@youtyan/code-viewer 0.2.10 → 0.3.0

package/README.md

@@ -14,6 +14,12 @@ Requires Node.js 20 or newer when installed from npm. Development uses
   pills" that copy `@path#start-end` for AI agents.
 - Browse commit history per branch and open any commit's changed files and
   diff, with shareable `/history?ref=<branch>&commit=<sha>` links.
+- Open per-file Blame and History tabs on a file detail page (GitHub-style):
+  Blame groups consecutive lines from the same commit with an Older→Newer
+  colour bar and lets you jump to the originating commit; History embeds the
+  same commit list and diff renderer used by `/history` inside the file's
+  tab shell, filtered to that path. Both tabs keep the Repository sidebar
+  visible.
 - Open files directly from the repository or diff view, including large
   generated files (virtualized source viewer with copy/open-full-view).
 - Preview Markdown with a table of contents, task lists, Mermaid diagrams
@@ -27,8 +33,9 @@ Requires Node.js 20 or newer when installed from npm. Development uses
   viewer.
 - Browse SQLite, PostgreSQL, MySQL, Redis, Elasticsearch, and S3-compatible
   object storage (MinIO, LocalStack) with a built-in datastore viewer.
-- Read the built-in Help page for repository browsing, diffs, annotations,
-  the datastore viewer, agent skills, and shortcuts.
+- Read the built-in Help page for getting started, the `.code-viewer/`
+  project files, AI annotations, datastores, the agent skill, and
+  keybindings.
 - Open repository folders (and parent folders of files) in the OS file
   manager, create folders, and trash/restore files from localhost-only
   actions.
@@ -110,6 +117,15 @@ resolved inside the repository, code blocks are highlighted with Shiki, and
 Mermaid diagrams are rendered lazily in the browser (click any diagram to
 open it in a lightbox).
 
+A file detail page lays out four tabs — **Preview**, **Code**, **Blame**,
+**History** — modelled after the GitHub file view. `Code` is the default and
+`?preview=1` opts in to the Markdown / media preview. `Blame` and `History`
+each have their own canonical URL (`view=blame`, `view=history`), so deep
+links and the browser back/forward stay in sync. The Blame tab reuses the
+source view's row component, so line numbers, drag-selection of `line=`
+ranges, syntax highlighting and the Viewer Settings code font size all match
+the Code tab.
+
 Very large text files use a virtualized source viewer. Only visible rows are
 rendered, and the page includes controls to copy the full file or reopen it in
 the full non-virtual view.
@@ -226,7 +242,9 @@ Open Datastores in the global navigation to access:
   whether each entry came from the browser or the CLI.
 - **ER diagram** — auto-generated entity-relationship diagram showing
   foreign-key relationships between tables.
-- **Schema view** — table columns, indexes, foreign keys, triggers, and DDL.
+- **Schema view** — table columns (with column comments when the database
+  defines them), indexes, foreign keys, triggers, and DDL. Tables themselves
+  surface a comment column on the database table list.
 - **Global search** — full-text search across all tables and text columns of
   a database.
 - **Snapshots and diffs** — take point-in-time snapshots of selected tables
@@ -239,9 +257,10 @@ Open Datastores in the global navigation to access:
 
 ### CLI
 
-AI agents can read data and query history from the command line. The current
-CLI ships `exec`, `list`, and `clear`; search, snapshot, and diff operations
-are performed from the browser UI.
+AI agents can run read-only queries, search content across tables, and
+capture snapshots / diffs from the command line. The same operations are
+mirrored under the Datastores tab in the browser UI, and every CLI result is
+written to the per-repository history visible in the browser.
 
 ```sh
 code-viewer query exec --db data.db --sql "SELECT * FROM users LIMIT 10" \
@@ -252,6 +271,22 @@ code-viewer query exec --db app.db --sql "SELECT count(*) FROM orders" \
 
 code-viewer query list --db app.db --json
 code-viewer query clear --db app.db
+
+code-viewer query search --db app.db --term "john@example.com" \
+    --tables users,orders --include-non-text --max-hits 20
+
+code-viewer query snapshot create --db app.db --tables users,orders \
+    --note "Before user registration test"
+code-viewer query snapshot list --db app.db --json
+code-viewer query snapshot note --id snap-abc123 --note "Updated context"
+code-viewer query snapshot delete --id snap-abc123
+
+code-viewer query diff create --before snap-abc123 --after snap-def456 \
+    --note "User registration test"
+code-viewer query diff tables --id diff-xyz789
+code-viewer query diff rows --id diff-xyz789 --table users --type inserted
+code-viewer query diff list --db app.db --json
+code-viewer query diff delete --id diff-xyz789
 ```
 
 `code-viewer query --help` shows all command syntax. `code-viewer query
@@ -307,9 +342,9 @@ with `--include-non-text` and `--run-search`. For AI agents,
 
 Annotations are grouped into sessions and persisted in
 `.code-viewer/annotations.json` at the repository root, so the walkthrough
-survives reloads and server restarts. The `.code-viewer` directory is
-tool-internal: it never shows up in the file tree, searches, or diffs. Add it
-to `.gitignore` if you do not want to share annotations through git.
+survives reloads and server restarts. See **Uploads and Scope Settings**
+above for how `.code-viewer/` is treated by the viewer and how to opt out of
+sharing it through git.
 
 In the browser, the annotation icon in the header opens the side panel. The
 current explanation is rendered at the top of the panel while the code stays
@@ -341,9 +376,11 @@ write concise Markdown explanations, and how to install the bundled agent skill.
 
 ### Agent Skill
 
-The package bundles [Agent Skills](https://agentskills.io) (the SKILL.md
-open standard) that teach AI coding agents when and how to use
-`annotate`, `query`, and `snapshot`. Install them into the current project:
+The package bundles three [Agent Skills](https://agentskills.io)
+(the SKILL.md open standard) — `code-viewer-annotate`, `code-viewer-query`,
+and `code-viewer-snapshot` — that teach AI coding agents when and how to use
+`annotate`, read-only `query`, and snapshot / diff workflows. A single
+`skill install` copies all three into the selected agent directories:
 
 ```sh
 npx -y @youtyan/code-viewer skill install                       # Claude Code (.claude/skills/)