Datasets:
Commit ·
616230d
1
Parent(s): ca5608b
Add traces config: Hub Session Traces (STS) files
Browse filesOne STS session file per trajectory (403 train, 34 test) generated by
scripts/to_sts.py from the cleaned trajectories, so the rows open in the
Hub's agent trace viewer. Content is identical to the trajectories config;
each file is verified to convert back to its source row.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
This view is limited to 50 files because it contains too many changes. See raw diff
- README.md +39 -2
- scripts/to_sts.py +134 -0
- traces/test/019e0337-5711-7a12-8527-47a6e56f83bc_1.jsonl +3 -0
- traces/test/019e111f-2cc7-7d43-b136-abe7afc4c58f_1.jsonl +3 -0
- traces/test/019e12ea-4bed-7ff0-8ea3-c19dcbb83790_1.jsonl +3 -0
- traces/test/019e1561-fb9c-7c80-8b06-1af917eebc3b_1.jsonl +3 -0
- traces/test/019e1757-3040-7ad2-92f3-cd00144574bc_1.jsonl +3 -0
- traces/test/019e1771-3f88-7693-ad68-a908363df442_1.jsonl +3 -0
- traces/test/019e1890-4125-7b50-8a69-e2bfa2f351d5_1.jsonl +3 -0
- traces/test/019e1acb-5db1-7383-9235-b448247e6196_1.jsonl +3 -0
- traces/test/019e1bbd-727e-7f02-97fc-7b6fb200a671_1.jsonl +3 -0
- traces/test/019e1c0e-2092-7990-aaac-0053e8464965_1.jsonl +3 -0
- traces/test/019e2321-efe8-7a52-a31e-28c5ca7fe879_1.jsonl +3 -0
- traces/test/019e2403-9ad5-7b32-a346-a002e59b2f47_1.jsonl +3 -0
- traces/test/019e26b9-7d12-7de0-88cb-e822ba1968a5_1.jsonl +3 -0
- traces/test/019e26b9-a610-74a3-867b-acc5d74209a2_1.jsonl +3 -0
- traces/test/019e31c3-39a3-7d20-a930-338226745928_1.jsonl +3 -0
- traces/test/019e3247-59f6-7482-bc64-0d0ef23d428d_1.jsonl +3 -0
- traces/test/019e3271-11f6-7a10-913d-e1b27ecb0f80_1.jsonl +3 -0
- traces/test/019e3451-db53-7780-8281-f0bffd32d3e9_1.jsonl +3 -0
- traces/test/019e35d6-08f7-7af2-9c3e-e9984232d116_1.jsonl +3 -0
- traces/test/019e35d6-58f9-70c3-bbf8-5c528a90076f_1.jsonl +3 -0
- traces/test/019e3a6e-18de-7e31-9f35-be614569f671_1.jsonl +3 -0
- traces/test/019e3a7e-2550-7fa2-8de5-02cc71ed8395_1.jsonl +3 -0
- traces/test/019e3b71-ec69-7490-8cfc-afe6c1bedf6c_1.jsonl +3 -0
- traces/test/019e3b85-2b4e-7b50-8e9a-f389bb2777eb_1.jsonl +3 -0
- traces/test/019e3c19-ab7d-7021-a516-5f72f9857d03_1.jsonl +3 -0
- traces/test/019e445f-f408-7472-a7bd-62a8c01d5ba3_1.jsonl +3 -0
- traces/test/019e450d-33d8-7581-80c2-c2a5a3990bc5_1.jsonl +3 -0
- traces/test/019e4ba7-fa8d-7b52-a382-25a1963d8a79_1.jsonl +3 -0
- traces/test/019e5d89-a004-77a0-a857-e61c0082ba30_1.jsonl +3 -0
- traces/test/019e618f-bdaf-7791-9faa-ad64f9881ece_1.jsonl +3 -0
- traces/test/019e6c49-5dee-7153-8d69-c9b6087ded8b_1.jsonl +3 -0
- traces/test/019e77a8-b85f-7003-b1f4-c0e79b7d2fdb_1.jsonl +3 -0
- traces/test/019e7f5f-f2ac-7ad1-97b0-8f65f14a7f25_1.jsonl +3 -0
- traces/test/019e7f5f-f2ac-7ad1-97b0-8f65f14a7f25_2.jsonl +3 -0
- traces/train/019dfe14-09bb-7063-bd51-9fa96185c593_1.jsonl +3 -0
