docs: introduce the 0.65.0 browser UI (#1066)

Roborev 0.65.0 adds a second primary review interface and several
operator-facing behaviors, but the release documentation still led with
the previous terminal UI and did not tell the complete release story.

This makes the native browser application the primary visual on the
README and documentation home page, expands the browser and TUI workflow
guides, and completes the 0.65.0 changelog with contributor
acknowledgements. It also corrects the Homebrew tap path and
source-build requirements.

![Roborev browser application showing the review queue and an open
review](https://raw.githubusercontent.com/kenn-io/roborev/docs-generated-assets/web-ui.png)

The browser image is reproducible rather than hand-edited. The
Dockerized capture builds the branch application and drives it at a
fixed viewport with Playwright. Its reduced database is limited to
completed reviews from the canonical public Roborev repository and
sanitizes local paths, credential-shaped values, and configured private
terms before the daemon starts.

This PR documents existing 0.65.0 behavior. It does not change the
daemon, browser application, or review lifecycle; the executable changes
are limited to documentation asset generation and its regression
coverage.

---------

Co-authored-by: Codex <noreply@openai.com>

Author: wesm

README.md

@@ -14,7 +14,7 @@ background, reviews every commit as agents write code, and surfaces
 issues in seconds -- before they compound. Pull code reviews into
 your agentic loop while context is fresh.
 
-![roborev TUI](https://roborev.io/assets/generated/tui-hero.svg)
+![Roborev browser application](https://roborev.io/assets/generated/web-ui.png)
 
 ## How It Works
 
@@ -78,8 +78,10 @@ You can also choose the exact binary path with
   Reviews are orchestrated on your machine using the coding agents
   you already have configured.
 - **Interactive TUI** - Real-time review queue with vim-style navigation.
+  Large terminals show the queue and selected review in a split-screen layout.
 - **Native Browser UI** - Browse, filter, inspect, comment on, close, cancel,
-  and rerun reviews from the application embedded in the local daemon.
+  and rerun reviews from the application embedded in the local daemon, then
+  explore cost, latency, reliability, and outcomes in Analytics.
 - **Review Verification** - `roborev compact` verifies findings against
   current code, filters false positives, and consolidates related issues
   into a single review.
@@ -153,7 +155,7 @@ curl -fsSL https://roborev.io/install.sh | bash
 
 **Homebrew (macOS / Linux):**
 ```bash
-brew install roborev-dev/tap/roborev
+brew install kenn-io/tap/roborev
 ```
 
 **Windows (PowerShell):**

docs/README.md

@@ -27,7 +27,7 @@ are ignored local outputs.
 
 - `docs-assets`: curated static media, including logos, favicons, Open Graph
   images, diagrams, agent icons, and manually captured integration images.
-- `docs-generated-assets`: generated CLI and TUI screenshots.
+- `docs-generated-assets`: generated browser UI, CLI, and TUI screenshots.
 
 Docs pages should reference media through:
 

docs/assets/hydrate-assets.sh

@@ -41,6 +41,7 @@ static_assets=(
 )
 
 generated_assets=(
+  "web-ui.png"
   "tui-hero.svg"
   "tui-queue.svg"
   "tui-review.svg"

docs/changelog.md

@@ -5,24 +5,39 @@ description: Release history for roborev
 
 All notable changes to roborev, grouped by minor release.
 
