yuta/twitter-lite
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat: add URL-driven workspace state and streamed Codex research

Browse Source
This commit is contained in:
yuta
2026-09-28 20:08:28 +09:00
parent 85d23a328b
commit 8085ad2f90
207 changed files with 15287 additions and 16026 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/plans/2026-09-24-home-agent-research.md
+3
View File
@@ -1,5 +1,8 @@
# Home-agent research
Historical design: the WebSocket transport and global active selection described
here were replaced by the [AI SDK research runtime](../research-runtime.md).
## Product boundary
The user starts a research task from the deck workspace. A home agent chooses
docs/research-decks.md
+19 -12
View File
@@ -24,15 +24,15 @@ list, user-profile, and conversation pages are not provided.
unique column IDs within each deck. Each column contains `id`, `title`,
`connectionId`, and a platform-specific `source`.
| Platform | Source kind | Conditions |
| --- | --- | --- |
| Twitter | `search` | Native `query`, `product` (`Top`/`Latest`), `following` |
| Twitter | `user` | `target`: handle or X/Twitter profile URL |
| Twitter | `list` | `target`: numeric list ID or X/Twitter list URL |
| Mastodon | `search` | `query`; results depend on the instance's search configuration |
| Mastodon | `user` | `target`: account handle or the connected instance's numeric account ID |
| Mastodon | `list` | `target`: numeric list ID for the connected account |
| Mastodon | `hashtag` | `target`: tag without `#`, containing letters, numbers, or underscores |
| Platform | Source kind | Conditions |
| -------- | ----------- | ----------------------------------------------------------------------- |
| Twitter | `search` | Native `query`, `product` (`Top`/`Latest`), `following` |
| Twitter | `user` | `target`: handle or X/Twitter profile URL |
| Twitter | `list` | `target`: numeric list ID or X/Twitter list URL |
| Mastodon | `search` | `query`; results depend on the instance's search configuration |
| Mastodon | `user` | `target`: account handle or the connected instance's numeric account ID |
| Mastodon | `list` | `target`: numeric list ID for the connected account |
| Mastodon | `hashtag` | `target`: tag without `#`, containing letters, numbers, or underscores |
`connectionId` is the account binding, distinct from a deck's name. Twitter
connections resolve to profiles discovered from the configured relay.
@@ -76,9 +76,9 @@ in a transaction. A stale revision is rejected. Save failures remain visible
and do not report unsaved edits as persisted.
Saved decks refresh on focus and every five seconds while visible. Refresh is
held while an editor is open. Active selection is a local preference under
`twitter-lite-active-deck`; selecting a deck does not switch another device's
view. Deleting the active deck selects a remaining one. When none remain, the
held while an editor is open. The `deck` URL parameter selects the active view;
selecting a deck does not switch another device's view. Browser history restores
selection, and saved decks can be opened from a copied URL. Deleting the active deck selects a remaining one. When none remain, the
app opens an empty temporary view.
WebMCP-created views are temporary by default. Temporary copies of saved decks
@@ -139,3 +139,10 @@ uses an isolated SQLite database and mock relay through real server functions
to exercise shared decks across browser contexts, editing, persistence,
pagination, and native WebMCP. Deterministic automated tests do not require
live SNS credentials.
## Conversation execution
The [research runtime](research-runtime.md) uses AI SDK and the Codex app-server
provider. Conversations are selected by URL, with per-conversation background
execution and reconnectable SSE. Opening New chat or another saved conversation
does not stop an existing turn or change another browser's selection.
docs/research-runtime.md
+54
View File
@@ -0,0 +1,54 @@
# Research runtime
Research uses Vercel AI SDK 7 `streamText` with
`ai-sdk-provider-codex-cli` 2.3's app-server provider. The provider owns its local
stdio child, thread resumption, protocol decoding, and abort handling. The app
owns accepted requests, account-scoped tools, snapshots, and report validation.
The separate WebSocket launcher is removed.
## Tools
The provider does not register AI SDK `tools` as Codex dynamic tools. Existing
research definitions and executors are exposed with `createSdkMcpServer` instead.
Validation, connected-account scope, request budgets, evidence tracking, and
read-only SNS behavior remain in `agent-tools.server.ts`. Provider execution is
one agent turn; there is no second AI SDK tool loop replaying those calls.
Resumed threads that already registered dynamic tools route their calls through
`onDynamicToolCall` to the same validated executor.