- traces/train/019dfe18-bdde-71b1-96af-031aba2b977b_1.jsonl +3 -0
- traces/train/019dfe19-bae0-7471-8c0e-491f25d39c7e_1.jsonl +3 -0
- traces/train/019dfe1d-8537-7073-a805-07c4fda91d06_1.jsonl +3 -0
- traces/train/019dfe1f-6d7b-70d2-a285-0da926506c6b_1.jsonl +3 -0
- traces/train/019dfe2b-5f3f-7e42-8a16-08b6b356e6af_1.jsonl +3 -0
- traces/train/019dfe36-ae14-72d3-9ede-da5e0442740f_1.jsonl +3 -0
- traces/train/019dfe4e-3fb2-7391-a40d-886e80b2b448_1.jsonl +3 -0
- traces/train/019dfe58-7b79-75b0-9522-315e87a99336_1.jsonl +3 -0
- traces/train/019dfe5a-9b15-71a0-8c81-8dd758f7e5c7_1.jsonl +3 -0
- traces/train/019dfe5f-0f09-75a2-846e-a91ed5b88b49_1.jsonl +3 -0
- traces/train/019dfe60-6db8-79b0-9ad2-dddff8cdb5a0_1.jsonl +3 -0
- traces/train/019dfe69-5d30-7191-b88f-0cf6bd3be3b5_1.jsonl +3 -0
- traces/train/019dfe74-7cda-75f1-978b-e2341e082ebd_1.jsonl +3 -0
README.md
CHANGED
|
@@ -30,6 +30,12 @@ configs:
|
|
| 30 |
path: pairs.train.jsonl
|
| 31 |
- split: test
|
| 32 |
path: pairs.eval.jsonl
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 33 |
---
|
| 34 |
|
| 35 |
# Codex Tool-Use Trajectories
|
|
@@ -41,6 +47,9 @@ Windows workspace, cleaned and redacted from raw session logs.
|
|
| 41 |
compaction), each a full chat with tool calls and tool outputs.
|
| 42 |
- **`pairs`**: 314 delegation examples: the brief a parent agent gave a subagent and the final
|
| 43 |
report the subagent returned.
|
|
|
|
|
|
|
|
|
|
| 44 |
|
| 45 |
The sessions were recorded between 2026-05-06 and 2026-06-01 on one developer's machine. Most of
|
| 46 |
the work is C#/C++/PowerShell development and binary reverse engineering of a video codec.
|
|
@@ -58,15 +67,23 @@ see [Acknowledgements](#acknowledgements).
|
|
| 58 |
| `trajectories` | test | 34 | 33 | 3.6 MB |
|
| 59 |
| `pairs` | train | 284 | 284 | 1.3 MB |
|
| 60 |
| `pairs` | test | 30 | 30 | 0.1 MB |
|
|
|
|
|
|
|
| 61 |
|
| 62 |
```python
|
| 63 |
from datasets import load_dataset
|
| 64 |
|
| 65 |
traj = load_dataset("israellaguan/codex-tool-use-trajectories", "trajectories")
|
| 66 |
pairs = load_dataset("israellaguan/codex-tool-use-trajectories", "pairs")
|
|
|
|
| 67 |
```
|
| 68 |
|
| 69 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 70 |
(seed 42, about 10% of sessions held out), so no session appears in both splits.
|
| 71 |
See the note on split independence under [Considerations](#considerations-for-using-the-data).
|
| 72 |
|
|
@@ -124,6 +141,26 @@ See the note on split independence under [Considerations](#considerations-for-us
|
|
| 124 |
median 730 characters) and `report` (its final message; median 3,423 characters).
|
| 125 |
There are no pairs for the 6 reconstructed rows.
|
| 126 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 127 |
### Auxiliary files
|
| 128 |
|
| 129 |
| file | purpose |
|
|
@@ -131,7 +168,7 @@ There are no pairs for the 6 reconstructed rows.
