benchflow 0.7.2.dev1980__tar.gz → 0.7.2.dev1986__tar.gz

{benchflow-0.7.2.dev1980 → benchflow-0.7.2.dev1986}/CHANGELOG.md

@@ -2,6 +2,28 @@
 
 ## [Unreleased]
 
+### Added
+- **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
+  `session_meta`), resolves its git `origin` remote (invocation directory as
+  fallback), and stores `repo/<owner>/<name>` as the manifest source id,
+  printing `Repo: owner/name (use --no-repo to omit)`. `--no-repo` opts out
+  — the `benchflow-traj-upload` skill now surfaces the detected tag during
+  the confirm step so contributors can decline it for private repos — and
+  undetectable repos fall back silently to the path-derived source id.
+
+### Changed
+- **Trajectory viewer restyled to match www.benchflow.ai.** All three
+  `bench eval view` pages (stream-json/JSONL, ACP events, multi-turn trial)
+  now share one inline stylesheet with the site's design language: light
+  monochrome palette (`#fafafa` page, white cards, `#0a0a0a` ink, dark
+  `#141414` code blocks), Satoshi/Google Sans Code font stacks with
+  system-safe fallbacks, mono pill badges, and a small BenchFlow wordmark
+  header with the inline SVG logo. Pages remain fully offline (no external
+  font or CDN requests) and content/structure semantics are unchanged;
+  follows the contributor paste-line flow from #1013.
+
 ## 0.7.1 — 2026-08-16
 
 ### Added

{benchflow-0.7.2.dev1980 → benchflow-0.7.2.dev1986}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: benchflow
-Version: 0.7.2.dev1980
+Version: 0.7.2.dev1986
 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

{benchflow-0.7.2.dev1980 → benchflow-0.7.2.dev1986}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "benchflow"
-version = "0.7.2.dev1980"
+version = "0.7.2.dev1986"
 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.dev1980 → benchflow-0.7.2.dev1986}/src/benchflow/cli/traj.py

@@ -2,6 +2,7 @@
 
 from __future__ import annotations
 
+import json
 import os
 import shutil
 import subprocess
@@ -29,6 +30,7 @@ from benchflow.publish.traj_capture import (
     stage_trajectory_artifacts,
     validate_email,
     validate_github_id,
+    validate_source_id,
 )
 from benchflow.publish.traj_report import (
     DEFAULT_PREVIEW_STEPS,
@@ -127,6 +129,7 @@ class _UploadOptions:
     github_id: str | None
     email: str | None
     source_id: str | None
+    repo: bool
     direct: bool
     container_url: str | None
     dry_run: bool
@@ -223,6 +226,14 @@ def register_traj(app: typer.Typer) -> None:
             str | None,
             typer.Option("--source-id", help="Stable contributor source identifier"),
         ] = None,
+        repo: Annotated[
+            bool,
+            typer.Option(
+                "--repo/--no-repo",
+                help="Tag the upload with the session's repository "
+                "(owner/name from its git remote) as the source id",
+            ),
+        ] = True,
         direct: Annotated[
             bool,
             typer.Option("--direct", help="Upload with local Azure credentials"),
@@ -254,6 +265,7 @@ def register_traj(app: typer.Typer) -> None:
                     github_id=github_id,
                     email=email,
                     source_id=source_id,
+                    repo=repo,
                     direct=direct,
                     container_url=container_url,
                     dry_run=dry_run,
@@ -268,7 +280,17 @@ def register_traj(app: typer.Typer) -> None:
 def _run_upload(options: _UploadOptions) -> None:
     prompted = options.path is None
     path = options.path or _prompt_for_path()
-    source_id = options.source_id or default_source_id(path)
+    repo_slug: str | None = None
+    if options.source_id is not None:
+        source_id = options.source_id
+    else:
+        if options.repo:
+            repo_slug = _detect_repo_slug(path)
+        source_id = f"repo/{repo_slug}" if repo_slug else default_source_id(path)
+    if repo_slug:
+        # Contributor-visible metadata: surface the tag so private-repo
+        # sessions can opt out before anything leaves the machine.
+        console.print(f"Repo: {repo_slug} (use --no-repo to omit)")
     destination = _resolve_destination(options)
 
     with (
@@ -381,6 +403,92 @@ def _infer_email() -> str:
     return ""
 
