Official Codebase of 3DCodeVerse
Python
dataset: the user-prompt id is quality.prompt_id — it shared the name prompt_hash with prompts.prompt_hash (a 12-hex template hash) in the same package
gaoypeng415ec15f5c64
diff --git a/harness/codeverse3d/addons/dataset/__init__.py b/harness/codeverse3d/addons/dataset/__init__.pyindex 74a98be7..bb43b772 100644--- a/harness/codeverse3d/addons/dataset/__init__.py+++ b/harness/codeverse3d/addons/dataset/__init__.py@@ -7,5 +7,5 @@* ``captions`` ``caption_sample(ws, record, model_id)`` → ``Captions``* ``index`` ``build_index(runs_dir, out_sqlite)`` + query helpers* ``sample`` one run → one dataset sample (the code files of its picked round)-* ``quality`` ``quality_tier`` (A/B/C/D) · ``prompt_hash`` · code fingerprints, duplicate groups+* ``quality`` ``quality_tier`` (A/B/C/D) · ``prompt_id`` · code fingerprints, duplicate groups"""diff --git a/harness/codeverse3d/addons/dataset/index.py b/harness/codeverse3d/addons/dataset/index.pyindex 5d828cfa..3d543cdb 100644--- a/harness/codeverse3d/addons/dataset/index.py+++ b/harness/codeverse3d/addons/dataset/index.py@@ -13,7 +13,7 @@ from pathlib import Pathfrom typing import Anyfrom codeverse3d.addons import select-from codeverse3d.addons.dataset.quality import prompt_hash, quality_tier+from codeverse3d.addons.dataset.quality import prompt_id, quality_tierfrom codeverse3d.addons.dataset.sample import gate_error_summaryfrom codeverse3d.contracts.run import RoundRecord, RunId, RunRecordfrom codeverse3d.record.record import effective_judgment, iter_runs, skip_unreadable@@ -54,7 +54,7 @@ def _run_row(ws: Workspace, rec: RunRecord, rid: RunId) -> tuple:tier = quality_tier(passed=j.passed if j else None, gate_errors=n_err, score=j.overall if j else None)return (rid.slug, str(ws.root), rec.spec.track.value, rec.spec.language.value, rec.spec.prompt,- prompt_hash(rec.spec.prompt), rec.spec.backends.generator, rec.spec.backends.planner,+ prompt_id(rec.spec.prompt), rec.spec.backends.generator, rec.spec.backends.planner,rec.spec.backends.judge, s.stop_reason, s.baseline_score, s.picked_score, s.picked_round,len(rec.rounds), rec.total_usage.cost_usd, rec.total_usage.input_tokens,rec.total_usage.output_tokens, rec.started_at.isoformat(),diff --git a/harness/codeverse3d/addons/dataset/pairs.py b/harness/codeverse3d/addons/dataset/pairs.pyindex 528b4a3c..3844d469 100644--- a/harness/codeverse3d/addons/dataset/pairs.py+++ b/harness/codeverse3d/addons/dataset/pairs.py@@ -31,7 +31,7 @@ from pathlib import Pathfrom typing import Anyfrom codeverse3d.addons import select-from codeverse3d.addons.dataset.quality import prompt_hash+from codeverse3d.addons.dataset.quality import prompt_idfrom codeverse3d.contracts.run import RoundRecord, RunRecordfrom codeverse3d.proc import read_jsonl_lenientfrom codeverse3d.record import _git@@ -39,7 +39,7 @@ from codeverse3d.record.record import effective_judgment, iter_runs, skip_unreadfrom codeverse3d.workspace import Workspace__all__ = ["build_pairs", "preference_pairs", "repair_pairs", "in_round_repair_pairs",- "cross_backend_pairs", "prompt_hash"]+ "cross_backend_pairs"]MAX_INLINE_CODE = 200_000@@ -74,7 +74,7 @@ def _base(ws: Workspace, rec: RunRecord, kind: str, *, slug: str | None = None)"run": slug or ws.root.name,"workspace": str(ws.root),"prompt": rec.spec.prompt,- "prompt_hash": prompt_hash(rec.spec.prompt),+ "prompt_hash": prompt_id(rec.spec.prompt),"track": rec.spec.track.value,"language": rec.spec.language.value,"generator": rec.spec.backends.generator,@@ -255,7 +255,7 @@ def build_pairs(runs_dir: Path | str, out_jsonl: Path | str, *, min_delta: floatwith tmp.open("w") as fh:for ws, rec, rid in iter_runs(runs_dir, on_error=skip_unreadable):slug = rid.slug- groups[(prompt_hash(rec.spec.prompt), rec.spec.track.value, rec.spec.language.value)].append((ws, rec, slug))+ groups[(prompt_id(rec.spec.prompt), rec.spec.track.value, rec.spec.language.value)].append((ws, rec, slug))pairs = (preference_pairs(ws, rec, min_delta=min_delta, slug=slug)+ repair_pairs(ws, rec, slug=slug) + in_round_repair_pairs(ws, rec, slug=slug))for pair in pairs:diff --git a/harness/codeverse3d/addons/dataset/quality.py b/harness/codeverse3d/addons/dataset/quality.pyindex 4d3d66b9..b45ee81a 100644--- a/harness/codeverse3d/addons/dataset/quality.py+++ b/harness/codeverse3d/addons/dataset/quality.py@@ -60,7 +60,7 @@ TIER_ORDER: dict[str, int] = {"A": 0, "B": 1, "C": 2, "D": 3}TIER_C_MIN_SCORE = 0.6-def prompt_hash(prompt: str) -> str:+def prompt_id(prompt: str) -> str:"""Stable 16-hex id of a (stripped) user prompt — shared by samples, pairs and the index."""return hashlib.sha256(prompt.strip().encode()).hexdigest()[:16]diff --git a/harness/codeverse3d/addons/dataset/sample.py b/harness/codeverse3d/addons/dataset/sample.pyindex 686e9ac5..5f1c3cca 100644--- a/harness/codeverse3d/addons/dataset/sample.py+++ b/harness/codeverse3d/addons/dataset/sample.py@@ -26,7 +26,7 @@ from codeverse3d.addons import selectfrom codeverse3d.addons.dataset.quality import (QualityTier,code_fingerprint,- prompt_hash,+ prompt_id,quality_tier,)from codeverse3d.contracts.common import (@@ -345,7 +345,7 @@ def build_meta(code_commit=rnd.commit if rnd is not None else "",code_source=code_source,code_fingerprint=code_fingerprint(code) if code else "",- prompt_hash=prompt_hash(spec.prompt),+ prompt_hash=prompt_id(spec.prompt),status=record.status.value,cost_usd=round(record.total_usage.cost_usd, 6),usage=record.total_usage.model_dump(mode="json"),diff --git a/harness/docs/INTERFACES.md b/harness/docs/INTERFACES.mdindex a3dbb955..aafe5148 100644--- a/harness/docs/INTERFACES.md+++ b/harness/docs/INTERFACES.md@@ -640,7 +640,7 @@ from codeverse3d.record.record import complexity_block, round_complexity # obj# round_complexity (a dataset sample's rounds_summary rows carry it)from codeverse3d.addons.dataset.export import export_samples # (runs_dir, out_dir, *, min_score=None, only_passed=False,# include_unbuilt=False, captions_dir=None, drop_duplicates=False) -> ExportReport{…, n_duplicates, duplicates, tiers}-from codeverse3d.addons.dataset.quality import quality_tier, prompt_hash, find_duplicates # tiers (of the exported round's verdict):+from codeverse3d.addons.dataset.quality import quality_tier, prompt_id, find_duplicates # tiers (of the exported round's verdict):# A passed & 0 gate errors, B passed, C score ≥ 0.6, D else; dedupe = (code fingerprint, prompt)from codeverse3d.addons.dataset.pairs import build_pairs # (runs_dir, out_jsonl, *, min_delta=0.05) -> nfrom codeverse3d.addons.dataset.refine import build_refine, transitions, RefineTransition, REFINE_KINDS, outcome_ofdiff --git a/harness/tests/flywheel_cli/test_quality_gallery.py b/harness/tests/flywheel_cli/test_quality_gallery.pyindex f6bb74b2..cbf85b29 100644--- a/harness/tests/flywheel_cli/test_quality_gallery.py+++ b/harness/tests/flywheel_cli/test_quality_gallery.py@@ -10,7 +10,7 @@ from codeverse3d.addons.dataset.export import export_samples, load_captionsfrom codeverse3d.addons.dataset.quality import (find_duplicates,mark_duplicates,- prompt_hash,+ prompt_id,quality_tier,)from codeverse3d.record.record import load_record@@ -46,7 +46,7 @@ def test_only_raw_hash_duplicates_are_dropped_normalised_ones_are_only_marked():mark_duplicates(rows)assert {r["id"]: r["duplicate_of"] for r in rows} == {"a": "e", "b": "e", "c": "", "d": "", "e": "", "f": ""}assert {r["id"]: r["near_duplicate_of"] for r in rows} == {"a": "e", "b": "e", "c": "e", "d": "", "e": "", "f": ""}- assert prompt_hash(" a chair ") == prompt_hash("a chair") and len(prompt_hash("x")) == 16+ assert prompt_id(" a chair ") == prompt_id("a chair") and len(prompt_id("x")) == 16def test_whitespace_inside_a_string_is_a_near_duplicate_not_a_duplicate(runs_dir: Path, tmp_path: Path):