benchflow 0.7.2.dev1987__tar.gz → 0.7.2.dev1994__tar.gz

{benchflow-0.7.2.dev1987 → benchflow-0.7.2.dev1994}/CHANGELOG.md

@@ -3,6 +3,17 @@
 ## [Unreleased]
 
 ### Added
+- **The trajectory viewer can collect the eval-prize confirmation in the
+  browser.** `bench eval view PATH --confirm` renders the normal page plus a
+  sticky site-styled bottom bar ("Submit this trajectory to the BenchFlow
+  eval prize?") with **Approve & submit** / **Not this one** buttons. A click
+  POSTs to `/decision`; the server prints a machine-readable
+  `DECISION: approved` or `DECISION: rejected` line to stdout, shuts down,
+  and the CLI exits `0` on approve and `3` on reject (non-1/2 so rejection
+  never collides with error or usage exits). Without the flag, behavior is
+  unchanged (no bar, no endpoint, Ctrl+C to stop). The
+  `benchflow-traj-upload` skill now prefers the button flow and falls back
+  to chat confirmation on CLIs older than 0.7.2.
 - **Trajectory uploads are tagged with the session's repository by
   default.** Unless `--source-id` is given, `bench traj upload` reads the
   session's recorded working directory (Claude `cwd` events, Codex
@@ -14,6 +25,17 @@
   undetectable repos fall back silently to the path-derived source id.
 
 ### Changed
+- **Trajectory viewer tool calls are color-coded and backgrounds are light.**
+  Each tool kind now gets a muted GitHub-label-style accent on its name pill
+  and a left border strip on the card: shell/exec → amber, write/edit → blue,
+  read → teal, agent/task/skill → purple, web/search/fetch → cyan, everything
+  else → neutral gray. Tool arguments and tool outputs render on light
+  surfaces (`#f5f5f5` / white with dark ink) instead of near-black blocks;
+  the dark `#141414` terminal treatment is reserved for shell-command output
+  only, and the ink-black result card stays as the deliberate bento-ink
+  accent. Presentation only (CSS classes + a tool-name→accent mapping);
+  content strings and behavior are unchanged. Applies to all three viewer
+  templates, which share one stylesheet since #1019.
 - **The contributor prompt now tells the agent to upgrade BenchFlow first.**
   `CONTRIBUTOR_PROMPT` (kept in sync in `README.md`, `docs/traj-upload.md`,
   and the `benchflow-traj-upload` skill evals) is a three-line block that

{benchflow-0.7.2.dev1987 → benchflow-0.7.2.dev1994}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: benchflow
-Version: 0.7.2.dev1987
+Version: 0.7.2.dev1994
 Summary: Multi-turn agent benchmarking with ACP — run any agent, any model, any provider.
 Project-URL: Homepage, https://github.com/benchflow-ai/benchflow
 Project-URL: Repository, https://github.com/benchflow-ai/benchflow
@@ -89,7 +89,7 @@ BenchFlow is a universal environment framework: it runs AI agents against task e
 
 ## Quick start: 1. Submit a trajectory
 
-**Don't follow these steps yourself — send them to your coding agent.** Copy
+**Send these to your coding agent.** Copy
 the block below and paste it as a message to Claude Code, Codex, OpenCode, or
 Cursor. That is the $2,000 eval prize path. No BenchFlow account, API key, or
 Azure login.

{benchflow-0.7.2.dev1987 → benchflow-0.7.2.dev1994}/README.md

@@ -15,7 +15,7 @@ BenchFlow is a universal environment framework: it runs AI agents against task e
 
 ## Quick start: 1. Submit a trajectory
 
-**Don't follow these steps yourself — send them to your coding agent.** Copy
+**Send these to your coding agent.** Copy
 the block below and paste it as a message to Claude Code, Codex, OpenCode, or
 Cursor. That is the $2,000 eval prize path. No BenchFlow account, API key, or
 Azure login.