-## Unreleased
+## 0.65.0
+
+<small>2026-08-16</small>
 
 **New features**
 
-- Reasoning accepts exact `low`, `medium`, `high`, `xhigh`, and `max` values and
-    forwards them unchanged to agents that support each tier. The legacy `fast`,
-    `standard`, `thorough`, and `maximum` presets remain compatible.
 - Roborev now embeds an authenticated native browser review workspace on a
     separate daemon listener. `roborev ui [job-id]` opens the job list or a
     review deep link with filtering, sorting, panel detail, Markdown output,
     comments, logs, prompts, keyboard controls, and review/job mutations. Local
     use bootstraps automatically; remote HTTPS deployments exchange a token for
-    a browser session without exposing the private CLI listener.
+    a browser session without exposing the private CLI listener. Analytics
+    reports review volume, outcomes, errors, latency, agent attempts, estimated
+    cost, and pricing coverage. See [Browser UI](/web-ui/).
+- The TUI automatically shows a kata-style split-screen review workspace on
+    terminals at least 140 columns by 36 rows. The queue drives a live detail
+    pane for completed, running, failed, and queued jobs; `L` toggles layouts,
+    and smaller terminals fall back to the existing stacked view. See
+    [Split-Screen Review](/integrations/tui/#split-screen-review).
+- Reasoning accepts exact `low`, `medium`, `high`, `xhigh`, and `max` values and
+    forwards them unchanged to agents that support each tier. The legacy `fast`,
+    `standard`, `thorough`, and `maximum` presets remain compatible. See
+    [Reasoning Levels](/configuration/#reasoning-levels).
 - Pi agents accept global `[agent.pi] launch_args`, passed as tokenized
     arguments to every Pi invocation before roborev-managed workflow and safety
     options. This allows isolated classifier jobs to load extension-defined
     model providers explicitly while retaining `--no-extensions` discovery
-    isolation.
+    isolation. See [Pi Classifier Options](/configuration/#pi-classifier-options).
+- `roborev export ci-costs` exports job-level costs for eligible CI attempts,
+    including terminal retries that panel summaries cannot retain. Stable
+    cursors, overlapping-window refreshes, null-versus-zero pricing, and a
+    legacy backfill mode support incremental downstream accounting. See
+    [Exporting CI Costs](/commands/#exporting-ci-costs).
 
 **Improvements**
 
@@ -35,6 +50,25 @@ All notable changes to roborev, grouped by minor release.
     repository, worktree, branch, and expiry, while the TUI shows a contextual
     snooze badge for an exactly filtered checkout. See
     [Snoozing Reminders](/agent-hook/#snoozing-reminders).
+- Daemon restarts stop new claims and wait for active reviews, hooks, and final
+    sync to finish before starting the replacement process. The CLI reports the
+    wait instead of force-killing work after a fixed timeout.
+- `roborev review --branch` now resolves the repository and range from the
+    linked worktree that invoked it, preventing shared or stale Git
+    configuration from redirecting the review to a sibling checkout. See
+    [Git Worktrees](/guides/repository-management/#git-worktrees).
+- Agent Hook installation and runtime handling now use kit's shared coding-agent
+    profiles. One workflow supports Claude Code, Codex, Copilot CLI, Cursor,
+    Factory Droid, Gemini CLI, Hermes, and Qwen, while Roborev retains its Grok
+    integration. Existing Codex, Claude, and Droid registrations can be upgraded
+    in place. See [Agent Hook](/agent-hook/).
+- The Homebrew tap now discovers official Roborev releases and owns formula
+    updates, removing duplicate publisher logic and the cross-repository release
+    credential. See
+    [Homebrew installation](/installation/#homebrew-macos-linux).
+- Development, CI, release, and screenshot builds now require Go 1.26.6. Go,
+    JavaScript, documentation, and GitHub Actions dependencies were refreshed,
+    including fixes for known Go toolchain and `go-git` vulnerabilities.
 
 **Bug fixes**
 
@@ -44,13 +78,31 @@ All notable changes to roborev, grouped by minor release.
     rerun jobs. Session expiry and logout now close active browser streams.
 - Persisted job output is shown only when its log belongs to the current
     attempt, so a failed rerun cannot fall back to an older attempt's output.
-- Daemon restarts now stop new job claims and wait indefinitely for running
-    reviews and worker finalization instead of force-killing the daemon after a
-    short graceful-shutdown window.
+- Daemon discovery now preserves sandbox permission failures instead of treating
+    an unreachable loopback or Unix socket as proof that no daemon exists.
+    Clients can use the daemon's private Unix-socket fallback without starting a
+    competing process. See [Daemon & Hooks](/commands/#daemon-hooks).
+- Fresh agent sessions now receive a short, bounded agentsview usage-indexing
+    retry before Roborev falls back to job-log token data, reducing permanently
+    missing cost estimates. See [Token Usage](/commands/#token-usage).
 - The Codex `maximum` preset now requests literal `max` for explicit GPT-5.6
     `sol`, `terra`, and `luna` models. Older, default, and unknown models retain
     the compatible `xhigh` mapping, while exact `xhigh` remains distinct.
 