## Lifecycle and selection
The background runner is keyed by conversation ID. Acceptance saves the user
message before starting the model. A request ID identifies a new conversation
and deduplicates retries within the process. Overlapping turns of the same
conversation are rejected, while different conversations can run independently.
The URL selects a conversation. GET status and SSE subscribe to that ID;
selecting history never starts or cancels a turn. Reconnection sends the latest
complete snapshot. Closing a browser or changing its URL only detaches a viewer.
Only Stop, the execution timeout, or a server-side failure aborts generation.
Authentication expiry or logout closes the viewer's stream.
SQLite retains messages, thread IDs, generated decks, and citations. It no
longer stores a global active conversation pointer. Browser views cannot change
another browser's selection. After a backend restart, unfinished conversations
are marked interrupted and require an explicit follow-up. No turn is replayed
automatically. Reports remain files under the configured report root.
## Streaming presentation
The model stream uses `smoothStream` with 15 ms pacing and the supplied chunking
pattern `/[\u3040-\u309F\u30A0-\u30FF]|\S+\s+/`. The trailing buffer flushes at the
text-end event. The existing snapshot SSE remains the UI's reconnectable data
channel; an HTTP request's lifetime never owns the model execution.
Streamdown displays partial Markdown and animates only the latest live answer.
Completed messages render immediately. Reduced-motion preferences disable the
animation. HTML is skipped, unsafe link protocols are rejected, images remain
links, and retrieved post citations keep their deck navigation behavior.
`@shadcn/helpers/ai-sdk` provides deterministic streaming fixtures for UI tests.
References: [Codex provider](https://github.com/ben-vargas/ai-sdk-provider-codex-cli),
[AI SDK smoothing](https://ai-sdk.dev/docs/reference/ai-sdk-core/smooth-stream),
[Streamdown animation](https://streamdown.ai/docs/animation),
[shadcn AI SDK helpers](https://ui.shadcn.com/docs/helpers/ai-sdk).
docs/vitals-preview.md
+4
View File
@@ -16,3 +16,7 @@ as the anatomy illustration source and clipped in CSS. Source media:
https://pbs.twimg.com/media/HTE4JWFWUAAnPBt.jpg
The rest of the interface is HTML/CSS with SVG data charts. Reference-specific
colors are scoped to `--vital-*` tokens in `vitals.css`.
Sections, tabs, history, chart filters, and detail dialogs are URL state and
restore on reload or browser Back/Forward. Simulated climate control values
remain local; opening a URL never issues a device command.
docs/webmcp-prototype.md
+9 -9
View File
@@ -13,15 +13,15 @@ operations use the same access controls and persistence rules as the UI.
## Workspace tools
| Tool | Input | Behavior |
| --- | --- | --- |
| `list_connections` | `{}` | Return connection IDs, platforms, origins, account IDs, display names, and states; never credentials |
| `list_decks` | `{}` | Return saved decks and this tab's temporary views, `activeDeckId`, and `storageError` |
| `get_deck` | Optional `deckId` | Read the named or active view, including `persisted` and saved `revision` |
| `set_deck` | Optional `deckId` and `expectedRevision`, required `title` and `columns` | Without an ID, create a temporary view; with an ID, replace and activate that existing view |
| `save_deck` | `deckId` | Explicitly persist a temporary view; an already saved deck is unchanged |
| `select_deck` | `deckId` | Activate a view; selection remains device-local |
| `delete_deck` | `deckId`, optional `expectedRevision` | Discard a temporary view or delete a saved deck for all devices |
| Tool | Input | Behavior |
| ------------------ | ------------------------------------------------------------------------ | ---------------------------------------------------------------------------------------------------- |
| `list_connections` | `{}` | Return connection IDs, platforms, origins, account IDs, display names, and states; never credentials |
| `list_decks` | `{}` | Return saved decks and this tab's temporary views, `activeDeckId`, and `storageError` |
| `get_deck` | Optional `deckId` | Read the named or active view, including `persisted` and saved `revision` |
| `set_deck` | Optional `deckId` and `expectedRevision`, required `title` and `columns` | Without an ID, create a temporary view; with an ID, replace and activate that existing view |
| `save_deck` | `deckId` | Explicitly persist a temporary view; an already saved deck is unchanged |
| `select_deck` | `deckId` | Activate a view and update the URL; other tabs remain unchanged |
| `delete_deck` | `deckId`, optional `expectedRevision` | Discard a temporary view or delete a saved deck for all devices |
Read tools return the workspace's current client snapshot, not a fresh server
request. Saved definitions refresh on focus and while visible, except during