{benchflow-0.7.2.dev1987 → benchflow-0.7.2.dev1994}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "benchflow"
-version = "0.7.2.dev1987"
+version = "0.7.2.dev1994"
 description = "Multi-turn agent benchmarking with ACP — run any agent, any model, any provider."
 readme = "README.md"
 requires-python = ">=3.12"

{benchflow-0.7.2.dev1987 → benchflow-0.7.2.dev1994}/src/benchflow/cli/main.py

@@ -1228,11 +1228,27 @@ def eval_view(
         ),
     ],
     port: Annotated[int, typer.Option(help="Server port")] = 8888,
+    confirm: Annotated[
+        bool,
+        typer.Option(
+            "--confirm",
+            help=(
+                "Show an Approve & submit / Not this one bar on the page; "
+                "print DECISION: approved|rejected and exit 0 on approve, "
+                "3 on reject."
+            ),
+        ),
+    ] = False,
 ) -> None:
     """View a trial or session trajectory in the browser."""
-    from benchflow.trajectories.viewer import serve
-
-    serve(str(rollout_dir), port)
+    from benchflow.trajectories import viewer
+
+    decision = viewer.serve(str(rollout_dir), port, confirm=confirm)
+    # Exit-code contract for --confirm: 0 approved, 3 rejected. 3 is chosen
+    # so a rejection never collides with the CLI's existing error exits
+    # (1 = errors, 2 = Typer usage errors).
+    if decision == "rejected":
+        raise typer.Exit(3)
 
 
 # ── Command-group wiring ──────────────────────────────────────────────

{benchflow-0.7.2.dev1987 → benchflow-0.7.2.dev1994}/src/benchflow/trajectories/viewer.py

@@ -16,9 +16,15 @@ _RESULT_PREVIEW = 300  # max chars for result summary
 
 # Shared stylesheet for all viewer pages, matching the www.benchflow.ai design
 # language (light monochrome: near-white page, white cards, near-black ink,
-# pure-black accent, Satoshi/Google Sans Code with system-safe fallbacks, dark
-# code blocks). Kept inline so pages work fully offline with no external
-# requests; one constant so the three templates stop drifting apart.
+# Satoshi/Google Sans Code with system-safe fallbacks). Kept inline so pages
+# work fully offline with no external requests; one constant so the three
+# templates stop drifting apart.
+#
+# The site itself is deliberately achromatic, so tool-call accents below are
+# muted GitHub-label-style pastels (pale chip background + darker same-hue
+# text + a soft left border strip) that read as annotations rather than
+# fighting the monochrome base. The dark #141414 code treatment is reserved
+# for terminal output of shell commands; everything else stays light.
 _VIEWER_CSS = """\
 * { margin: 0; padding: 0; box-sizing: border-box; }
 :root {
@@ -46,8 +52,10 @@ body { font-family: var(--font-sans); background: var(--background); color: var(
 .meta span { font-family: var(--font-mono); font-size: 11px; font-weight: 500; color: var(--ink-secondary); background: var(--secondary); padding: 3px 9px; border-radius: 4px; border: 1px solid var(--border); }
 .step { margin-bottom: 8px; padding: 12px 16px; border-radius: var(--radius); background: var(--card); border: 1px solid var(--border); box-shadow: 0 1px 2px rgba(10, 10, 10, 0.04); }
 .step.prompt { background: var(--secondary); border-color: var(--rule-strong); margin-bottom: 14px; }
-.step.output { background: var(--code-bg); border-color: var(--code-bg); padding: 10px 16px; }
-.step.output pre { color: var(--code-ink); font-family: var(--font-mono); font-size: 12px; line-height: 1.7; white-space: pre-wrap; word-break: break-word; }
+.step.output { background: var(--card); border-color: var(--border); padding: 10px 16px; }
+.step.output pre { color: var(--ink-secondary); font-family: var(--font-mono); font-size: 12px; line-height: 1.7; white-space: pre-wrap; word-break: break-word; }
+.step.output.term { background: var(--code-bg); border-color: var(--code-bg); }
+.step.output.term pre { color: var(--code-ink); }
 .step.result { background: var(--ink); border-color: var(--ink); margin-top: 14px; }
 .step.result .msg { color: var(--background); }
 .step-header { display: flex; align-items: center; gap: 8px; margin-bottom: 8px; }
@@ -59,12 +67,70 @@ body { font-family: var(--font-sans); background: var(--background); color: var(
 .msg { font-size: 14px; line-height: 1.65; white-space: pre-wrap; word-break: break-word; }
 .thinking { font-size: 13px; color: var(--muted); font-style: italic; margin-bottom: 8px; padding: 8px 12px; background: var(--secondary); border-radius: 4px; border-left: 2px solid var(--rule-strong); white-space: pre-wrap; word-break: break-word; }
 .tool { margin-bottom: 6px; }
-.tool-name { display: inline-flex; align-items: center; font-family: var(--font-mono); font-size: 11px; font-weight: 600; color: var(--ink); background: var(--secondary); border: 1px solid var(--border); padding: 2px 9px; border-radius: 4px; }
-.tool-args { margin-top: 6px; font-family: var(--font-mono); font-size: 12px; line-height: 1.7; color: var(--code-ink); background: var(--code-bg); padding: 10px 12px; border-radius: 6px; white-space: pre-wrap; word-break: break-word; }
+.tool-name { display: inline-flex; align-items: center; font-family: var(--font-mono); font-size: 11px; font-weight: 600; color: var(--acc-ink, var(--ink)); background: var(--acc-bg, var(--secondary)); border: 1px solid var(--acc-line, var(--border)); padding: 2px 9px; border-radius: 4px; }
+.tool-args { margin-top: 6px; font-family: var(--font-mono); font-size: 12px; line-height: 1.7; color: var(--ink-secondary); background: var(--secondary); border: 1px solid var(--border); padding: 10px 12px; border-radius: 6px; white-space: pre-wrap; word-break: break-word; }
+.step.tool-step { border-left: 3px solid var(--acc-strip, var(--rule-strong)); }
+.acc-bash  { --acc-bg: #f7efda; --acc-line: #ecdcb2; --acc-ink: #8a5a12; --acc-strip: #dcb45e; }
+.acc-edit  { --acc-bg: #e8f0fa; --acc-line: #d0dff1; --acc-ink: #1d4e89; --acc-strip: #7fa8d8; }
+.acc-read  { --acc-bg: #e5f2ec; --acc-line: #cbe3d7; --acc-ink: #1a6b52; --acc-strip: #74bda0; }
+.acc-agent { --acc-bg: #efeaf8; --acc-line: #ded3ef; --acc-ink: #5b3e96; --acc-strip: #a78fd6; }
+.acc-web   { --acc-bg: #e3f1f6; --acc-line: #c8e2ea; --acc-ink: #176478; --acc-strip: #6fb6ca; }
+.acc-other { --acc-bg: var(--secondary); --acc-line: var(--border); --acc-ink: var(--ink-secondary); --acc-strip: var(--rule-strong); }
 .metrics { font-family: var(--font-mono); font-size: 11px; color: var(--faint); margin-top: 4px; }
 .turn-divider { border-top: 1px solid var(--border); margin: 24px 0; }
 """
 
