cjm-transcript-graph-schema 0.0.6__tar.gz → 0.0.8__tar.gz

{cjm_transcript_graph_schema-0.0.6 → cjm_transcript_graph_schema-0.0.8}/MANIFEST.in

@@ -1,5 +1,4 @@
 include pyproject.toml
 include LICENSE
-include CONTRIBUTING.md
 include README.md
 recursive-exclude * __pycache__

cjm_transcript_graph_schema-0.0.8/PKG-INFO

@@ -0,0 +1,52 @@
+Metadata-Version: 2.4
+Name: cjm-transcript-graph-schema
+Version: 0.0.8
+Summary: Audio-transcript layer schema for context graphs: Source, AudioSegment, Transcript, and Segment nodes with deterministic identity tuples and graph-node mapping, shared by the transcription, decomposition, and correction workflow cores.
+Author-email: "Christian J. Mills" <9126128+cj-mills@users.noreply.github.com>
+License: Apache-2.0
+Project-URL: Repository, https://github.com/cj-mills/cjm-transcript-graph-schema
+Keywords: python
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3 :: Only
+Requires-Python: >=3.12
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: cjm-context-graph-layer>=0.0.7
+Requires-Dist: cjm-context-graph-primitives>=0.0.12
+Dynamic: license-file
+
+# cjm-transcript-graph-schema
+
+<!-- generated from the context graph by `cjm-context-graph readme` — do not edit by hand; edit the graph (the urge to hand-edit = move it on-graph) -->
+
+Audio-transcript layer schema for context graphs: Source, AudioSegment, Transcript, and Segment nodes with deterministic identity tuples and graph-node mapping, shared by the transcription, decomposition, and correction workflow cores.
+
+## Modules
+
+- **`cjm_transcript_graph_schema.schema`** — The audio-transcript layer schema (where-graph-begins locked layer schema): Source -> AudioSegment(coarse boundary) -> AudioRendition(model-input: raw | vocals | ...) -> Transcript(per-transcriber variants) emitted by transcription, extended with the fine Segment spine (per-rendition) by decomposition. Deterministic identity tuples per the stage-5 ratified rule; the AudioRendition node lets raw + preprocessed model-inputs of one boundary coexist in one graph. Document from the pre-CR-18 era dissolves into Source.
+
+## API
+
+### `cjm_transcript_graph_schema.schema`
+
+- `AudioRenditionNode` _class_ — A model-input rendition OF an AudioSegment — the materialized 16k-mono WAV
+- `AudioSegmentNode` _class_ — Coarse ~5-min spine member: a BOUNDARY range of the Source (an audio fact),
+- `CollectionNode` _class_ — An intrinsic collection (book, podcast series, lecture course) — the
+- `SegmentNode` _class_ — Fine spine member: one VAD chunk — IMMUTABLE audio range + CORRECTABLE
+- `SourceNode` _class_ — The provenance root: one ingested media file.
+- `TranscriptGraphLabels` _class_ — Node labels of the audio-transcript layer schema.
+- `TranscriptNode` _class_ — One transcriber's text for one AudioRendition (per-transcriber variants at
+- `TranscriptSliceRef` _class_ — One per-transcriber char-range reference for a fine Segment: where this
+- `audio_rendition_node_id` _function_ — AudioRendition identity = (audio segment, preprocessing chain).
+- `audio_segment_node_id` _function_ — AudioSegment identity = (source, boundary range).
+- `collection_edges` _function_ — Membership edges for a collection (ae3464fc: ORDER IS OPTIONAL).
+- `collection_node_id` _function_ — Collection identity = the NORMALIZED TITLE (ratified fork, hub-session
+- `normalize_collection_title` _function_ — Normalize a collection title for identity derivation.
+- `segment_node_id` _function_ — Fine Segment identity = audio-side only (audio rendition, VAD config, chunk
+- `source_node_id` _function_ — Source identity = the ingested file's content hash (Thread-1 ingested-root identity).
+- `transcript_node_id` _function_ — Transcript identity = (audio rendition, transcriber, config) — MIRRORS the
+
+## Dependencies
+
+**Depends on:** `cjm-context-graph-layer`, `cjm-context-graph-primitives`
+**Used by:** `cjm-context-graph-layer`, `cjm-transcript-correction-core`, `cjm-transcript-correction-tui`, `cjm-transcript-decomp-core`, `cjm-transcription-core`

