specs/F007-DEMO.md

Feature Demo: F007 — HuggingFace Deployment & Submission

Generated: 2026-03-29T07:33:23Z Context source: spec + discovery only (implementation not read) Feature entry: FEATURES.json #F007

---

What This Feature Does

F007 packages SQLEnv so a judge can actually consume it end-to-end: discover the project from README, run or visit the deployed Hugging Face Space, and use the training notebook workflow.

From a user perspective, the core value is trust and usability: deployment assets validate/build/push cleanly, and the submission package is runnable by someone outside the team.

---

What Is Already Proven

Verified in This Demo Run

• Ran deployment validation locally with uv run openenv validate --verbose.
• Built deployment image locally with uv run openenv build -t openenv-sql-env-f007-hf-submission.
• Ran authenticated deployment push with uv run openenv push to https://huggingface.co/spaces/hjerpe/sql_env.
• Ran notebook/training E2E checks (tests/e2e/test_training_e2e.py): 5 passed.
• Ran full regression suite: 250 passed, 1 skipped.

Previously Verified Evidence

• specs/FEATURES.json → verification_evidence for F007: 250/250 tests passed, verifier approved.
• specs/F007-IMPLEMENTATION_SPEC.md (Section 1a) records authenticated build + push completion evidence.

---

What Still Needs User Verification

• Open the live Space in a browser and manually run a reset/step/answer episode flow.
• Open notebooks/train_grpo.ipynb in Colab and execute cells in order on a clean runtime.

---

Quickstart / Verification Steps

Run these commands to see the feature in action:

uv run openenv validate --verbose
uv run openenv build -t openenv-sql-env-f007-hf-submission
uv run openenv push

Prereq: authenticated Hugging Face CLI/account with write access to target Space.

---

Live Local Proof

Validate Deployment Configuration

This confirms deployment mode support and flags non-Docker modes clearly.

uv run openenv validate --verbose

[OK] sql-env-F007-huggingface-deployment-submission: Ready for multi-mode deployment

Supported deployment modes:
  [YES] docker
  [YES] openenv_serve
  [YES] uv_run
  [YES] python_module

What to notice: All four deployment modes are supported.

Build the Hugging Face Deployment Image

uv run openenv build -t openenv-sql-env-f007-hf-submission

Building Docker image for: sql-env-F007-huggingface-deployment-submission
...
#18 naming to docker.io/library/openenv-sql-env-f007-hf-submission done
✓ Docker build successful

Done!

What to notice: image build completed successfully with the expected tag.

Push to Hugging Face Space

uv run openenv push

✓ Authenticated as: hjerpe
Creating/verifying space: hjerpe/sql_env
✓ Space hjerpe/sql_env is ready
Uploading files to hjerpe/sql_env...
✓ Upload completed successfully
Space URL: https://huggingface.co/spaces/hjerpe/sql_env

✓ Deployment complete!
Visit your space at: https://huggingface.co/spaces/hjerpe/sql_env

What to notice: authenticated push succeeded and produced a live Space URL.

---

Existing Evidence

• Verification spec target command (uv run --with pytest pytest tests/ -v) was re-run in this demo and passed.
• F007 entry in specs/FEATURES.json already recorded verifier approval before this refresh.

---

Manual Verification Checklist

1. Open https://huggingface.co/spaces/hjerpe/sql_env.
2. Confirm the app loads without startup errors.
3. Start an episode (reset), then run at least one exploration step.
4. Submit an answer action and confirm terminal response/reward appears.
5. Open notebooks/train_grpo.ipynb in Colab and run setup + connect + one training/eval pass.

---

Edge Cases Exercised

All deployment modes pass validation

uv run openenv validate --verbose

Supported deployment modes:
  [YES] docker
  [YES] openenv_serve
  [YES] uv_run
  [YES] python_module

This matters because all four modes pass cleanly — no warnings or caveats for the submission reviewer.

Verification-spec command drift (error case)

uv run --with pytest pytest tests/e2e/test_readme_completeness.py -v

ERROR: file or directory not found: tests/e2e/test_readme_completeness.py
collected 0 items
============================ no tests ran in 0.00s ============================

This matters because it reveals a spec-to-repo mismatch that should be corrected in verification artifacts.

Notebook pipeline smoke validation still passes

uv run --with pytest pytest tests/e2e/test_training_e2e.py -v

collected 5 items
...
============================== 5 passed in 11.33s ==============================

This confirms the training notebook path still has executable smoke coverage.

---

Test Evidence (Optional)

Supplementary proof that the feature works correctly across all scenarios.

Test Suite	Tests	Status
Full regression (uv run --with pytest pytest tests/ -v)	251 collected	250 passed, 1 skipped
Training E2E (tests/e2e/test_training_e2e.py)	5	All passed

---

Feature Links

• Implementation spec: specs/F007-IMPLEMENTATION_SPEC.md
• Verification spec: specs/F007-VERIFICATION_SPEC.md

---

Demo generated by feature-demo agent. Re-run with /feature-demo F007 to refresh.