as it asks for verification and makes speculative suggestions without evidence.

4. web-app/src/routes/threads/$threadId.tsx:223

• Draft comment:
  The generateAIResponse function's new branch, which deletes non-user messages until a user message is found, appears to address incomplete tool call messages. Confirm that removing the user message before re-sending is the intended regeneration flow.
• Reason this comment was not posted:
  Confidence changes required: 33% <= threshold 50% None

5. web-app/src/routes/threads/$threadId.tsx:250

• Draft comment:
  The condition for showing the 'Generate AI Response' button now also checks for tool_calls in the message metadata. Ensure that tool_calls are consistently set in the metadata (or equivalent) so that the button appears as expected in all tool call cases.
• Reason this comment was not posted:
  Confidence changes required: 33% <= threshold 50% None

Workflow ID: wflow_xgl8LZlai43LAqpV

^{You can customize by changing your verbosity settings, reacting with 👍 or 👎, replying to comments, or adding code review rules.}

[ellipsis-dev]

Important

Looks good to me! 👍

Reviewed 10763fd in 2 minutes and 26 seconds. Click for details.

• Reviewed 463 lines of code in 1 files
• Skipped 0 files when reviewing.
• Skipped posting 4 draft comments. View those below.
• Modify your settings and rules to customize what types of comments Ellipsis leaves. And don't forget to react with 👍 or 👎 to teach Ellipsis.

1. web-app/src/lib/__tests__/messages.test.ts:143

• Draft comment:
  Consecutive user message de-duplication is correctly tested; the expectation now returns only the latter message as required by Gemma3.
• Reason this comment was not posted:
  Comment did not seem useful. Confidence is useful = 0% <= threshold 50% This comment is purely informative, stating that a test is correctly implemented. It does not provide any actionable feedback or suggestions for improvement.

2. web-app/src/lib/__tests__/messages.test.ts:412

• Draft comment:
  The integration test for a complex conversation flow comprehensively validates the correct ordering and handling of system, user, assistant (with and without tool calls), and tool messages.
• Reason this comment was not posted:
  Comment was not on a location in the diff, so it can't be submitted as a review comment.

3. web-app/src/lib/__tests__/messages.test.ts:100

• Draft comment:
  The behavior of converting messages with empty content to '.' is unexpected; consider adding a comment to clarify the intent behind this fallback.
• Reason this comment was not posted:
  Comment was on unchanged code.

4. web-app/src/lib/__tests__/messages.test.ts:336

• Draft comment:
  The tests covering assistant message normalization with tags (including nested and multiple blocks) are thorough; the various edge cases are well validated.
• Reason this comment was not posted:
  Comment did not seem useful. Confidence is useful = 0% <= threshold 50% This comment is purely informative, as it only states that the tests are thorough and edge cases are well validated. It does not provide any actionable feedback or suggestions for improvement.

Workflow ID: wflow_DcsUFYLJ7KWgqGya

^{You can customize by changing your verbosity settings, reacting with 👍 or 👎, replying to comments, or adding code review rules.}

[github-actions]

Barecheck - Code coverage report

Total: 38.59%

Your code coverage diff: -0.04% ▾

Uncovered files and lines

File	Lines
web-app/src/containers/dialogs/LoadModelErrorDialog.tsx	1, 9-15, 17-20, 22-38, 40-42, 44-54, 56-59, 61-68, 70-72, 74-78, 80-87, 89-93, 95-98, 100-103, 105
web-app/src/hooks/useChat.ts	100-112, 126-132, 142-164, 167, 169, 171, 174, 177-182, 184-185, 190-192, 195-199, 202, 205-207, 209-217, 228, 257-260, 275-286, 303-314, 316-323, 325-341, 344-359, 361-372, 374-379, 381-390, 392-396, 398-400, 403-410, 412-418, 435
web-app/src/routes/threads/$threadId.tsx	1-2, 4-8, 10-15, 17-26, 29-31, 33-46, 48-52, 55-58, 61-63, 65-67, 69-71, 73-80, 83, 85-87, 89-91, 93, 95-96, 98-99, 101, 104-106, 109-117, 120-126, 129, 132, 134-135, 137, 139-144, 146-153, 155-157, 159-160, 163-169, 172-174, 176-177, 180-186, 188-206, 209, 211-217, 219, 222-229, 231-242, 244, 246, 248-253, 255-268, 270-275, 277-278, 280-286, 288-297, 300-303, 305-317, 319-327, 329-335, 337-339, 341-344, 346-348, 350-354, 356