|
|
| 131 |
| `redactions.jsonl` | One entry per redaction: row id, field path, kind, replacement, and the SHA-256 of the original value. Original values are not stored. |
|
| 132 |
| `manifest.json` | Counts and reconciliation numbers from the build. |
|
| 133 |
| `REPORT.md` | Full cleaning report: funnel, validation checks, manual recoveries, sample rows. |
|
| 134 |
-
| `scripts/` | The pipeline (`build.py`, `validate.py`, `report.py`, `redact.py`, tests). They read the raw logs, which are not included. |
|
| 135 |
|
| 136 |
## Dataset creation
|
| 137 |
|
|
|
|
| 30 |
path: pairs.train.jsonl
|
| 31 |
- split: test
|
| 32 |
path: pairs.eval.jsonl
|
| 33 |
+
- config_name: traces
|
| 34 |
+
data_files:
|
| 35 |
+
- split: train
|
| 36 |
+
path: traces/train/*.jsonl
|
| 37 |
+
- split: test
|
| 38 |
+
path: traces/test/*.jsonl
|
| 39 |
---
|
| 40 |
|
| 41 |
# Codex Tool-Use Trajectories
|
|
|
|
| 47 |
compaction), each a full chat with tool calls and tool outputs.
|
| 48 |
- **`pairs`**: 314 delegation examples: the brief a parent agent gave a subagent and the final
|
| 49 |
report the subagent returned.
|
| 50 |
+
- **`traces`**: the same 437 runs as `trajectories`, one session per file in the Hub's
|
| 51 |
+
[Session Traces Format](https://huggingface.co/docs/hub/session-traces-format), so they open in
|
| 52 |
+
the Hub's agent trace viewer.
|
| 53 |
|
| 54 |
The sessions were recorded between 2026-05-06 and 2026-06-01 on one developer's machine. Most of
|
| 55 |
the work is C#/C++/PowerShell development and binary reverse engineering of a video codec.
|
|
|
|
| 67 |
| `trajectories` | test | 34 | 33 | 3.6 MB |
|
| 68 |
| `pairs` | train | 284 | 284 | 1.3 MB |
|
| 69 |
| `pairs` | test | 30 | 30 | 0.1 MB |
|
| 70 |
+
| `traces` | train | 403 | 301 | 41.4 MB |
|
| 71 |
+
| `traces` | test | 34 | 33 | 3.6 MB |
|
| 72 |
|
| 73 |
```python
|
| 74 |
from datasets import load_dataset
|
| 75 |
|
| 76 |
traj = load_dataset("israellaguan/codex-tool-use-trajectories", "trajectories")
|
| 77 |
pairs = load_dataset("israellaguan/codex-tool-use-trajectories", "pairs")
|
| 78 |
+
traces = load_dataset("israellaguan/codex-tool-use-trajectories", "traces")
|
| 79 |
```
|
| 80 |
|
| 81 |
+
`traces` has the same content as `trajectories` in a different layout; use `trajectories` for
|
| 82 |
+
training and `traces` for browsing. Each row of `trajectories` is one file in `traces`, so a
|
| 83 |
+
split's rows and files match one to one. Loading `traces` with `datasets` makes the library ask
|
| 84 |
+
for the `teich` package, which it uses to decode agent traces.
|
| 85 |
+
|
| 86 |
+
The `test` split is stored in the `*.eval.jsonl` files and in `traces/test/`. Splits were made **by session id**
|
| 87 |
(seed 42, about 10% of sessions held out), so no session appears in both splits.
|
| 88 |
See the note on split independence under [Considerations](#considerations-for-using-the-data).
|
| 89 |
|
|
|
|
| 141 |
median 730 characters) and `report` (its final message; median 3,423 characters).
|
| 142 |
There are no pairs for the 6 reconstructed rows.
|
| 143 |
|
| 144 |
+
### `traces` format
|
| 145 |
+
|
| 146 |
+
`traces/train/` and `traces/test/` hold one file per trajectory, named `<turn_id>_<segment>.jsonl`.
|
| 147 |
+
Each file is a session in the Hub's STS format: a header line, then one line per message.
|
| 148 |
+
|
| 149 |
+
```json
|
| 150 |
+
{"type": "session", "harness": "codex-desktop", "id": "<turn_id>#<segment>", "name": "<first line of the task>", "session_id": "...", "model": "gpt-5.5", "cwd": "...", "flags": [], "...": "..."}
|
| 151 |
+
{"type": "message", "message": {"role": "assistant", "content": "", "model": "gpt-5.5", "toolCalls": [{"id": "call_...", "function": {"name": "shell_command", "arguments": "{...}"}}]}}
|
| 152 |
+
{"type": "message", "message": {"role": "tool", "toolCallId": "call_...", "content": "Exit code: 0\n..."}}
|
| 153 |
+
```
|
| 154 |
+
|
| 155 |
+
- The header carries all the metadata fields of `trajectories` (`session_id`, `turn_id`, `segment`,
|
| 156 |
+
`thread_source`, `agent_role`, `model`, `effort`, `cwd`, `timestamp`, `task_source`, `flags`, ...)