 
+_SESSION_CWD_SCAN_LINES = 50
+
+
+def _detect_repo_slug(path: Path) -> str | None:
+    """Best-effort ``owner/name`` for the repository the session was about.
+
+    Reads the working directory the session recorded (Claude events carry a
+    ``cwd`` field; Codex ``session_meta`` payloads do too), asks that
+    directory's git for the ``origin`` remote, and falls back to the
+    invocation directory. Every failure is silent — repo tagging must never
+    break an upload — and local-path remotes never produce a tag, so no
+    local absolute path can leak into the manifest.
+    """
+    for candidate in (_session_cwd(path), Path.cwd()):
+        if candidate is None or not candidate.is_dir():
+            continue
+        remote = _command_stdout(
+            "git", "-C", str(candidate), "remote", "get-url", "origin"
+        )
+        slug = _repo_slug_from_remote(remote) if remote else None
+        if slug:
+            return slug
+    return None
+
+
+def _session_cwd(path: Path) -> Path | None:
+    """Working directory recorded by the first session event that has one."""
+    session = path.expanduser()
+    if not session.is_file():
+        return None
+    try:
+        with session.open(encoding="utf-8", errors="replace") as stream:
+            for line_number, line in enumerate(stream):
+                if line_number >= _SESSION_CWD_SCAN_LINES:
+                    break
+                body = line.strip()
+                if not body:
+                    continue
+                try:
+                    event = json.loads(body)
+                except json.JSONDecodeError:
+                    continue
+                if not isinstance(event, dict):
+                    continue
+                cwd = event.get("cwd")
+                if isinstance(cwd, str) and cwd:
+                    return Path(cwd)
+                payload = event.get("payload")
+                if (
+                    event.get("type") == "session_meta"
+                    and isinstance(payload, dict)
+                    and isinstance(payload.get("cwd"), str)
+                    and payload["cwd"]
+                ):
+                    return Path(payload["cwd"])
+    except OSError:
+        return None
+    return None
+
+
+def _repo_slug_from_remote(remote: str) -> str | None:
+    """Normalize an https/ssh git remote URL to ``owner/name``.
+
+    Only URL-shaped remotes qualify; a filesystem-path remote returns
+    ``None`` so local paths never enter the uploaded source id.
+    """
+    value = remote.strip()
+    if "://" in value:
+        _, _, rest = value.partition("://")
+        _, _, repo_path = rest.partition("/")
+    elif ":" in value and "@" in value.partition(":")[0]:
+        repo_path = value.partition(":")[2]
+    else:
+        return None
+    repo_path = repo_path.strip("/").removesuffix(".git")
+    segments = [segment for segment in repo_path.split("/") if segment]
+    if len(segments) < 2:
+        return None
+    slug = f"{segments[-2]}/{segments[-1]}"
+    try:
+        validate_source_id(f"repo/{slug}")
+    except ValueError:
+        return None
+    return slug
+
+
 def _git_config(key: str) -> str | None:
     return _command_stdout("git", "config", "--get", key)
 

{benchflow-0.7.2.dev1980 → benchflow-0.7.2.dev1986}/src/benchflow/trajectories/viewer.py