cjm_transcript_graph_schema-0.0.8/README.md

@@ -0,0 +1,35 @@
+# cjm-transcript-graph-schema
+
+<!-- generated from the context graph by `cjm-context-graph readme` — do not edit by hand; edit the graph (the urge to hand-edit = move it on-graph) -->
+
+Audio-transcript layer schema for context graphs: Source, AudioSegment, Transcript, and Segment nodes with deterministic identity tuples and graph-node mapping, shared by the transcription, decomposition, and correction workflow cores.
+
+## Modules
+
+- **`cjm_transcript_graph_schema.schema`** — The audio-transcript layer schema (where-graph-begins locked layer schema): Source -> AudioSegment(coarse boundary) -> AudioRendition(model-input: raw | vocals | ...) -> Transcript(per-transcriber variants) emitted by transcription, extended with the fine Segment spine (per-rendition) by decomposition. Deterministic identity tuples per the stage-5 ratified rule; the AudioRendition node lets raw + preprocessed model-inputs of one boundary coexist in one graph. Document from the pre-CR-18 era dissolves into Source.
+
+## API
+
+### `cjm_transcript_graph_schema.schema`
+
+- `AudioRenditionNode` _class_ — A model-input rendition OF an AudioSegment — the materialized 16k-mono WAV
+- `AudioSegmentNode` _class_ — Coarse ~5-min spine member: a BOUNDARY range of the Source (an audio fact),
+- `CollectionNode` _class_ — An intrinsic collection (book, podcast series, lecture course) — the
+- `SegmentNode` _class_ — Fine spine member: one VAD chunk — IMMUTABLE audio range + CORRECTABLE
+- `SourceNode` _class_ — The provenance root: one ingested media file.
+- `TranscriptGraphLabels` _class_ — Node labels of the audio-transcript layer schema.
+- `TranscriptNode` _class_ — One transcriber's text for one AudioRendition (per-transcriber variants at
+- `TranscriptSliceRef` _class_ — One per-transcriber char-range reference for a fine Segment: where this
+- `audio_rendition_node_id` _function_ — AudioRendition identity = (audio segment, preprocessing chain).
+- `audio_segment_node_id` _function_ — AudioSegment identity = (source, boundary range).
+- `collection_edges` _function_ — Membership edges for a collection (ae3464fc: ORDER IS OPTIONAL).
+- `collection_node_id` _function_ — Collection identity = the NORMALIZED TITLE (ratified fork, hub-session
+- `normalize_collection_title` _function_ — Normalize a collection title for identity derivation.
+- `segment_node_id` _function_ — Fine Segment identity = audio-side only (audio rendition, VAD config, chunk
+- `source_node_id` _function_ — Source identity = the ingested file's content hash (Thread-1 ingested-root identity).
+- `transcript_node_id` _function_ — Transcript identity = (audio rendition, transcriber, config) — MIRRORS the
+
+## Dependencies
+
+**Depends on:** `cjm-context-graph-layer`, `cjm-context-graph-primitives`
+**Used by:** `cjm-context-graph-layer`, `cjm-transcript-correction-core`, `cjm-transcript-correction-tui`, `cjm-transcript-decomp-core`, `cjm-transcription-core`

cjm_transcript_graph_schema-0.0.8/cjm_transcript_graph_schema/__init__.py

@@ -0,0 +1 @@
+__version__ = "0.0.8"

{cjm_transcript_graph_schema-0.0.6 → cjm_transcript_graph_schema-0.0.8}/cjm_transcript_graph_schema/schema.py

@@ -1,26 +1,17 @@
-"""The audio-transcript layer schema (where-graph-begins locked layer schema): `Source -> AudioSegment(coarse boundary) -> AudioRendition(model-input: raw | vocals | …) -> Transcript(per-transcriber variants)` emitted by transcription, extended with the fine `Segment` spine (per-rendition) by decomposition. Deterministic identity tuples per the stage-5 ratified rule; the `AudioRendition` node lets raw + preprocessed model-inputs of one boundary coexist in one graph. `Document` from the pre-CR-18 era dissolves into `Source`.
+"""The audio-transcript layer schema (where-graph-begins locked layer schema): Source -> AudioSegment(coarse boundary) -> AudioRendition(model-input: raw | vocals | ...) -> Transcript(per-transcriber variants) emitted by transcription, extended with the fine Segment spine (per-rendition) by decomposition, and with the intrinsic Collection layer ABOVE Source (book/series/course; proposed-by-tool / confirmed-by-actor, order optional) captured at transcription hand-off. Deterministic identity tuples per the stage-5 ratified rule; the AudioRendition node lets raw + preprocessed model-inputs of one boundary coexist in one graph. Document from the pre-CR-18 era dissolves into Source."""
 
