feat: v2 - ai assistant chat window

[maxy-shpfy]

Description

Introduces the foundational scaffolding for an AI Assistant panel in the V2 editor, gated behind a new ai-assistant feature flag (disabled by default, categorized as beta).

The panel opens as a floating window in the editor and provides a chat interface where users can ask questions about their pipeline. This PR establishes the core architecture:

• AiChatStore (MobX observable) holds messages, thread ID, pending state, and thinking text outside the React tree so state survives window minimize/hide/unmount. The sendMessage implementation is currently a stub that echoes the user's prompt back — the real LLM integration is deferred to a follow-up PR.
• AiChatStoreProvider / useAiChatStore expose the store via React context, wrapping the editor session.
• AiChatContent composes ChatMessageList and ChatInput into the panel body.
• ChatMessageList auto-scrolls to the latest message and renders an empty state when no messages exist.
• ChatInput supports Enter to submit and Shift+Enter for newlines.
• ThinkingMessage displays an animated spinner with status text while a response is pending.
• Markdown rendering via react-markdown + remark-gfm with custom component overrides for headings, lists, tables, blockquotes, code blocks, and links.
• Custom link protocols (entity:// and component://) are resolved into interactive EntityChip and ComponentChip elements. EntityChip navigates to the corresponding task/input/output node in the canvas. ComponentChip is draggable onto the canvas and opens a ComponentDetailsDialog on click.
• useGatedAiChatWindow opens the window only when the feature flag is enabled and the window does not already exist.

Related Issue and Pull requests

Type of Change

• Bug fix
• New feature
• Improvement
• Cleanup/Refactor
• Breaking change
• Documentation update

Checklist

• I have tested this does not break current pipelines / runs functionality
• I have tested the changes on staging

Screenshots (if applicable)

AI Assistant - Echo and markdown for chat.mov (uploaded via Graphite)

Test Instructions

1. Enable the ai-assistant feature flag in the beta flags settings panel.
2. Open a pipeline in the V2 editor — an "AI Assistant" floating window should appear.
3. Type a message and press Enter or click the send button. The message should be echoed back as an assistant reply.
4. Verify the window persists when minimized and re-shown.
5. Disable the flag and confirm the window does not open.

Additional Comments

The sendMessage method currently echoes the user prompt as the assistant response. The actual LLM/worker integration is intentionally left for a subsequent PR. The threadId field on AiChatStore is reserved for that integration.

For test you can use markdown:

# H1: Pipeline assistant test

## H2: Section heading

### H3: Subsection

#### H4: Smallest heading

A paragraph with **bold**, *italic*, ~~strikethrough~~, and `inline code`.
Soft line break here
and another line.

---

## Lists

Unordered list:

- First item
- Second item with `inline code`
- Nested:
  - Nested item one
  - Nested item two
- Last item

Ordered list:

1. Step one
2. Step two
3. Step three

Task list (GFM):

- [x] Completed task
- [ ] Pending task

---

## Blockquote

> This is a blockquote. It should render with a left border, italic, and muted color.
>
> Second paragraph inside the blockquote.

---

## Table (GFM)

| Component        | Status | Notes                        |
| ---------------- | :----: | ---------------------------- |
| Data loader      |   OK   | Reads from GCS               |
| Training task    |  FAIL  | OOM at epoch 3               |
| Evaluation task  |   --   | Skipped (upstream failed)    |


---

## Links and chips

Regular external link: [OpenAI docs](https://platform.openai.com/docs).

* Entity chip (task): [Train model](entity://task-train-001).
* Entity chip (input): [training_data](entity://input-data-001).
* Entity chip (output): [trained_model](entity://output-model-001).
* Entity chip (unknown id, renders as question mark): [missing-thing](entity://does-not-exist).

* Component chip with backing ref (renders draggable `ComponentChip`):
[Format date and time](component://7784069b3bd1301674cfb4bbf7170b7db91ffd289138eb2d8838e3368158b70c).

* Component chip whose id is missing from `componentReferences` (falls through to the `<span>` fallback):
[unknown component](component://not-in-refs).

[maxy-shpfy]

• chore: v2 - ai assistant - unit tests #2336
• feat: v2 - ai assistant - architect subagent #2335
• refactor: v2 - ai assistant - split tools bridges #2333
• feat: v2 - ai assistent - debug agent #2332
• feat: v2 - ai assistent - repair subagent #2331
• feat: v2 - ai assistant - tangle help subagent #2330
• feat: v2 - ai assistant llm proxy and dispatcher agent #2329
• feat: v2 - ai assistant web worker scaffolding #2328
• feat: v2 - ai assistant chat window #2327 👈 (View in Graphite)
• master

This stack of pull requests is managed by Graphite. Learn more about stacking.

[github-actions]

🎩 Preview

A preview build has been created at: 05-26-feat_v2_-_ai_assistant_chat_window/8dfdf7a