@@ -14,6 +14,78 @@ _ARGS_PREVIEW = 300  # max chars for tool args display
 _CONTENT_PREVIEW = 200  # max chars for write/agent content preview
 _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.
+_VIEWER_CSS = """\
+* { margin: 0; padding: 0; box-sizing: border-box; }
+:root {
+  --background: #fafafa; --card: #ffffff;
+  --ink: #0a0a0a; --ink-secondary: #404040; --muted: #737373; --faint: #a1a1a1;
+  --border: #e5e5e5; --rule-strong: #c7c7c7; --secondary: #f5f5f5;
+  --code-bg: #141414; --code-ink: #ececec;
+  --radius: 8px;
+  --font-sans: "Satoshi", ui-sans-serif, system-ui, -apple-system, "Segoe UI", sans-serif;
+  --font-mono: "Google Sans Code", ui-monospace, SFMono-Regular, Menlo, Monaco, Consolas, monospace;
+}
+body { font-family: var(--font-sans); background: var(--background); color: var(--ink); padding: 28px 20px 48px; max-width: 960px; margin: 0 auto; line-height: 1.6; -webkit-font-smoothing: antialiased; }
+::selection { background: var(--secondary); color: var(--ink); }
+::-webkit-scrollbar { width: 6px; height: 6px; }
+::-webkit-scrollbar-track { background: transparent; }
+::-webkit-scrollbar-thumb { background: var(--border); border-radius: 9999px; }
+::-webkit-scrollbar-thumb:hover { background: var(--muted); }
+.wordmark { display: flex; align-items: center; gap: 10px; margin-bottom: 18px; }
+.wordmark svg { width: 18px; height: 18px; flex: none; color: var(--ink); }
+.wordmark .brand { font-weight: 600; font-size: 15px; letter-spacing: -0.01em; color: var(--ink); }
+.wordmark .app { font-family: var(--font-mono); font-size: 10.5px; font-weight: 500; color: var(--ink-secondary); background: var(--secondary); border: 1px solid var(--border); border-radius: 9999px; padding: 3px 10px; }
+.header { border-bottom: 1px solid var(--border); padding-bottom: 18px; margin-bottom: 24px; }
+.header h1 { font-size: 20px; font-weight: 600; letter-spacing: -0.02em; color: var(--ink); margin-bottom: 10px; overflow-wrap: anywhere; }
+.meta { display: flex; gap: 8px; flex-wrap: wrap; font-size: 13px; color: var(--muted); }
+.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.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; }
+.label { display: inline-flex; align-items: center; font-family: var(--font-mono); padding: 2px 10px; border-radius: 9999px; font-weight: 600; font-size: 10.5px; text-transform: uppercase; letter-spacing: 0.04em; }
+.label.prompt { background: var(--ink); color: var(--background); }
+.label.result { background: var(--background); color: var(--ink); }
+.meta-inline { font-family: var(--font-mono); font-size: 11px; color: var(--muted); }
+.step.result .meta-inline { color: var(--faint); }
+.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; }
+.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; }
+"""
+
+# 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 = (
+    '<div class="wordmark">'
+    '<svg viewBox="0 0 514 512" fill="currentColor" aria-hidden="true">'
+    '<path fill-rule="evenodd" clip-rule="evenodd" d="M445.422 66.4597L511.882 0'
+    "L389.022 293.965L295.129 387.859L0 511.882L69.3042 442.577L81.0101 454.283"
+    "L89.0554 446.238L77.3493 434.532L130.65 381.232L162.469 413.051L170.514 40"
+    "5.006L138.695 373.187L191.995 319.887L203.701 331.593L211.746 323.547L200."
+    "04 311.841L253.34 258.541L285.16 290.36L293.205 282.315L261.386 250.496L31"
+    "4.686 197.196L326.392 208.902L334.437 200.856L322.731 189.15L376.031 135.8"
+    "5L407.851 167.669L415.896 159.624L384.077 127.805L437.377 74.5049L449.083 "
+    "86.2108L457.128 78.1656L445.422 66.4597ZM399.127 389.865V299.369L513.197 2"
+    "6.4333V503.935L399.127 389.865ZM391.061 397.931L505.132 512.001H29.1594L30"
+    '0.605 397.931H391.061Z"/>'
+    "</svg>"
+    '<span class="brand">BenchFlow</span>'
+    '<span class="app">trajectory viewer</span>'
+    "</div>"
+)
+
 