+# Sticky bottom confirmation bar injected only in --confirm mode. Styling
+# reuses the page's CSS variables (white surface, top border, ink text, pill
+# buttons) so it reads as part of the same www.benchflow.ai design language.
+# The <style> block rides along with the snippet so plain (non-confirm) pages
+# carry zero confirm markup or CSS.
+_CONFIRM_BAR_HTML = """\
+<style>
+body { padding-bottom: 104px; }
+.confirm-bar { position: fixed; bottom: 0; left: 0; right: 0; background: var(--card); border-top: 1px solid var(--border); box-shadow: 0 -1px 3px rgba(10, 10, 10, 0.05); z-index: 10; }
+.confirm-inner { max-width: 960px; margin: 0 auto; padding: 14px 20px; display: flex; align-items: center; justify-content: space-between; gap: 16px; flex-wrap: wrap; }
+.confirm-question { font-size: 14px; font-weight: 500; color: var(--ink); }
+.confirm-actions { display: flex; gap: 8px; flex: none; }
+.confirm-btn { font-family: var(--font-sans); font-size: 13px; font-weight: 600; letter-spacing: -0.01em; padding: 8px 18px; border-radius: 9999px; cursor: pointer; transition: background 0.15s ease; }
+.confirm-btn.approve { background: var(--ink); color: var(--background); border: 1px solid var(--ink); }
+.confirm-btn.approve:hover { background: #262626; }
+.confirm-btn.reject { background: var(--card); color: var(--ink); border: 1px solid var(--rule-strong); }
+.confirm-btn.reject:hover { background: var(--secondary); }
+</style>
+<div class="confirm-bar">
+<div class="confirm-inner" id="confirm-inner">
+<span class="confirm-question">Submit this trajectory to the BenchFlow eval prize?</span>
+<div class="confirm-actions">
+<button class="confirm-btn approve" onclick="__benchDecide('approve')">Approve &amp; submit</button>
+<button class="confirm-btn reject" onclick="__benchDecide('reject')">Not this one</button>
+</div>
+</div>
+</div>
+<script>
+async function __benchDecide(choice) {
+  try {
+    await fetch("/decision", { method: "POST", body: choice });
+  } catch (err) {
+    // The server exits right after answering; a dropped socket is still a
+    // delivered decision, so fall through to the confirmation note.
+  }
+  document.getElementById("confirm-inner").innerHTML =
+    choice === "approve"
+      ? '<span class="confirm-question">Approved — go back to your agent.</span>'
+      : '<span class="confirm-question">Rejected — tell your agent which session instead.</span>';
+}
+</script>
+"""
+
+
+def _inject_confirm_bar(page: str) -> str:
+    """Append the confirm bar just before </body> (or at the end as fallback)."""
+    if "</body>" in page:
+        return page.replace("</body>", f"{_CONFIRM_BAR_HTML}</body>", 1)
+    return page + _CONFIRM_BAR_HTML
+
+
 # Small BenchFlow wordmark header (inline SVG logo from the site's icon set —
 # no external requests) shown above the page title on every viewer page.
 _WORDMARK_HTML = (
@@ -87,6 +153,28 @@ _WORDMARK_HTML = (
 )
 