|
| 157 |
+
next to the STS fields.
|
| 158 |
+
- `harness` is `codex-desktop`, the agent that produced the sessions.
|
| 159 |
+
- Fields are renamed `tool_calls` → `toolCalls` and `tool_call_id` → `toolCallId`, and a `null`
|
| 160 |
+
content becomes `""`. Nothing else changes: the text is identical to `trajectories`.
|
| 161 |
+
`scripts/to_sts.py` generates the files and checks that each converts back to its source row.
|
| 162 |
+
- STS has no per-message timestamps, so only the turn start (`timestamp` in the header) is kept.
|
| 163 |
+
|
| 164 |
### Auxiliary files
|
| 165 |
|
| 166 |
| file | purpose |
|
|
|
|
| 168 |
| `redactions.jsonl` | One entry per redaction: row id, field path, kind, replacement, and the SHA-256 of the original value. Original values are not stored. |
|
| 169 |
| `manifest.json` | Counts and reconciliation numbers from the build. |
|
| 170 |
| `REPORT.md` | Full cleaning report: funnel, validation checks, manual recoveries, sample rows. |
|
| 171 |
+
| `scripts/` | The pipeline (`build.py`, `validate.py`, `report.py`, `redact.py`, tests). They read the raw logs, which are not included. `to_sts.py` reads only the cleaned `trajectories.*.jsonl` and writes `traces/`. |
|
| 172 |
|
| 173 |
## Dataset creation
|
| 174 |
|
scripts/to_sts.py
ADDED
|
@@ -0,0 +1,134 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
"""Convert the cleaned trajectories into Hugging Face Session Trace (STS) files.
|
| 2 |
+
|
| 3 |
+
Reads : trajectories.train.jsonl, trajectories.eval.jsonl (the cleaned output of build.py; raw logs are NOT read)
|
| 4 |
+
Writes : traces/train/<id>.jsonl, traces/test/<id>.jsonl (one session per file, one file per trajectory row)
|
| 5 |
+
|
| 6 |
+
Format : https://huggingface.co/docs/hub/session-traces-format
|
| 7 |
+
line 1 {"type": "session", "harness": ..., "id": ..., "name": ..., <extra metadata>}
|
| 8 |
+
line 2+ {"type": "message", "message": {"role", "content", "toolCalls"?, "toolCallId"?, "model"?}}
|
| 9 |
+
|
| 10 |
+
Mapping : tool_calls -> toolCalls (the redundant "type": "function" is dropped), tool_call_id -> toolCallId,
|
| 11 |
+
content null -> "" (the spec requires a string), assistant messages get the row's model when it is known.
|
| 12 |
+
Nothing is redacted, truncated or reordered here: the text is exactly what is in trajectories.*.jsonl.
|
| 13 |
+
|
| 14 |
+
Every file is converted back and compared with its source row right after it is written; any difference raises and
|
| 15 |
+
the script exits non-zero. `python3 -I scripts/to_sts.py` is idempotent: it recreates traces/ from scratch.