 def render_turn(events: list[dict], turn_number: int, prompt: str = "") -> str:
     """Render one turn's events as HTML blocks."""
@@ -260,33 +332,11 @@ def render_rollout(rollout_dir: Path, prompts: list[str] | None = None) -> str:
 <meta charset="utf-8">
 <title>benchflow — {rollout_dir.name}</title>
 <style>
-* {{ margin: 0; padding: 0; box-sizing: border-box; }}
-body {{ font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', sans-serif; background: #0d1117; color: #c9d1d9; padding: 20px; max-width: 960px; margin: 0 auto; }}
-.header {{ border-bottom: 1px solid #30363d; padding-bottom: 16px; margin-bottom: 24px; }}
-.header h1 {{ font-size: 20px; color: #f0f6fc; margin-bottom: 8px; }}
-.meta {{ display: flex; gap: 12px; flex-wrap: wrap; font-size: 13px; color: #8b949e; }}
-.meta span {{ background: #161b22; padding: 4px 10px; border-radius: 6px; border: 1px solid #30363d; }}
-.step {{ margin-bottom: 4px; padding: 10px 14px; border-radius: 6px; }}
-.step.prompt {{ background: #0d1f3c; border: 1px solid #1f3a5f; margin-bottom: 12px; }}
-.step.agent {{ background: #161b22; border: 1px solid #30363d; }}
-.step.output {{ background: #0d1117; border-left: 3px solid #238636; padding: 6px 14px; }}
-.step.output pre {{ color: #7ee787; font-size: 12px; white-space: pre-wrap; word-break: break-word; }}
-.step.result {{ background: #1a2f1a; border: 1px solid #238636; margin-top: 12px; }}
-.step-header {{ display: flex; align-items: center; gap: 8px; margin-bottom: 6px; }}
-.label {{ padding: 2px 8px; border-radius: 4px; font-weight: 600; font-size: 11px; text-transform: uppercase; }}
-.label.prompt {{ background: #1f3a5f; color: #58a6ff; }}
-.label.result {{ background: #1a2f1a; color: #3fb950; }}
-.meta-inline {{ font-size: 12px; color: #8b949e; }}
-.msg {{ font-size: 14px; line-height: 1.6; white-space: pre-wrap; word-break: break-word; }}
-.thinking {{ font-size: 13px; color: #8b949e; font-style: italic; margin-bottom: 6px; padding: 8px; background: #0d1117; border-radius: 4px; border-left: 3px solid #484f58; }}
-.tool {{ margin-bottom: 4px; }}
-.tool-name {{ background: #2d333b; color: #f0883e; padding: 2px 8px; border-radius: 4px; font-family: monospace; font-size: 13px; font-weight: 600; }}
-.tool-args {{ margin-top: 4px; font-size: 12px; color: #c9d1d9; background: #0d1117; padding: 8px; border-radius: 4px; white-space: pre-wrap; word-break: break-word; }}
-.turn-divider {{ border-top: 2px solid #30363d; margin: 20px 0; padding-top: 8px; }}
-</style>
+{_VIEWER_CSS}</style>
 </head>
 <body>
 <div class="header">
+{_WORDMARK_HTML}
 <h1>{html.escape(rollout_dir.name)}</h1>
 <div class="meta">
 <span>model: {html.escape(model)}</span>
@@ -392,25 +442,8 @@ def _render_acp_events(
     return f"""<!DOCTYPE html>
 <html><head><meta charset="utf-8"><title>benchflow — {html.escape(title)}</title>
 <style>
-* {{ margin: 0; padding: 0; box-sizing: border-box; }}
-body {{ font-family: -apple-system, sans-serif; background: #0d1117; color: #c9d1d9; padding: 20px; max-width: 960px; margin: 0 auto; }}
-.header {{ border-bottom: 1px solid #30363d; padding-bottom: 16px; margin-bottom: 24px; }}
-.header h1 {{ font-size: 20px; color: #f0f6fc; }}
-.step {{ margin-bottom: 4px; padding: 10px 14px; border-radius: 6px; }}
-.step.prompt {{ background: #0d1f3c; border: 1px solid #1f3a5f; margin-bottom: 12px; }}
-.step.agent {{ background: #161b22; border: 1px solid #30363d; }}
-.step.result {{ background: #1a2f1a; border: 1px solid #238636; margin-top: 12px; }}
-.step-header {{ margin-bottom: 6px; }}
-.label {{ padding: 2px 8px; border-radius: 4px; font-weight: 600; font-size: 11px; text-transform: uppercase; }}
-.label.prompt {{ background: #1f3a5f; color: #58a6ff; }}
-.label.result {{ background: #1a2f1a; color: #3fb950; }}
-.msg {{ font-size: 14px; line-height: 1.6; white-space: pre-wrap; word-break: break-word; }}
-.thinking {{ font-size: 13px; color: #8b949e; font-style: italic; padding: 8px; background: #0d1117; border-radius: 4px; border-left: 3px solid #484f58; }}
-.tool {{ margin-bottom: 4px; }}
-.tool-name {{ background: #2d333b; color: #f0883e; padding: 2px 8px; border-radius: 4px; font-family: monospace; font-size: 13px; font-weight: 600; }}
-.metrics {{ font-size: 11px; color: #484f58; margin-top: 4px; }}
-</style></head><body>
-<div class="header"><h1>{html.escape(title)}</h1></div>
+{_VIEWER_CSS}</style></head><body>
+<div class="header">{_WORDMARK_HTML}<h1>{html.escape(title)}</h1></div>
 {"".join(blocks)}
 </body></html>"""
 