 
+# Tool kind → accent class. Substring matching so it covers Claude Code tool
+# names (Bash, Write, Edit, Read, Agent, WebSearch, ...), ACP kinds (execute,
+# edit, read, search, fetch, ...), and Codex function names (shell, ...).
+# Order matters: earlier entries win (e.g. "websearch" hits web before read).
+_TOOL_ACCENTS: tuple[tuple[str, tuple[str, ...]], ...] = (
+    ("acc-web", ("web", "search", "fetch", "grep", "glob", "browser")),
+    ("acc-bash", ("bash", "shell", "exec", "terminal", "command")),
+    ("acc-edit", ("write", "edit", "patch", "delete", "move", "notebook")),
+    ("acc-read", ("read", "cat", "view", "ls", "list")),
+    ("acc-agent", ("agent", "task", "skill", "oracle")),
+)
+
+
+def _tool_accent_class(name: str) -> str:
+    """Map a tool name/kind to its accent CSS class (presentation only)."""
+    lowered = name.lower()
+    for css_class, needles in _TOOL_ACCENTS:
+        if any(needle in lowered for needle in needles):
+            return css_class
+    return "acc-other"
+
+
 def render_turn(events: list[dict], turn_number: int, prompt: str = "") -> str:
     """Render one turn's events as HTML blocks."""
     blocks = []
@@ -103,6 +191,9 @@ def render_turn(events: list[dict], turn_number: int, prompt: str = "") -> str:
     # Group: thinking → text → tool_use → tool_result → thinking → ...
     pending_thinking = ""
     pending_text = ""
+    # tool_use id → accent class, so each tool_result can match its call's
+    # accent and shell results alone keep the dark terminal treatment.
+    accent_by_tool_id: dict[str, str] = {}
 
     for event in events:
         etype = event.get("type", "")
@@ -160,7 +251,13 @@ def render_turn(events: list[dict], turn_number: int, prompt: str = "") -> str:
                         f"</div>"
                     )
 
-                    blocks.append(f'<div class="step agent">{"".join(parts)}</div>')
+                    accent = _tool_accent_class(block.get("name", ""))
+                    if block.get("id"):
+                        accent_by_tool_id[str(block["id"])] = accent
+                    blocks.append(
+                        f'<div class="step agent tool-step {accent}">'
+                        f"{''.join(parts)}</div>"
+                    )
 
