Repository navigation
Conversation
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
66b9b01 to
8920c57
Compare
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 66b9b0178e
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| LEFT JOIN message reply_msg ON msg.reply_to_message_id = reply_msg.platform_message_id | ||
| LEFT JOIN member reply_m ON reply_msg.sender_id = reply_m.id` | ||
| LEFT JOIN member reply_m ON reply_msg.sender_id = reply_m.id | ||
| LEFT JOIN message_mark mm ON mm.message_id = msg.id` |
There was a problem hiding this comment.
Because FULL_MSG_SELECT is also used by searchMessagesByKeywords, the default packages/tools/src/providers/core-data-provider.test.ts fixture—which creates only member and message—now throws no such table: message_mark on its first provider.searchMessages() call. The two core query fixtures were updated but this default-suite fixture was not, so add the table there (and to any other full-message fixtures) or make the shared select tolerate schemas without marks.
Useful? React with 👍 / 👎.
| mark: | ||
| row.markUpdatedAt == null | ||
| ? null | ||
| : { |
There was a problem hiding this comment.
In the CLI/MCP/HTTP-tool path, CoreDataProvider.getSearchMessageContext and getMessageContext return these MappedMessage objects directly as RawMessage, and the tool handlers JSON-serialize them. Once a user has added a mark, this new field sends its categories and free-form comment to the configured model (and sends mark:null for every unmarked context row) even though RawMessage exposes no such field; sanitize those provider returns or keep mark mapping viewer-specific.
Useful? React with 👍 / 👎.
| function applyMessageMark(messageId: number, mark: ChatRecordMessage['mark']) { | ||
| messages.value = messages.value.map((message) => (message.id === messageId ? { ...message, mark } : message)) |
There was a problem hiding this comment.
When query.markCategories, markColor, or markComments is active, changing or removing the matched mark emits update-mark, but this handler merely updates the row in place. A message that no longer satisfies the active filter therefore remains visible, with stale loaded count and pagination, until the user reapplies the filter; rerun the filtered search or remove and refill the row after save.
Useful? React with 👍 / 👎.
| version: 11, | ||
| description: 'Add user-defined message marks', |
There was a problem hiding this comment.
When Desktop opens a v10 database, getPendingMigrationInfos(10) synthesizes database.migrationV11Desc and database.migrationV11Message, but all four apps/desktop/main/i18n/locales/*.ts catalogs stop at V10, so the migration modal renders the literal internal key instead of an explanation; the existing migration-version assertion also still stops at 10. Add the V11 strings and extend that migration test.
AGENTS.md reference: AGENTS.md:L65-L65
Useful? React with 👍 / 👎.
8920c57 to
c3dbf1e
Compare
|
Codex Review: Something went wrong. Try again later by commenting “@codex review”. ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
If Codex has suggestions, it will comment; otherwise it will react with 👍. Codex can also answer questions or update the PR. Try commenting "@codex address that feedback". |
1 similar comment
|
Codex Review: Something went wrong. Try again later by commenting “@codex review”. ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
If Codex has suggestions, it will comment; otherwise it will react with 👍. Codex can also answer questions or update the PR. Try commenting "@codex address that feedback". |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: c3dbf1e6bd
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| try { | ||
| const query = toRaw(props.query) | ||
| const { filter, senderId, keywords } = buildFilterParams(query) | ||
| const { filter, senderId, keywords, markFilter } = buildFilterParams(query) |
There was a problem hiding this comment.
When a mark filter is active and the user clicks a session in the timeline, ChatRecordWorkspace.handleSessionSelect preserves that filter while adding scrollToMessageId. This makes the target-ID branch run, but the computed markFilter is not passed to either getMessagesBefore or getMessagesAfter, so a mark-only query suddenly displays unmarked surrounding messages while the filter controls still show the filter as active; pass the mark criteria through these centered pagination calls or explicitly clear them when selecting a timeline session.
Useful? React with 👍 / 👎.
There was a problem hiding this comment.
Fixed in 3a79f8d. Target-centered initial loads and subsequent before/after pagination now receive the active mark filter, with regression coverage for filtered pages around a target.
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 3a79f8dcea
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| categories TEXT NOT NULL DEFAULT '[]', | ||
| color TEXT NOT NULL DEFAULT '', | ||
| comment TEXT NOT NULL DEFAULT '', | ||
| updated_at INTEGER NOT NULL, |
There was a problem hiding this comment.
In Agent/MCP runs where the default raw-SQL tools are available, this new table is returned by getDatabaseSchema(), while SHARED_TOOLS exposes both get_schema and unrestricted read-only execute_sql (packages/tools/src/registry.ts:52-75), so a model can execute SELECT categories, comment FROM message_mark and send these notes to the configured provider. Fresh evidence beyond the earlier context-return issue is the added test at packages/tools/src/providers/core-data-provider.test.ts:134-143, which explicitly calls marks viewer-only, while only the RawMessage mapping was sanitized. Hide and reject this table in both the CoreDataProvider and WorkerDataProvider schema/raw-SQL paths, or store marks outside the AI-visible database.
Useful? React with 👍 / 👎.
Chat records currently provide the original message content but offer few ways to organize or annotate individual messages. Member aliases also have to be edited from the separate member management view.
This PR adds an inline alias editor beside the sender name and keeps alias changes synchronized with member management. It also adds a copy action for the original content of a single message.
Messages can now be marked with any combination of categories, a color, and a free-form comment. Marks are displayed separately from the original message bubble and can be filtered by category, color, or comment text. Long comments remain readable without taking over the record view, and the controls work with narrow or resized record panels.
The mark data is stored separately from the message body. This keeps imported messages intact while establishing a clear place for information associated with a message but not originally part of it.
This separation may later support richer message preprocessing—for example, audio transcription and translation, image captions or classifications, and document ETL—without converting the original media into replacement text. That extension is intentionally not implemented here. It still needs a design for preserving media, namespaced and versioned preprocessing metadata, provenance, generated-result lifecycle, and how search or AI features consume those results.
The current PR stays focused on human organization: aliases, copying, categories, colors, and comments. It does not change imported message content or add media/AI processing. Schema migration, filtering, localization, keyboard and screen-reader state, virtual-list synchronization, and Desktop/CLI Web shared service paths are included in the implementation.