@@ -526,33 +559,11 @@ def _stream_json_page(title: str, events: list[dict], turn_blocks: list[str]) ->
 <meta charset="utf-8">
 <title>benchflow — {html.escape(title)}</title>
 <style>
-* {{ margin: 0; padding: 0; box-sizing: border-box; }}
-body {{ font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', sans-serif; background: #0d1117; color: #c9d1d9; padding: 20px; max-width: 960px; margin: 0 auto; }}
-.header {{ border-bottom: 1px solid #30363d; padding-bottom: 16px; margin-bottom: 24px; }}
-.header h1 {{ font-size: 20px; color: #f0f6fc; margin-bottom: 8px; }}
-.meta {{ display: flex; gap: 12px; flex-wrap: wrap; font-size: 13px; color: #8b949e; }}
-.meta span {{ background: #161b22; padding: 4px 10px; border-radius: 6px; border: 1px solid #30363d; }}
-.step {{ margin-bottom: 4px; padding: 10px 14px; border-radius: 6px; }}
-.step.prompt {{ background: #0d1f3c; border: 1px solid #1f3a5f; margin-bottom: 12px; }}
-.step.agent {{ background: #161b22; border: 1px solid #30363d; }}
-.step.output {{ background: #0d1117; border-left: 3px solid #238636; padding: 6px 14px; }}
-.step.output pre {{ color: #7ee787; font-size: 12px; white-space: pre-wrap; word-break: break-word; }}
-.step.result {{ background: #1a2f1a; border: 1px solid #238636; margin-top: 12px; }}
-.step-header {{ display: flex; align-items: center; gap: 8px; margin-bottom: 6px; }}
-.label {{ padding: 2px 8px; border-radius: 4px; font-weight: 600; font-size: 11px; text-transform: uppercase; }}
-.label.prompt {{ background: #1f3a5f; color: #58a6ff; }}
-.label.result {{ background: #1a2f1a; color: #3fb950; }}
-.meta-inline {{ font-size: 12px; color: #8b949e; }}
-.msg {{ font-size: 14px; line-height: 1.6; white-space: pre-wrap; word-break: break-word; }}
-.thinking {{ font-size: 13px; color: #8b949e; font-style: italic; margin-bottom: 6px; padding: 8px; background: #0d1117; border-radius: 4px; border-left: 3px solid #484f58; }}
-.tool {{ margin-bottom: 4px; }}
-.tool-name {{ background: #2d333b; color: #f0883e; padding: 2px 8px; border-radius: 4px; font-family: monospace; font-size: 13px; font-weight: 600; }}
-.tool-args {{ margin-top: 4px; font-size: 12px; color: #c9d1d9; background: #0d1117; padding: 8px; border-radius: 4px; white-space: pre-wrap; word-break: break-word; }}
-.turn-divider {{ border-top: 2px solid #30363d; margin: 20px 0; padding-top: 8px; }}
-</style>
+{_VIEWER_CSS}</style>
 </head>
 <body>
 <div class="header">
+{_WORDMARK_HTML}
 <h1>{html.escape(title)}</h1>
 <div class="meta">
 <span>model: {html.escape(model)}</span>

{benchflow-0.7.2.dev1980 → benchflow-0.7.2.dev1986}/tests/test_traj_upload_cli.py