|
| 16 |
+
"""
|
| 17 |
+
import json
|
| 18 |
+
import os
|
| 19 |
+
import re
|
| 20 |
+
import shutil
|
| 21 |
+
import sys
|
| 22 |
+
|
| 23 |
+
ROOT = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
|
| 24 |
+
OUT = os.path.join(ROOT, "traces")
|
| 25 |
+
SPLITS = {"train": "trajectories.train.jsonl", "test": "trajectories.eval.jsonl"}
|
| 26 |
+
HARNESS = "codex-desktop"
|
| 27 |
+
HEADER_META = ["session_id", "turn_id", "segment", "segments_total", "compaction_summary_available", "thread_source",
|
| 28 |
+
"agent_role", "model", "effort", "cwd", "timestamp", "task_source", "n_tool_calls", "n_failed_calls", "flags"]
|
| 29 |
+
MSG_KEYS = {"role", "content", "tool_calls", "tool_call_id"}
|
| 30 |
+
TITLE_MAX = 100
|
| 31 |
+
|
| 32 |
+
|
| 33 |
+
REQUEST_MARKER = "## My request for Codex:" # the Codex app prepends its context (open browser, URL) before this line
|
| 34 |
+
|
| 35 |
+
|
| 36 |
+
def title(row):
|
| 37 |
+
for m in row["messages"]:
|
| 38 |
+
if m["role"] == "user" and (m.get("content") or "").strip():
|
| 39 |
+
text = m["content"]
|
| 40 |
+
if REQUEST_MARKER in text:
|
| 41 |
+
text = text.split(REQUEST_MARKER, 1)[1]
|
| 42 |
+
for line in text.splitlines():
|
| 43 |
+
line = re.sub(r"\s+", " ", line).strip()
|
| 44 |
+
if line:
|
| 45 |
+
return line if len(line) <= TITLE_MAX else line[:TITLE_MAX - 1].rstrip() + "…"
|
| 46 |
+
return row["id"]
|
| 47 |
+
|
| 48 |
+
|
| 49 |
+
def convert_message(m, model):
|
| 50 |
+
extra = set(m) - MSG_KEYS
|
| 51 |
+
if extra:
|
| 52 |
+
raise ValueError(f"unexpected message keys {sorted(extra)}")
|
| 53 |
+
out = {"role": m["role"], "content": m.get("content") or ""}
|
| 54 |
+
if m.get("tool_calls"):
|
| 55 |
+
calls = []
|
| 56 |
+
for tc in m["tool_calls"]:
|
| 57 |
+
if set(tc) != {"id", "type", "function"} or tc["type"] != "function" or set(tc["function"]) != {"name", "arguments"}:
|
| 58 |
+
raise ValueError(f"unexpected tool call shape {sorted(tc)}")
|
| 59 |
+
calls.append({"id": tc["id"], "function": {"name": tc["function"]["name"], "arguments": tc["function"]["arguments"]}})
|
| 60 |
+
out["toolCalls"] = calls
|
| 61 |
+
if m.get("tool_call_id"):
|
| 62 |
+
out["toolCallId"] = m["tool_call_id"]
|
| 63 |
+
if m["role"] == "assistant" and model:
|
| 64 |
+
out["model"] = model
|
| 65 |
+
return {"type": "message", "message": out}
|
| 66 |
+
|
| 67 |
+
|
| 68 |
+
def convert_row(row):
|
| 69 |
+
header = {"type": "session", "harness": HARNESS, "id": row["id"], "name": title(row)}
|
| 70 |
+
for k in HEADER_META:
|
| 71 |
+
header[k] = row[k]
|
| 72 |
+
return [header] + [convert_message(m, row["model"]) for m in row["messages"]]
|
| 73 |
+
|
| 74 |
+
|
| 75 |
+
def revert(lines):
|
| 76 |
+
"""Rebuild the original row fields from an STS file (used by the verification)."""
|
| 77 |
+
header, msgs = lines[0], []
|
| 78 |
+
for ln in lines[1:]:
|
| 79 |
+
m = ln["message"]
|
| 80 |
+
o = {"role": m["role"], "content": m["content"]}
|
| 81 |
+
if "toolCalls" in m:
|
| 82 |
+
o["tool_calls"] = [{"id": c["id"], "type": "function", "function": c["function"]} for c in m["toolCalls"]]
|
| 83 |
+
if "toolCallId" in m:
|
| 84 |
+
o["tool_call_id"] = m["toolCallId"]
|
| 85 |
+
msgs.append(o)
|
| 86 |
+
return header, msgs
|
| 87 |
+
|
| 88 |
+
|
| 89 |
+
def norm(m):
|
| 90 |
+
m = dict(m)
|
| 91 |
+
m["content"] = m.get("content") or ""
|
| 92 |
+
return m
|
| 93 |
+
|
| 94 |
+
|
| 95 |
+
def main():
|
| 96 |
+
if os.path.exists(OUT):
|
| 97 |
+
assert os.path.abspath(OUT) == os.path.join(ROOT, "traces")
|
| 98 |
+
shutil.rmtree(OUT)
|
| 99 |
+
names, total, biggest, n_calls = set(), 0, 0, 0
|
| 100 |
+
counts = {}
|
| 101 |
+
