Pin Spark reasoning parser download revision

Author: intern_nem_dev_2

tests/usage_cookbook/test_spark_reasoning_parser_revision.py

@@ -0,0 +1,47 @@
+from __future__ import annotations
+
+import re
+from pathlib import Path
+
+REPO_ROOT = Path(__file__).resolve().parents[2]
+SPARK_GUIDE = REPO_ROOT / "usage-cookbook/Nemotron-3-Super/SparkDeploymentGuide/README.md"
+
+EXPECTED_REPO = "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4"
+EXPECTED_REVISION = "4f0cf9daaeb7a4d5e23f80a00e7ed15f0e03caf6"
+EXPECTED_FILE = "super_v3_reasoning_parser.py"
+EXPECTED_URL = (
+    f"https://huggingface.co/{EXPECTED_REPO}/resolve/{EXPECTED_REVISION}/{EXPECTED_FILE}"
+)
+
+
+def _spark_guide_text() -> str:
+    return SPARK_GUIDE.read_text(encoding="utf-8")
+
+
+def _reasoning_parser_wget_urls(text: str) -> list[str]:
+    return re.findall(r"wget\s+(https://huggingface\.co/\S+/super_v3_reasoning_parser\.py)", text)
+
+
+def test_spark_reasoning_parser_downloads_are_commit_pinned() -> None:
+    urls = _reasoning_parser_wget_urls(_spark_guide_text())
+
+    assert urls == [EXPECTED_URL, EXPECTED_URL]
+    for url in urls:
+        assert EXPECTED_REPO in url
+        assert EXPECTED_REVISION in url
+        assert re.search(r"/resolve/[0-9a-f]{40}/super_v3_reasoning_parser\.py$", url)
+
+
+def test_spark_reasoning_parser_downloads_have_no_floating_main_ref() -> None:
+    text = _spark_guide_text()
+
+    assert "/raw/main/super_v3_reasoning_parser.py" not in text
+    assert "/resolve/main/super_v3_reasoning_parser.py" not in text
+    assert "/main/super_v3_reasoning_parser.py" not in text
+
+
+def test_spark_guide_keeps_vllm_and_trt_llm_sections() -> None:
+    text = _spark_guide_text()
+
+    assert "## vLLM" in text
+    assert "## TensorRT-LLM" in text

usage-cookbook/Nemotron-3-Super/SparkDeploymentGuide/README.md

@@ -19,7 +19,7 @@ Three properties of Nemotron 3 Super that directly affect inference configuratio
 Both vLLM and TRT-LLM require the Nemotron 3 Super reasoning parser. Download it before starting the server:
 
 ```bash
-wget https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4/raw/main/super_v3_reasoning_parser.py
+wget https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4/resolve/4f0cf9daaeb7a4d5e23f80a00e7ed15f0e03caf6/super_v3_reasoning_parser.py
 ```
 