@@ -780,3 +780,138 @@ def test_list_recent_sessions_scans_all_projects_most_recent_first(
     assert all(hit.source == "claude" for hit in hits)
     assert "GC content" in hits[0].snippet
     assert "prize session please" in hits[1].snippet
+
+
+REPO_LINE = "Repo: benchflow-ai/benchflow (use --no-repo to omit)"
+
+
+def _session_with_cwd(tmp_path: Path, event: dict) -> Path:
+    session = tmp_path / "session.jsonl"
+    session.write_text(json.dumps(event) + "\n", encoding="utf-8")
+    return session
+
+
+def _git_remote_stub(url: str | None, workdir: Path):
+    """A ``_command_stdout`` stub answering only the origin-remote lookup."""
+
+    def fake(*args: str) -> str | None:
+        if args == ("git", "-C", str(workdir), "remote", "get-url", "origin"):
+            return url
+        return None
+
+    return fake
+
+
+def _capture_broker_source_id(
+    monkeypatch: pytest.MonkeyPatch, captured: dict[str, str]
+) -> None:
+    def fake_upload(staged, *, broker_url, on_file_complete, on_bytes):
+        captured["source_id"] = staged.manifest["source_id"]
+        for staged_file in staged.files:
+            on_file_complete(staged_file)
+        return SimpleNamespace(url=broker_url, uploaded=("payload",), skipped=())
+
+    monkeypatch.setattr(
+        "benchflow.publish.broker.upload_capture_via_broker", fake_upload
+    )
+
+
+@pytest.mark.parametrize(
+    ("event_for_cwd", "remote"),
+    [
+        (
+            lambda cwd: {"type": "user", "cwd": cwd, "message": {"content": "hi"}},
+            "https://github.com/benchflow-ai/benchflow.git",
+        ),
+        (
+            lambda cwd: {"type": "session_meta", "payload": {"cwd": cwd}},
+            "git@github.com:benchflow-ai/benchflow.git",
+        ),
+    ],
+    ids=["claude-https", "codex-ssh"],
+)
+def test_upload_tags_source_id_with_the_session_repo(
+    tmp_path: Path, monkeypatch: pytest.MonkeyPatch, event_for_cwd, remote: str
+) -> None:
+    """Guards PR #1015 (repo tagging follow-up to PR #1013): by default the
+    upload's source id becomes ``repo/<owner>/<name>``, resolved from the
+    session's recorded cwd git remote, and the CLI prints the repo line."""
+    workdir = tmp_path / "workdir"
+    workdir.mkdir()
+    session = _session_with_cwd(tmp_path, event_for_cwd(str(workdir)))
+    monkeypatch.setattr(
+        "benchflow.cli.traj._command_stdout", _git_remote_stub(remote, workdir)
+    )
+    captured: dict[str, str] = {}
+    _capture_broker_source_id(monkeypatch, captured)
+
+    result = runner.invoke(app, _upload_command(session))
+
+    assert result.exit_code == 0, result.output
+    assert captured["source_id"] == "repo/benchflow-ai/benchflow"
+    assert REPO_LINE in click.unstyle(result.output)
+    assert str(workdir) not in captured["source_id"]
+
+
+def test_no_repo_keeps_the_default_source_id(
+    tmp_path: Path, monkeypatch: pytest.MonkeyPatch
+) -> None:
+    """Guards PR #1015 (repo tagging follow-up to PR #1013): ``--no-repo``
+    opts out even when a repo is detectable, keeping the path-derived id."""
+    workdir = tmp_path / "workdir"
+    workdir.mkdir()
+    session = _session_with_cwd(
+        tmp_path, {"type": "user", "cwd": str(workdir), "message": {"content": "hi"}}
+    )
+    monkeypatch.setattr(
+        "benchflow.cli.traj._command_stdout",
+        _git_remote_stub("https://github.com/benchflow-ai/benchflow.git", workdir),
+    )
+    captured: dict[str, str] = {}
+    _capture_broker_source_id(monkeypatch, captured)
+
+    result = runner.invoke(app, _upload_command(session, "--no-repo"))
+
+    assert result.exit_code == 0, result.output
+    assert captured["source_id"] == "session"
+    assert "Repo:" not in click.unstyle(result.output)
+
+
+def test_explicit_source_id_wins_over_repo_tagging(
+    tmp_path: Path, monkeypatch: pytest.MonkeyPatch
+) -> None:
+    """Guards PR #1015 (repo tagging follow-up to PR #1013): an explicit
+    ``--source-id`` always wins and repo detection never runs."""
+
+    def fail_detection(path: Path) -> str | None:
+        raise AssertionError("repo detection ran despite explicit --source-id")
+
+    monkeypatch.setattr("benchflow.cli.traj._detect_repo_slug", fail_detection)
+    captured: dict[str, str] = {}
+    _capture_broker_source_id(monkeypatch, captured)
+
+    result = runner.invoke(
+        app, _upload_command(_trial(tmp_path), "--source-id", "my-project/run-42")
+    )
+
+    assert result.exit_code == 0, result.output
+    assert captured["source_id"] == "my-project/run-42"
+    assert "Repo:" not in click.unstyle(result.output)
+
+
+def test_undetectable_repo_falls_back_silently(
+    tmp_path: Path, monkeypatch: pytest.MonkeyPatch
+) -> None:
+    """Guards PR #1015 (repo tagging follow-up to PR #1013): a session with
+    no recorded cwd and no reachable git remote uploads with the default
+    source id, no repo line, and no error."""
+    session = _session_with_cwd(tmp_path, {"type": "message", "text": "demo"})
+    monkeypatch.setattr("benchflow.cli.traj._command_stdout", lambda *_args: None)
+    captured: dict[str, str] = {}
+    _capture_broker_source_id(monkeypatch, captured)
+
+    result = runner.invoke(app, _upload_command(session))
+
+    assert result.exit_code == 0, result.output
+    assert captured["source_id"] == "session"
+    assert "Repo:" not in click.unstyle(result.output)