for split, fn in SPLITS.items():
|
| 102 |
+
os.makedirs(os.path.join(OUT, split))
|
| 103 |
+
n = 0
|
| 104 |
+
with open(os.path.join(ROOT, fn), encoding="utf-8") as fh:
|
| 105 |
+
for raw in fh:
|
| 106 |
+
row = json.loads(raw)
|
| 107 |
+
lines = convert_row(row)
|
| 108 |
+
name = row["id"].replace("#", "_") + ".jsonl"
|
| 109 |
+
if name in names:
|
| 110 |
+
raise ValueError("duplicate file name " + name)
|
| 111 |
+
names.add(name)
|
| 112 |
+
path = os.path.join(OUT, split, name)
|
| 113 |
+
body = "".join(json.dumps(x, ensure_ascii=False, separators=(",", ":")) + "\n" for x in lines)
|
| 114 |
+
with open(path, "w", encoding="utf-8", newline="\n") as w:
|
| 115 |
+
w.write(body)
|
| 116 |
+
total += len(body.encode("utf-8"))
|
| 117 |
+
biggest = max(biggest, len(body.encode("utf-8")))
|
| 118 |
+
n += 1
|
| 119 |
+
# ---- verification: file -> back to the source row
|
| 120 |
+
with open(path, encoding="utf-8") as r:
|
| 121 |
+
back = [json.loads(x) for x in r]
|
| 122 |
+
assert back[0]["type"] == "session" and back[0]["harness"] and back[0]["id"] == row["id"], path
|
| 123 |
+
assert all(x["type"] == "message" for x in back[1:]), path
|
| 124 |
+
header, msgs = revert(back)
|
| 125 |
+
assert [norm(m) for m in row["messages"]] == msgs, "messages differ: " + path
|
| 126 |
+
assert all(header[k] == row[k] for k in HEADER_META), "header metadata differs: " + path
|
| 127 |
+
n_calls += sum(len(m.get("tool_calls") or []) for m in row["messages"])
|
| 128 |
+
counts[split] = n
|
| 129 |
+
print(json.dumps({"files": counts, "bytes_total": total, "bytes_largest_file": biggest,
|
| 130 |
+
"tool_calls_verified": n_calls, "harness": HARNESS, "verified_round_trip": True}, indent=1))
|
| 131 |
+
|
| 132 |
+
|
| 133 |
+
if __name__ == "__main__":
|
| 134 |
+
sys.exit(main())
|
traces/test/019e0337-5711-7a12-8527-47a6e56f83bc_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3a73f3c1640f42c0ce6af0ef2193d49535bb4fcb2d8999d90507b343988ff47d
|
| 3 |
+
size 151383
|
traces/test/019e111f-2cc7-7d43-b136-abe7afc4c58f_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:88cd41a529291700e7a753a01c7d366274ca3ea067428097eca89484e8f48ea0
|
| 3 |
+
size 44762
|
traces/test/019e12ea-4bed-7ff0-8ea3-c19dcbb83790_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4cc90cd40ded0c47cdb941bc75895a265d88bfff1d01f231bc2e5b33ff6f7c47
|
| 3 |
+
size 32089
|
traces/test/019e1561-fb9c-7c80-8b06-1af917eebc3b_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b33f2d56e07d5263049a451d36c0e9f3a4d15557bf79142e59a6971c7f626b31
|
| 3 |
+
size 33974
|
traces/test/019e1757-3040-7ad2-92f3-cd00144574bc_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8ae2d30c864b78f9bad6ce1d7785a1a66e6e1f118ec2b8dd306d511c1e0fde29
|
| 3 |
+
size 170585
|
traces/test/019e1771-3f88-7693-ad68-a908363df442_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c440dd455054588c2f40376a7b6ce7be4506cd0b6b84bc9c6acab77b3bc006f1
|
| 3 |
+
size 66853
|
traces/test/019e1890-4125-7b50-8a69-e2bfa2f351d5_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e67d49c5c1255988f101057cfa922750349bf56283b36580c46853f7dc5157e9
|
| 3 |
+
size 134681
|
traces/test/019e1acb-5db1-7383-9235-b448247e6196_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:59f85db3372d6062cfd789e616b46236b2ccc41a0c76afff9bc113791695a042
|
| 3 |
+
size 120858
|
traces/test/019e1bbd-727e-7f02-97fc-7b6fb200a671_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:dd578b3c128b13f1a74a4f0ce84cc95de8b2941bd5f9a7b613786de6ae331668
|
| 3 |
+
size 123126
|
traces/test/019e1c0e-2092-7990-aaac-0053e8464965_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:efef8048fc83bee045ef1747a681495ba8aec58365d89c5455df9cf4d3a38b05
