benchflow 0.7.2.dev1986__py3-none-any.whl → 0.7.2.dev1987__py3-none-any.whl
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- benchflow/cli/traj.py +23 -10
- {benchflow-0.7.2.dev1986.dist-info → benchflow-0.7.2.dev1987.dist-info}/METADATA +14 -11
- {benchflow-0.7.2.dev1986.dist-info → benchflow-0.7.2.dev1987.dist-info}/RECORD +6 -6
- {benchflow-0.7.2.dev1986.dist-info → benchflow-0.7.2.dev1987.dist-info}/WHEEL +0 -0
- {benchflow-0.7.2.dev1986.dist-info → benchflow-0.7.2.dev1987.dist-info}/entry_points.txt +0 -0
- {benchflow-0.7.2.dev1986.dist-info → benchflow-0.7.2.dev1987.dist-info}/licenses/LICENSE +0 -0
benchflow/cli/traj.py
CHANGED
|
@@ -54,14 +54,26 @@ SKILL_RAW_URL = (
|
|
|
54
54
|
"/.agents/skills/benchflow-traj-upload/SKILL.md"
|
|
55
55
|
)
|
|
56
56
|
|
|
57
|
+
UPGRADE_COMMAND = "uv tool install --python 3.12 --upgrade --force benchflow"
|
|
58
|
+
|
|
59
|
+
# Three unwrapped lines separated by blank lines: each logical unit is one
|
|
60
|
+
# full physical line (no hard wraps mid-sentence), so the URL and the upgrade
|
|
61
|
+
# command stay selectable in agents and terminals.
|
|
57
62
|
CONTRIBUTOR_PROMPT = (
|
|
58
|
-
"Submit my
|
|
59
|
-
"
|
|
60
|
-
|
|
61
|
-
"
|
|
63
|
+
"Submit my relevant local Claude Code, Codex, OpenCode, or Cursor session "
|
|
64
|
+
"from the re:Agent e2e agentic science / ai4bio hackathon (last 72 hours, "
|
|
65
|
+
"on this laptop) to the BenchFlow eval prize."
|
|
66
|
+
"\n\n"
|
|
67
|
+
f"1. First make sure the latest benchflow CLI is installed: {UPGRADE_COMMAND}"
|
|
68
|
+
"\n\n"
|
|
69
|
+
f"2. Then read {SKILL_RAW_URL} "
|
|
70
|
+
"and follow it: find a session, open the viewer, and upload only after "
|
|
71
|
+
"I (the human) review it."
|
|
62
72
|
)
|
|
63
73
|
|
|
64
|
-
|
|
74
|
+
CONTRIBUTOR_PROMPT_FRAMING = (
|
|
75
|
+
"Send this to your coding agent (it's a prompt for the agent, not steps for you):"
|
|
76
|
+
)
|
|
65
77
|
|
|
66
78
|
|
|
67
79
|
def _fetch_latest_version() -> str | None:
|
|
@@ -166,14 +178,14 @@ def register_traj(app: typer.Typer) -> None:
|
|
|
166
178
|
] = False,
|
|
167
179
|
prompt_only: Annotated[
|
|
168
180
|
bool,
|
|
169
|
-
typer.Option("--prompt", help="Print the copy-paste agent
|
|
181
|
+
typer.Option("--prompt", help="Print the copy-paste agent prompt and exit"),
|
|
170
182
|
] = False,
|
|
171
183
|
list_sessions: Annotated[
|
|
172
184
|
bool,
|
|
173
185
|
typer.Option("--list", help="List recent local sessions and exit"),
|
|
174
186
|
] = False,
|
|
175
187
|
) -> None:
|
|
176
|
-
"""Install the submit skill, or print the
|
|
188
|
+
"""Install the submit skill, or print the prompt to send to an agent."""