         elif etype == "user":
             content = event.get("message", {}).get("content", "")
@@ -181,8 +278,13 @@ def render_turn(events: list[dict], turn_number: int, prompt: str = "") -> str:
                             display = "[binary content]"
                         else:
                             display = html.escape(raw[:400])
+                        accent = accent_by_tool_id.get(
+                            str(block.get("tool_use_id", "")), "acc-other"
+                        )
+                        term = " term" if accent == "acc-bash" else ""
                         blocks.append(
-                            f'<div class="step output"><pre>{display}</pre></div>'
+                            f'<div class="step output tool-step {accent}{term}">'
+                            f"<pre>{display}</pre></div>"
                         )
                     elif block.get("type") == "text" and block.get("text"):
                         texts.append(str(block["text"]))
@@ -408,8 +510,13 @@ def _render_acp_events(
             kind = html.escape(event.get("kind", ""))
             event_title = html.escape(event.get("title", ""))
             status = event.get("status", "")
+            # ACP kinds are coarse ("other" covers Skill/Task/...); when the
+            # kind doesn't map, fall back to the human title for the accent.
+            accent = _tool_accent_class(event.get("kind", ""))
+            if accent == "acc-other":
+                accent = _tool_accent_class(event.get("title", ""))
             blocks.append(
-                f'<div class="step agent">'
+                f'<div class="step agent tool-step {accent}">'
                 f'<div class="tool"><span class="tool-name">{kind}</span> {event_title}</div>'
                 f'<div class="metrics">{status}</div>'
                 f"</div>"
@@ -578,9 +685,21 @@ def _stream_json_page(title: str, events: list[dict], turn_blocks: list[str]) ->
 
 
 def serve(
-    rollout_path: str, port: int = 8888, prompts: list[str] | None = None
-) -> None:
-    """Serve a trial directory or a session JSONL file as a web page."""
+    rollout_path: str,
+    port: int = 8888,
+    prompts: list[str] | None = None,
+    confirm: bool = False,
+) -> str | None:
+    """Serve a trial directory or a session JSONL file as a web page.
+
+    With ``confirm=True`` the page carries an Approve/Reject bar posting to
+    ``/decision``; the first valid decision shuts the server down and is
+    returned as ``"approved"`` or ``"rejected"`` after printing a
+    machine-readable ``DECISION: <value>`` line to stdout. Without it the
+    server runs until Ctrl+C and the return value is ``None`` — exactly the
+    pre-confirm behavior (no bar, no POST endpoint).
+    """
+    import threading
     from http.server import HTTPServer, SimpleHTTPRequestHandler
 
     path = Path(rollout_path)
@@ -600,11 +719,20 @@ def serve(
         print(f"No trajectories found in {path}")
         sys.exit(1)
     if write_sidecar:
+        # The sidecar stays the plain page: the confirm bar is a one-shot
+        # interaction against this live server, not part of the artifact.
         (path / "trajectory.html").write_text(html_content)
+    if confirm:
+        html_content = _inject_confirm_bar(html_content)
 
     print(f"Trajectory viewer: http://localhost:{port}")
     print(f"Trial: {path}")
-    print("Press Ctrl+C to stop\n")
+    if confirm:
+        print("Waiting for Approve / Not this one in the browser (Ctrl+C to stop)\n")
+    else:
+        print("Press Ctrl+C to stop\n")
+
+    decision: str | None = None
 
     class Handler(SimpleHTTPRequestHandler):
         def do_GET(self):
@@ -616,11 +744,44 @@ def serve(
         def log_message(self, format, *args):
             pass
 
-    server = HTTPServer(("localhost", port), Handler)
+    handler_cls: type[SimpleHTTPRequestHandler] = Handler
+
+    if confirm:
+
+        class ConfirmHandler(Handler):
+            def do_POST(self):
+                nonlocal decision
+                if self.path != "/decision":
+                    self.send_error(404)
+                    return
+                length = int(self.headers.get("Content-Length") or 0)
+                body = self.rfile.read(length).decode("utf-8", "replace").strip()
+                if body not in ("approve", "reject"):
+                    self.send_error(400, "Body must be 'approve' or 'reject'")
+                    return
+                decision = "approved" if body == "approve" else "rejected"
+                self.send_response(200)
+                self.send_header("Content-Type", "text/plain; charset=utf-8")
+                self.end_headers()
+                self.wfile.write(decision.encode())
+                # shutdown() blocks until serve_forever returns, and this
+                # handler runs inside that loop — stop from a helper thread.
+                threading.Thread(target=server.shutdown, daemon=True).start()
+
+        handler_cls = ConfirmHandler
+
+    server = HTTPServer(("localhost", port), handler_cls)
     try:
         server.serve_forever()
     except KeyboardInterrupt:
         print("\nStopped.")
+        return None
+    finally:
+        server.server_close()
+
+    if decision is not None:
+        print(f"DECISION: {decision}", flush=True)
+    return decision
 
 
 if __name__ == "__main__":