 ---
@@ -35,7 +35,7 @@ vllm/vllm-openai:cu130-nightly
 ### Serve Command
 
 ```bash
-wget https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4/raw/main/super_v3_reasoning_parser.py
+wget https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4/resolve/4f0cf9daaeb7a4d5e23f80a00e7ed15f0e03caf6/super_v3_reasoning_parser.py
 
 docker run --rm -it --gpus all \
   -e VLLM_NVFP4_GEMM_BACKEND=marlin \

workspace/interns/intern_nem_dev_2/status.md

@@ -1,13 +1,13 @@
 # intern_nem_dev_2 - 状态
 
-<!-- METADATA:STATUS=Working,TASK=task168_lora_text2sql_bird_dataset_revision_pins_s1,ROLE=independent -->
+<!-- METADATA:STATUS=Working,TASK=task170_super_spark_reasoning_parser_revision_pin_s1,ROLE=independent -->
 
 | 字段 | 值 |
 |------|-----|
 | Name | intern_nem_dev_2 |
 | Status | Working |
-| Current Task | task168_lora_text2sql_bird_dataset_revision_pins_s1 |
-| PR | https://github.com/songCNMS/Nemotron/pull/275 |
+| Current Task | task170_super_spark_reasoning_parser_revision_pin_s1 |
+| PR | pending |
 | Session | 1 |
 
-最近进展：Opened PR #275 for `task168_lora_text2sql_bird_dataset_revision_pins_s1`: https://github.com/songCNMS/Nemotron/pull/275. Branch was refreshed from `origin/main` base `6328c018a86da7448e11a03bc1c71afc38e067f2`. Added static BIRD Text2SQL dataset revision pins and focused AST tests. Checks passed: focused pytest (`3 passed`), py_compile, Ruff, structured AST probe, added-line live-surface scan, and diff checks. No live dataset download, Text2SQL data prep run, train/eval, endpoint, W&B, cluster, deploy, artifact upload/download, main/master push, or self-merge.
+最近进展：Started `task170_super_spark_reasoning_parser_revision_pin_s1` from `origin/main` base `6500fdaa27735197da87ca25d641a2883b00e8e6` on branch `intern_nem_dev_2/task170_super_spark_reasoning_parser_revision_pin_s1`. Replaced Spark guide reasoning-parser `raw/main` download examples with commit-pinned `resolve/4f0cf9daaeb7a4d5e23f80a00e7ed15f0e03caf6` URLs and added focused static tests; PR creation is pending. No live wget/curl, HF/model download, vLLM/TRT-LLM launch, endpoint, W&B, cluster, deploy, artifact operation, main/master push, or self-merge.

workspace/tasks/task170_super_spark_reasoning_parser_revision_pin_s1/README.md

@@ -0,0 +1,44 @@
+# task170_super_spark_reasoning_parser_revision_pin_s1
+
+<!-- METADATA:STATUS=InProgress,ASSIGNEE=intern_nem_dev_2 -->
+
+Status: In progress
+Owner: intern_nem_dev_2
+Branch: `intern_nem_dev_2/task170_super_spark_reasoning_parser_revision_pin_s1`
+Base: `6500fdaa27735197da87ca25d641a2883b00e8e6`
+PR: pending
+
+## Summary
+
+Pin the Spark deployment guide examples that download
+`super_v3_reasoning_parser.py` so users do not fetch a drifting Hugging Face
+`main` branch file.
+
+## Scope
+
+- `usage-cookbook/Nemotron-3-Super/SparkDeploymentGuide/README.md`
+- Focused static docs test under `tests/usage_cookbook/`
+- Task/status docs for `intern_nem_dev_2`
+
+## Pin
+
+- Repo: `nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4`
+- Revision: `4f0cf9daaeb7a4d5e23f80a00e7ed15f0e03caf6`
+- File: `super_v3_reasoning_parser.py`
+
+## Boundaries
+
+- Docs/static-test only.
+- No live wget/curl, HF/model download, vLLM/TRT-LLM launch, endpoint call,
+  W&B run, cluster job, deploy, artifact operation, direct `main`/`master`
+  push, or self-merge.
+
+## Acceptance Checks
+
+- PASS: `PYTHONPATH=src /work-agents/.venv/bin/python -m pytest -q tests/usage_cookbook/test_spark_reasoning_parser_revision.py` (3 passed)
+- PASS: `/work-agents/.venv/bin/python -m py_compile tests/usage_cookbook/test_spark_reasoning_parser_revision.py`
+- PASS: `/work-agents/.venv/bin/ruff check tests/usage_cookbook/test_spark_reasoning_parser_revision.py`
+- PASS: structured static probe for exact pinned reasoning-parser URLs and no `raw/main`
+- PASS: added-line live-surface scan showed only Spark guide static wget examples
+- PASS: `git diff --check`
+- PASS: `git diff --cached --check`

workspace/tasks/task170_super_spark_reasoning_parser_revision_pin_s1/history_log.md

@@ -0,0 +1,16 @@
+# task170_super_spark_reasoning_parser_revision_pin_s1 history
+
+<!-- METADATA:SESSION=1 -->
+
+## Session 1 - 2026-05-29
+
+- Accepted PM assignment and created branch
+  `intern_nem_dev_2/task170_super_spark_reasoning_parser_revision_pin_s1`
+  from `origin/main` at `6500fdaa27735197da87ca25d641a2883b00e8e6`.
+- Replaced two Spark guide `super_v3_reasoning_parser.py` Hugging Face
+  `raw/main` wget URLs with commit-pinned `resolve/4f0cf9...` URLs.
+- Preserved the guide's vLLM/TRT-LLM command semantics and parser filename.
+- Added focused static tests that inspect the markdown without downloading or
+  launching any serving stack.
+- Verified focused pytest, `py_compile`, Ruff, structured static probe,
+  added-line live-surface scan, and diff checks.

workspace/tasks/task170_super_spark_reasoning_parser_revision_pin_s1/task_knowledge.md

@@ -0,0 +1,12 @@
+# task170_super_spark_reasoning_parser_revision_pin_s1 knowledge
+
+<!-- METADATA:SESSION=1 -->
+
+## Working Notes
+
+- PM-provided metadata-only model revision:
+  `4f0cf9daaeb7a4d5e23f80a00e7ed15f0e03caf6`.
+- The pinned parser URL is:
+  `https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4/resolve/4f0cf9daaeb7a4d5e23f80a00e7ed15f0e03caf6/super_v3_reasoning_parser.py`
+- Tests must inspect the Spark guide statically and must not run `wget`, `curl`,
+  HF/model downloads, vLLM, or TRT-LLM.