|
| 3 |
+
size 124400
|
traces/test/019e2321-efe8-7a52-a31e-28c5ca7fe879_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3c2da6440ac64c5b9062e35df5df0312c708ac6b948be0ed59994e92631cb586
|
| 3 |
+
size 55378
|
traces/test/019e2403-9ad5-7b32-a346-a002e59b2f47_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1469e2e6efa473d320e5fcaab03fb0814fdaef01d5880ddd6f32bf6c88ac15d6
|
| 3 |
+
size 59519
|
traces/test/019e26b9-7d12-7de0-88cb-e822ba1968a5_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0110c5095770ebaba4f54ee38c77d3bc5728cb36eba149fc0c21f57d0941db1d
|
| 3 |
+
size 77340
|
traces/test/019e26b9-a610-74a3-867b-acc5d74209a2_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:247399edd6f3847581df562c424e31dadc89abe83a7b463230881f1cf1635f8a
|
| 3 |
+
size 206753
|
traces/test/019e31c3-39a3-7d20-a930-338226745928_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:89a0d0fcef9573c106ee3348afc7e6f029d793b3ffe2a0fe2f22a22dcf6b4e0a
|
| 3 |
+
size 120594
|
traces/test/019e3247-59f6-7482-bc64-0d0ef23d428d_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:53706819fbcb150652c50b793834087f60ea0ce9dbdd5894b93273b9273982b9
|
| 3 |
+
size 91733
|
traces/test/019e3271-11f6-7a10-913d-e1b27ecb0f80_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d0a1e967949891d62521d04f1b0b3a7ae6f5eee1a8ea168cca3e51f93cde1eb2
|
| 3 |
+
size 95106
|
traces/test/019e3451-db53-7780-8281-f0bffd32d3e9_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ed95029358774cbdc2ccbdfe755df85624af656442157260e84b3f703136739e
|
| 3 |
+
size 225577
|
traces/test/019e35d6-08f7-7af2-9c3e-e9984232d116_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:dd309162038c98ba84abcee7f68c05c0b5e768b117a4112623477710856fa273
|
| 3 |
+
size 176427
|
traces/test/019e35d6-58f9-70c3-bbf8-5c528a90076f_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:25ef4db1d42b6b54f04409c667a058af0e8dd5c423563a00ebf81a2bed6d764e
|
| 3 |
+
size 221792
|
traces/test/019e3a6e-18de-7e31-9f35-be614569f671_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:64a86ae88449657dd7614347ee4fcc0db49678ce9cfe5a08b39b3014ed611c90
|
| 3 |
+
size 69211
|
traces/test/019e3a7e-2550-7fa2-8de5-02cc71ed8395_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:37018480afe618687e1f37cab782a33af026741d98e68b159389b74545b17d28
|
| 3 |
+
size 55079
|
traces/test/019e3b71-ec69-7490-8cfc-afe6c1bedf6c_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:aed6a48b61e33bc6cfae1001626b33ccbe5b465eebe2063666cbbda17ea12361
|
| 3 |
+
size 67402
|
traces/test/019e3b85-2b4e-7b50-8e9a-f389bb2777eb_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1d76b72d6271448d7e0583033e3ce46ad7aa41c158597dafc9a80b05440f6b0c
|
| 3 |
+
size 200209
|
traces/test/019e3c19-ab7d-7021-a516-5f72f9857d03_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3bea2797f1e81f85d1584d67fca2a0d27454837c96b4c11258bd51421e4bd33e
|
| 3 |
+
size 61368
|
traces/test/019e445f-f408-7472-a7bd-62a8c01d5ba3_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:08cf9904b5a0cb4c1ab3c2df1f74f62c268b3b90de4c7dc64e0b719973a27f6d
|
| 3 |
+
size 70143
|
traces/test/019e450d-33d8-7581-80c2-c2a5a3990bc5_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4eae150781ba1a5d853d83d20c5509d1aca779911e906288422db4b9bb5ebacf
|
| 3 |
+
size 101956
|
traces/test/019e4ba7-fa8d-7b52-a382-25a1963d8a79_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7ce1087e4330f4d78088fb0c1c2110d3b9e00be65d2bdbd0cde75b2a53d16c38
|
| 3 |
+
size 112867
|
traces/test/019e5d89-a004-77a0-a857-e61c0082ba30_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e573d6ad4eb6373e3a724c14107e4ad7b3b0b0f2a9a47eb8c4c722df80d261df