{benchflow-0.7.2.dev1987 → benchflow-0.7.2.dev1994}/tests/test_traj_upload_cli.py

@@ -747,7 +747,8 @@ def test_setup_prompt_prints_the_copy_paste_line() -> None:
     assert len(prompt_lines) == 3
     for line in prompt_lines:
         assert line in readme_text
-    assert "send them to your coding agent" in readme_text
+    # Wording updated on main in 8cae2a42 ("Send these to your coding agent.").
+    assert "Send these to your coding agent" in readme_text
 
     assert UPGRADE_COMMAND in CONTRIBUTOR_PROMPT
     assert CONTRIBUTOR_PROMPT.index(UPGRADE_COMMAND) < CONTRIBUTOR_PROMPT.index(

benchflow-0.7.2.dev1994/tests/test_viewer_confirm.py

@@ -0,0 +1,205 @@
+"""Tests for the trajectory viewer's --confirm flow (in-browser approve/reject).
+
+Guards the feature shipped in PR #1021: --confirm must add the decision bar +
+/decision endpoint and the DECISION stdout/exit-code contract, while plain
+mode stays byte-for-byte the pre-#1021 viewer.
+
+The eval-prize contributor loop is agent-driven: the agent serves the viewer,
+the human clicks **Approve & submit** or **Not this one**, and the process
+reports the decision via a machine-readable ``DECISION:`` stdout line plus the
+exit code (0 approve, 3 reject) so the agent can wait on it instead of a chat
+reply.
+"""
+
+from __future__ import annotations
+
+import json
+import socket
+import threading
+import time
+import urllib.error
+import urllib.request
+from pathlib import Path
+
+import pytest
+from typer.testing import CliRunner
+
+from benchflow.cli.main import app
+from benchflow.trajectories import viewer
+from benchflow.trajectories.viewer import render_jsonl_file, serve
+
+runner = CliRunner()
+
+_BAR_NEEDLES = (
+    "Submit this trajectory to the BenchFlow eval prize?",
+    "Approve & submit",
+    "Not this one",
+    "/decision",
+)
+
+
+def _write_session(tmp_path: Path) -> Path:
+    session = tmp_path / "session.jsonl"
+    session.write_text(
+        json.dumps({"type": "user", "message": {"content": "review me"}}) + "\n"
+    )
+    return session
+
+
+def _free_port() -> int:
+    with socket.socket() as s:
+        s.bind(("localhost", 0))
+        return s.getsockname()[1]
+
+
+def _serve_in_thread(
+    path: Path, *, confirm: bool
+) -> tuple[int, threading.Thread, dict]:
+    """Run serve() in a daemon thread; poll GET / until it answers."""
+    port = _free_port()
+    result: dict = {}
+
+    def run() -> None:
+        result["decision"] = serve(str(path), port, confirm=confirm)
+
+    thread = threading.Thread(target=run, daemon=True)
+    thread.start()
+    deadline = time.monotonic() + 10
+    while True:
+        try:
+            with urllib.request.urlopen(f"http://localhost:{port}/", timeout=1) as r:
+                result["page"] = r.read().decode()
+                break
+        except OSError:
+            if time.monotonic() > deadline:
+                raise
+            time.sleep(0.05)
+    return port, thread, result
+
+
+def _post_decision(port: int, body: str) -> str:
+    req = urllib.request.Request(
+        f"http://localhost:{port}/decision", data=body.encode(), method="POST"
+    )
+    with urllib.request.urlopen(req, timeout=5) as r:
+        return r.read().decode()
+
+
+def test_confirm_page_has_bar_and_plain_page_does_not(tmp_path):
+    """--confirm injects the sticky decision bar; plain rendering carries none of it."""
+    session = _write_session(tmp_path)
+
+    plain_html = render_jsonl_file(session)
+    for needle in _BAR_NEEDLES:
+        assert needle not in plain_html
+
+    port, thread, result = _serve_in_thread(session, confirm=True)
+    for needle in _BAR_NEEDLES:
+        assert needle in result["page"]
+
+    # Shut the server down so the thread exits.
+    _post_decision(port, "reject")
+    thread.join(timeout=10)