|
|
177
189
|
_maybe_print_update_hint()
|
|
178
190
|
if prompt_only:
|
|
179
191
|
_print_contributor_prompt()
|
|
@@ -194,7 +206,6 @@ def register_traj(app: typer.Typer) -> None:
|
|
|
194
206
|
_run_npx_skill_install()
|
|
195
207
|
else:
|
|
196
208
|
_install_project_skill(Path.cwd())
|
|
197
|
-
console.print("Paste this to your agent:")
|
|
198
209
|
_print_contributor_prompt()
|
|
199
210
|
if interactive and typer.confirm(
|
|
200
211
|
"List recent sessions and open the viewer now?", default=False
|
|
@@ -625,7 +636,9 @@ def _print_dry_run(staged: StagedCapture) -> None:
|
|
|
625
636
|
|
|
626
637
|
|
|
627
638
|
def _print_contributor_prompt() -> None:
|
|
628
|
-
#
|
|
639
|
+
# Plain print: Rich wrapping would break the unbroken URL line and make
|
|
640
|
+
# the block awkward to copy.
|
|
641
|
+
print(CONTRIBUTOR_PROMPT_FRAMING)
|
|
629
642
|
print(CONTRIBUTOR_PROMPT)
|
|
630
643
|
|
|
631
644
|
|
|
@@ -671,7 +684,7 @@ def _run_npx_skill_install() -> None:
|
|
|
671
684
|
check=False,
|
|
672
685
|
)
|
|
673
686
|
if completed.returncode != 0:
|
|
674
|
-
print_error("npx skills add failed; the copy-paste
|
|
687
|
+
print_error("npx skills add failed; the copy-paste prompt below still works.")
|
|
675
688
|
|
|
676
689
|
|
|
677
690
|
def _print_session_hits() -> None:
|
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
Metadata-Version: 2.5
|
|
2
2
|
Name: benchflow
|
|
3
|
-
Version: 0.7.2.
|
|
3
|
+
Version: 0.7.2.dev1987
|
|
4
4
|
Summary: Multi-turn agent benchmarking with ACP — run any agent, any model, any provider.
|
|
5
5
|
Project-URL: Homepage, https://github.com/benchflow-ai/benchflow
|
|
6
6
|
Project-URL: Repository, https://github.com/benchflow-ai/benchflow
|
|
@@ -89,20 +89,23 @@ BenchFlow is a universal environment framework: it runs AI agents against task e
|
|
|
89
89
|
|
|
90
90
|
## Quick start: 1. Submit a trajectory
|
|
91
91
|
|
|
92
|
-
|
|
93
|
-
|
|
94
|
-
BenchFlow account, API key, or
|
|
92
|
+
**Don't follow these steps yourself — send them to your coding agent.** Copy
|
|
93
|
+
the block below and paste it as a message to Claude Code, Codex, OpenCode, or
|
|
94
|
+
Cursor. That is the $2,000 eval prize path. No BenchFlow account, API key, or
|
|
95
|
+
Azure login.
|
|
95
96
|
|
|
96
|
-
|
|
97
|
-
|
|
98
|
-
|
|
97
|
+
> Submit my relevant local Claude Code, Codex, OpenCode, or Cursor session from the re:Agent e2e agentic science / ai4bio hackathon (last 72 hours, on this laptop) to the BenchFlow eval prize.
|
|
98
|
+
>
|
|
99
|
+
> 1. First make sure the latest benchflow CLI is installed: uv tool install --python 3.12 --upgrade --force benchflow
|
|
100
|
+
>
|
|
101
|
+
> 2. Then read https://raw.githubusercontent.com/benchflow-ai/benchflow/main/.agents/skills/benchflow-traj-upload/SKILL.md and follow it: find a session, open the viewer, and upload only after I (the human) review it.
|
|
99
102
|
|
|
100
103
|
The agent finds sessions on this machine, opens the viewer, and uploads after
|
|
101
104
|
you like what you see.
|
|
102
105
|
|
|
103
|
-
|
|
104
|
-
--force benchflow
|
|
105
|
-
when a newer release is available.
|
|
106
|
+
The prompt has the agent install the latest BenchFlow first (`uv tool
|
|
107
|
+
install --python 3.12 --upgrade --force benchflow`). The `bench traj` commands
|
|
108
|
+
also print a one-line upgrade hint when a newer release is available.