+**Acknowledgements**
+
+- Thanks to [Wes McKinney](https://github.com/wesm) for the native browser
+    application, job-level CI cost export, linked-worktree review fix,
+    usage-indexing retry, and release updates.
+- Thanks to [Graham Wheeler](https://github.com/gramster) for the TUI
+    split-screen review workspace.
+- Thanks to [Phillip Cloud](https://github.com/cpcloud) for exact
+    reasoning-effort tiers across supported agents.
+- Thanks to [Marius van Niekerk](https://github.com/mariusvniekerk) for unified
+    Agent Hook profiles, sandbox-safe daemon discovery, visible snooze status,
+    graceful daemon restarts, Pi launch arguments, and the Go 1.26.6 security
+    upgrade.
+
 ______________________________________________________________________
 
 ## 0.64.0

docs/development.md

@@ -213,7 +213,14 @@ branches before builds:
 
 - `docs-assets` contains curated static media such as logos, favicons, diagrams,
     Open Graph images, and integration screenshots.
-- `docs-generated-assets` contains generated CLI and TUI screenshots.
+- `docs-generated-assets` contains generated browser, CLI, and TUI screenshots.
+
+The screenshot pipeline reads the normal review database only through a
+read-only SQLite connection, extracts completed clean reviews from the canonical
+public Roborev repository, sanitizes copied content, and runs the branch binary
+against that reduced database inside Docker. See the
+[screenshot pipeline guide](https://github.com/kenn-io/roborev/blob/main/docs/screenshots/README.md)
+for the capture and asset publication workflow.
 
 Install the docs toolchain and build the site:
 

docs/guides/repository-management.md

@@ -145,6 +145,9 @@ When running commands from a worktree:
     active worktree branch
 - `roborev fix --open` filters to reviews reachable from the current worktree's
     HEAD
+- `roborev review --branch` resolves the repository root, branch, and commit
+    range from the linked worktree that invoked it, even if shared Git
+    configuration points at a sibling checkout
 - `refine` correctly finds and addresses reviews for commits in any worktree
 
 ### Detached HEAD worktrees

docs/index.md

@@ -28,10 +28,32 @@ coding agent so the write -> review -> fix loop runs hands-off.
 
 [Set up automation ->](automation/post-commit-reviews.md)
 
+## Browser UI
+
+Run `roborev ui` to open the native browser application. Browse and filter the
+live review queue, inspect review output, prompts, and logs, manage jobs and
+comments, or switch to Analytics to explore cost, latency, reliability, and
+review outcomes. The application runs from the local daemon and uses the same
+SQLite history as the CLI and terminal UI.
+
 <figure class="hero-shot" data-lightbox>
-  <img src="/assets/generated/tui-hero.svg" alt="roborev TUI queue view" loading="eager">
+  <img src="/assets/generated/web-ui.png" alt="Roborev browser application showing the review queue and an open review" loading="eager">
 </figure>
 