+    assert not thread.is_alive()
+
+
+@pytest.mark.parametrize(
+    "body, expected", [("approve", "approved"), ("reject", "rejected")]
+)
+def test_post_decision_returns_decision_and_prints_line(
+    tmp_path, capsys, body, expected
+):
+    """POST /decision shuts the server down; serve() returns the decision and
+    prints exactly one machine-readable DECISION line to stdout."""
+    session = _write_session(tmp_path)
+    port, thread, result = _serve_in_thread(session, confirm=True)
+
+    response = _post_decision(port, body)
+    thread.join(timeout=10)
+    assert not thread.is_alive()
+
+    assert response == expected
+    assert result["decision"] == expected
+    out = capsys.readouterr().out
+    assert out.count(f"DECISION: {expected}") == 1
+
+
+def test_post_decision_rejects_garbage_body(tmp_path):
+    """A body other than approve/reject is a 400 and keeps the server alive."""
+    session = _write_session(tmp_path)
+    port, thread, result = _serve_in_thread(session, confirm=True)
+
+    with pytest.raises(urllib.error.HTTPError) as exc:
+        _post_decision(port, "maybe")
+    assert exc.value.code == 400
+    assert thread.is_alive()  # still waiting for a real decision
+
+    _post_decision(port, "reject")
+    thread.join(timeout=10)
+    assert result["decision"] == "rejected"
+
+
+def test_plain_mode_has_no_decision_endpoint(tmp_path):
+    """Without --confirm, POST /decision does not exist (501 from the stdlib
+    handler) and GET / serves the page exactly as before."""
+    session = _write_session(tmp_path)
+    port, _thread, result = _serve_in_thread(session, confirm=False)
+
+    assert "review me" in result["page"]
+    with pytest.raises(urllib.error.HTTPError) as exc:
+        _post_decision(port, "approve")
+    assert exc.value.code == 501
+    # The plain server only stops on Ctrl+C; the daemon thread is reaped at
+    # process exit, matching the documented always-on behavior.
+
+
+def test_confirm_sidecar_stays_plain(tmp_path):
+    """The trajectory.html sidecar written for directories must not embed the
+    one-shot confirm bar."""
+    rollout = tmp_path / "rollout"
+    rollout.mkdir()
+    (rollout / "turn1.txt").write_text(
+        '{"type":"system","session_id":"s","model":"m"}\n'
+    )
+    port, thread, _result = _serve_in_thread(rollout, confirm=True)
+    sidecar = (rollout / "trajectory.html").read_text()
+    for needle in _BAR_NEEDLES:
+        assert needle not in sidecar
+
+    _post_decision(port, "reject")
+    thread.join(timeout=10)
+
+
+@pytest.mark.parametrize(
+    "decision, expected_exit", [("approved", 0), ("rejected", 3), (None, 0)]
+)
+def test_eval_view_maps_decision_to_exit_code(
+    tmp_path, monkeypatch, decision, expected_exit
+):
+    """The Typer command owns the exit-code mapping: approve → 0, reject → 3
+    (non-1/2 so it cannot collide with error or usage exits), Ctrl+C → 0."""
+    calls: dict = {}
+
+    def fake_serve(path, port=8888, prompts=None, confirm=False):
+        calls["confirm"] = confirm
+        return decision
+
+    monkeypatch.setattr(viewer, "serve", fake_serve)
+    res = runner.invoke(app, ["eval", "view", str(tmp_path), "--confirm"])
+    assert res.exit_code == expected_exit
+    assert calls["confirm"] is True
+
+
+def test_eval_view_defaults_to_no_confirm(tmp_path, monkeypatch):
+    """Without the flag, serve() must be called with confirm=False so today's
+    behavior (no bar, no endpoint) is preserved."""
+    calls: dict = {}
+
+    def fake_serve(path, port=8888, prompts=None, confirm=False):
+        calls["confirm"] = confirm
+        return None
+
+    monkeypatch.setattr(viewer, "serve", fake_serve)
+    res = runner.invoke(app, ["eval", "view", str(tmp_path)])
+    assert res.exit_code == 0
+    assert calls["confirm"] is False