|
|
106
109
|
|
|
107
110
|
Prefer the terminal instead? The guided upload inspects before anything leaves
|
|
108
111
|
your machine — it renders a redacted trajectory report (step counts, masked
|
|
@@ -128,7 +131,7 @@ bench traj setup
|
|
|
128
131
|
```
|
|
129
132
|
|
|
130
133
|
`npx skills add` asks which agents to install for. `bench traj setup` copies
|
|
131
|
-
the skill into this project and prints the same
|
|
134
|
+
the skill into this project and prints the same agent prompt. See the
|
|
132
135
|
[upload skill](./.agents/skills/benchflow-traj-upload/SKILL.md).
|
|
133
136
|
|
|
134
137
|
## Quick start: 2. Run with a ChatGPT or Claude subscription
|
|
@@ -116,7 +116,7 @@ benchflow/cli/skills.py,sha256=2Apv1iqvmstWEhY6O-PRdxKc8RYwR6f7MvbYfl2dWm0,7497
|
|
|
116
116
|
benchflow/cli/tasks.py,sha256=b1XP1EZ79ZtK_1shF65qZyv27EcKKFZYV35M4DzxYKI,18534
|
|
117
117
|
benchflow/cli/trace_import.py,sha256=NHDieFdiKQ-2Sr1-GF8Z5EDcbGsz2qq_kFE0rXacC2o,14705
|
|
118
118
|
benchflow/cli/train.py,sha256=MIP_JxptitH348bWWURpcWt-TmDAw1s8d-RUU7jBaIg,23635
|
|
119
|
-
benchflow/cli/traj.py,sha256=
|
|
119
|
+
benchflow/cli/traj.py,sha256=tX4mtAXbEwEhbwegUTHl_IBIFoxKtVlrwrlv8MPf764,24290
|
|
120
120
|
benchflow/continue_run/__init__.py,sha256=dWidng7kJphVUtbP1a1Ma0Lyjxj5yLgQZIApaKCJUew,1479
|
|
121
121
|
benchflow/continue_run/batch.py,sha256=8BNSInyp_Ol6Fcrg4ezBJdi3cBS12nC8mf963_x4fIg,4146
|
|
122
122
|
benchflow/continue_run/orchestrator.py,sha256=wYoseBNE_rZ0EjNwWDyqfave54xNRe43sCd_O4VJjyk,27676
|
|
@@ -292,8 +292,8 @@ benchflow/trajectories/tree.py,sha256=c0jyoP9OurDQq-zRtPc66edzWUvLGwQQz0jgtyqSw2
|
|
|
292
292
|
benchflow/trajectories/trl_sft_tokenization.py,sha256=-Ef0bP9cuzfzOz5nIv2mEF3taQIEw68h6SM1YO-Mwn4,8643
|
|
293
293
|
benchflow/trajectories/types.py,sha256=aSJclec3dOCUlfFNDL8GfzvkAcBo5r-dvlHE0MnGdMo,27554
|
|
294
294
|
benchflow/trajectories/viewer.py,sha256=59wr8pcWWsCYzuI_sFzTSFJFHeXzH6Yklrh8UsF1ksg,26561
|
|
295
|
-
benchflow-0.7.2.
|
|
296
|
-
benchflow-0.7.2.
|
|
297
|
-
benchflow-0.7.2.
|
|
298
|
-
benchflow-0.7.2.
|
|
299
|
-
benchflow-0.7.2.
|
|
295
|
+
benchflow-0.7.2.dev1987.dist-info/METADATA,sha256=9Bg2Qyc8MQrfqiWzedGTC7QHfPJq4CxxyncVLbnpmAA,15602
|
|
296
|
+
benchflow-0.7.2.dev1987.dist-info/WHEEL,sha256=zOwg4jB6zX2kU910N-cMawjivD6tO8NEWvE12je1bVk,87
|
|
297
|
+
benchflow-0.7.2.dev1987.dist-info/entry_points.txt,sha256=qPMBVtH4qz3P4nMALEAdTbvpxvsVch9RhqshZKQn1dE,84
|
|
298
|
+
benchflow-0.7.2.dev1987.dist-info/licenses/LICENSE,sha256=ohnbHkuBylrTtOvaZJk9Hz7dRwmaEEGAb5ZcBV4Xuhw,10779
|
|
299
|
+
benchflow-0.7.2.dev1987.dist-info/RECORD,,
|
|
File without changes
|
|
File without changes
|
|
File without changes
|