|
| 3 |
+
size 76751
|
traces/test/019e618f-bdaf-7791-9faa-ad64f9881ece_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:928a129ad6d7020084413147943b73c25c34f0695110ef1279a9628fc2f7c631
|
| 3 |
+
size 35448
|
traces/test/019e6c49-5dee-7153-8d69-c9b6087ded8b_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b2f2bff780d550f54b3b44b27a9942b70c44f179f0626b577daa57204c68b92b
|
| 3 |
+
size 196162
|
traces/test/019e77a8-b85f-7003-b1f4-c0e79b7d2fdb_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4aef39a5644bb6a37af0b828d0ae3163727b9cc49262023efb8f7375e239ea56
|
| 3 |
+
size 116228
|
traces/test/019e7f5f-f2ac-7ad1-97b0-8f65f14a7f25_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4ea2177e857d42aaf4261023301e96843520b8ba5875acf88455b7a911865cb1
|
| 3 |
+
size 142338
|
traces/test/019e7f5f-f2ac-7ad1-97b0-8f65f14a7f25_2.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f1fe2fa598f3d35d1b7292ea80d430e9456c91cfeef9217b5602d867161ae1d6
|
| 3 |
+
size 8310
|
traces/train/019dfe14-09bb-7063-bd51-9fa96185c593_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0e3688241b8ce58c30ae79a1be04298f13d15e09ca6fdc8d8e1620094ea8e4b1
|
| 3 |
+
size 63507
|
traces/train/019dfe18-bdde-71b1-96af-031aba2b977b_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:eea53f701447ae266941525c32be35703f50f59ecb5cdda51840b91a2f871cf3
|
| 3 |
+
size 10989
|
traces/train/019dfe19-bae0-7471-8c0e-491f25d39c7e_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bd2d641ccff4e7ce81a61ec46e7e657b03db764b5cff071aa36e1365ccadd0bc
|
| 3 |
+
size 79026
|
traces/train/019dfe1d-8537-7073-a805-07c4fda91d06_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:96a07c3c75f5e62bf4760cb8207d45d6a2d3b3297cd87b6421b3dc8b6aabb01b
|
| 3 |
+
size 3401
|
traces/train/019dfe1f-6d7b-70d2-a285-0da926506c6b_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7fc29e2c8da72050f61515ff5744e202d311ec41017c9782a258b9b2e79c01ce
|
| 3 |
+
size 2422
|
traces/train/019dfe2b-5f3f-7e42-8a16-08b6b356e6af_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3301039004850874f7ea3cd038a0af763778ceed5fb35bc8e2a595fcd89380a6
|
| 3 |
+
size 2423
|
traces/train/019dfe36-ae14-72d3-9ede-da5e0442740f_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4bbcf502b97ad0ac6435d4975b370253f02fb48757518eccbef3c6297313f699
|
| 3 |
+
size 11604
|
traces/train/019dfe4e-3fb2-7391-a40d-886e80b2b448_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:53ff617a9283d996a530bbb86dcbed6ef72be56e014cb14ff4e1ea75cb6a004e
|
| 3 |
+
size 13622
|
traces/train/019dfe58-7b79-75b0-9522-315e87a99336_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cab60ef4cc76ca6a1cad923c12a3a204678cd3816990b31b17fb0795464aea31
|
| 3 |
+
size 3744
|
traces/train/019dfe5a-9b15-71a0-8c81-8dd758f7e5c7_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1f23b534f4fa5f02bf74cc6bfe25a97abad9314eac627245ed906bd8bf99068e
|
| 3 |
+
size 10346
|
traces/train/019dfe5f-0f09-75a2-846e-a91ed5b88b49_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:72bc20e246707ba43a3b3b46c37eadc276cf7da0f4b3777ddb5051c5cbf457fc
|
| 3 |
+
size 7405
|
traces/train/019dfe60-6db8-79b0-9ad2-dddff8cdb5a0_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:50a28c4b440208ed4bd38cad1309a09cd6415e5cd371b526240f127b02649480
|
| 3 |
+
size 6386
|
traces/train/019dfe69-5d30-7191-b88f-0cf6bd3be3b5_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ee24a3082deda6b04728ff3da1b939a369eb657c0eaeed77f2a92bc80ded4fb9
|
| 3 |
+
size 64477
|
traces/train/019dfe74-7cda-75f1-978b-e2341e082ebd_1.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1079936c243f45c531fdfb6a51534d6ab23d930d26847113b6a64f6a4a17303c
|
| 3 |
+
size 76733
|