+[Explore the browser UI ->](web-ui.md)
+
+## Terminal UI
+
+Prefer to stay in the terminal? `roborev tui` provides the same live review
+ledger with vim-style navigation. On a large terminal, its split-screen layout
+keeps the queue and selected review visible together.
+
+<figure class="hero-shot" data-lightbox>
+  <img src="/assets/generated/tui-hero.svg" alt="Roborev TUI split-screen review view" loading="lazy">
+</figure>
+
+[Explore the terminal UI ->](integrations/tui.md)
+
 <div class="agent-matrix">
   <a href="/agents/"><img src="/assets/static/agents/codex.svg" alt="Codex" data-agent="codex" width="1604" height="719" /></a>
   <a href="/agents/"><img src="/assets/static/agents/claude-code.svg" alt="Claude Code" data-agent="claude-code" width="696" height="95" /></a>
@@ -57,6 +79,7 @@ Then from within your git repositories:
 roborev init          # Install post-commit hook
 # do some work, generate commits
 roborev tui           # Browse reviews in the terminal UI
+roborev ui            # Or browse reviews in the native browser UI
 ```
 
 For Windows, see the

docs/installation.md

@@ -50,6 +50,10 @@ brew install roborev
 This also works on Linux with
 [Linuxbrew](https://docs.brew.sh/Homebrew-on-Linux).
 
+The `kenn-io/tap` repository watches official Roborev releases and manages
+formula updates from their published checksums. Roborev release jobs do not need
+a cross-repository credential to update the formula.
+
 ## Linux Packages: DEB and RPM
 
 Starting with 0.57.0, GitHub releases include `.deb` and `.rpm` packages for
@@ -89,15 +93,17 @@ need the browser application.
 
 ## Build from Source
 
+Building Roborev requires Go 1.26.6 or newer. The embedded browser application
+also requires Bun 1.3.14.
+
 ```bash
 git clone https://github.com/kenn-io/roborev
 cd roborev
 make install
 ```
 
-The `make install` target requires Bun 1.3.14, builds and validates the browser
-application, and embeds it alongside version information (for example,
-`v0.7.0-5-gabcdef`).
+The `make install` target builds and validates the browser application and
+embeds it alongside version information (for example, `v0.7.0-5-gabcdef`).
 
 For quick iteration during development:
 

docs/integrations/tui.md

@@ -60,6 +60,7 @@ branch.
 | `f` | Open filter (repo/branch tree) |
 | `b` | Open filter with branches expanded |
 | `h` | Toggle hide closed/failed/canceled |
+| `L` | Toggle split-screen or stacked layout |
 | `D` | Toggle distraction-free mode |
 | `P` | Pause or resume queue processing |
 | `g` | Jump to top of queue |
@@ -73,6 +74,30 @@ shows live progress while member reviewers run and a compact member summary
 after completion. Expand the row to inspect individual member jobs. Close,
 cancel, rerun, or fix the panel from the parent row, not from a member row.
 
+## Split-Screen Review
+
+At 140 columns by 36 rows or larger, the TUI automatically places the queue in a
+left pane and the selected job in a right detail pane. Moving through the queue
+updates the detail pane after a short debounce, so holding `j` or `k` does not
+issue a request for every intermediate row.
+
+The detail pane adapts to job state: completed jobs show the rendered review,
+running jobs show a live log tail, failed jobs show the error, and queued jobs
+show their metadata. When a running job finishes, its log is replaced by the
+review automatically.
+
+Press `Tab` to focus the detail pane and `Esc` to return to the queue. Review
+actions and scrolling work in the focused detail pane. `Enter` is intentionally
+a no-op while the split queue is focused because the detail already follows the
+selection. Mouse clicks focus either pane, and the wheel scrolls the pane below
+the pointer.
+
+Press `L` to lock the stacked or split preference for the TUI session. A split
+preference falls back to stacked when the terminal is too small and returns when
+it grows again. Prompt, comment, help, log, and task screens remain full-screen.
+The queue also drops lower-priority columns as needed to preserve the most
+useful fields in its narrower pane.
+
 ## Queue Pause
 
 Press `P` in the queue to pause or resume daemon queue processing. Pausing is

docs/screenshots/Dockerfile

@@ -1,3 +1,18 @@
+FROM oven/bun:1.3.14-alpine AS web-builder
+
+RUN apk add --no-cache git
+
+WORKDIR /src
+
+COPY package.json bun.lock ./
+COPY web/package.json web/package.json
+COPY packages/roborev-ui/package.json packages/roborev-ui/package.json
+RUN bun install --frozen-lockfile
+
+COPY web web
+COPY packages packages
+RUN bun run web:build
+
 FROM golang:1.26.6-alpine AS builder
 
 WORKDIR /src
@@ -8,15 +23,18 @@ RUN go mod download
 
 # Copy source and build
 COPY . .
+COPY --from=web-builder /src/web/dist ./internal/web/dist
 ARG VERSION=dev
-RUN go build \
+RUN CGO_ENABLED=0 go build \
     -ldflags="-X go.kenn.io/roborev/internal/version.Version=${VERSION}" \
     -o /roborev ./cmd/roborev
 
-FROM alpine:3.19
+FROM node:22-bookworm
 
 # Install dependencies for freeze + tmux screenshots
-RUN apk add --no-cache ca-certificates git bash tmux
+RUN apt-get update && \
+    apt-get install -y --no-install-recommends ca-certificates git tmux wget && \
+    rm -rf /var/lib/apt/lists/*
 
 # Install freeze for SVG screenshots (detect architecture)
 RUN ARCH=$(uname -m) && \
@@ -28,6 +46,11 @@ RUN ARCH=$(uname -m) && \
 
 COPY --from=builder /roborev /usr/local/bin/roborev
 
+WORKDIR /screenshots
+COPY docs/screenshots/package.json docs/screenshots/package-lock.json ./
+RUN npm ci && npx playwright install --with-deps chromium
+COPY docs/screenshots/ ./
+
 # Create config pointing to local daemon
 RUN mkdir -p /root/.roborev && \
     echo 'server_addr = "127.0.0.1:7373"' > /root/.roborev/config.toml
@@ -36,4 +59,4 @@ RUN mkdir -p /root/.roborev && \
 RUN printf '#!/bin/sh\ncat > /dev/null\n' > /usr/local/bin/xclip && \
     chmod +x /usr/local/bin/xclip
 
-WORKDIR /screenshots
+ENTRYPOINT ["/bin/bash"]