-Docs: https://cj-mills.github.io/cjm-transcript-graph-schemaschema.html.md"""
-
-# AUTOGENERATED! DO NOT EDIT! File to edit: ../nbs/schema.ipynb.
-
-# %% auto #0
-__all__ = ['TranscriptGraphLabels', 'source_node_id', 'audio_segment_node_id', 'audio_rendition_node_id', 'transcript_node_id',
-           'segment_node_id', 'SourceNode', 'AudioSegmentNode', 'AudioRenditionNode', 'TranscriptNode',
-           'TranscriptSliceRef', 'SegmentNode']
-
-# %% ../nbs/schema.ipynb #76d0dd1c
-from pathlib import Path
 from dataclasses import dataclass, field
+from pathlib import Path
 from typing import Any, Dict, List, Optional
 
+from cjm_context_graph_layer.grammar import (attribution, make_edge, OverlayRelations, spine_edges,
+                                             SpineRelations)
 from cjm_context_graph_layer.identity import derive_node_id
-from cjm_context_graph_layer.grammar import OverlayRelations, attribution, make_edge
 from cjm_context_graph_primitives.locators import FileRef, GraphNodeRef
-from cjm_context_graph_primitives.slices import TimeSlice, CharSlice, FullContent
 from cjm_context_graph_primitives.provenance import SourceRef
+from cjm_context_graph_primitives.slices import CharSlice, FullContent, TimeSlice
+
 
-# %% ../nbs/schema.ipynb #ff94773a
 class TranscriptGraphLabels:
     """Node labels of the audio-transcript layer schema."""
     SOURCE = "Source"                  # Provenance root: one ingested media file
@@ -28,20 +19,21 @@ class TranscriptGraphLabels:
     AUDIO_RENDITION = "AudioRendition" # A model-input rendition OF an AudioSegment (raw convert | vocals | future enhancement); OWNS the model-input WAV
     TRANSCRIPT = "Transcript"          # Per-transcriber text variant of one AudioRendition
     SEGMENT = "Segment"                # Fine spine member: one VAD chunk (immutable audio + correctable text), per-rendition
+    COLLECTION = "Collection"          # Intrinsic collection (book / series / course): the fractal spine layer ABOVE Source
 
     @classmethod
     def all(cls) -> list:  # All schema labels
         """All schema labels."""
-        return [cls.SOURCE, cls.AUDIO_SEGMENT, cls.AUDIO_RENDITION, cls.TRANSCRIPT, cls.SEGMENT]
+        return [cls.SOURCE, cls.AUDIO_SEGMENT, cls.AUDIO_RENDITION, cls.TRANSCRIPT, cls.SEGMENT, cls.COLLECTION]
+
 
-# %% ../nbs/schema.ipynb #a5ea12db
 def source_node_id(
     content_hash: str,  # Content hash of the source media file ("algo:hexdigest")
 ) -> str:  # Deterministic Source node id
     """Source identity = the ingested file's content hash (Thread-1 ingested-root identity)."""
     return derive_node_id("source", content_hash)
 
-# %% ../nbs/schema.ipynb #794a019f
+
 def audio_segment_node_id(
     source_id: str,  # Owning Source node id
     start: float,    # Boundary start (source-coordinate seconds)
@@ -55,7 +47,7 @@ def audio_segment_node_id(
     the id; the model-input WAV lives on the `AudioRendition` child, not here."""
     return derive_node_id("audio-segment", source_id, start, end)
 
-# %% ../nbs/schema.ipynb #8e147ac9
+
 def audio_rendition_node_id(
     audio_segment_id: str,  # Owning AudioSegment node id
     chain: List[str],       # Ordered preprocessing-chain descriptors; [] = the raw convert-only rendition
@@ -73,7 +65,7 @@ def audio_rendition_node_id(
     search)."""
     return derive_node_id("audio-rendition", audio_segment_id, *chain)
 
-# %% ../nbs/schema.ipynb #6fe30fc8
+
 def transcript_node_id(
     rendition_id: str,  # Owning AudioRendition node id
     transcriber: str,   # Transcriber capability name (e.g. "cjm-transcription-plugin-whisper")
@@ -85,7 +77,7 @@ def transcript_node_id(
     RENDITION, so a raw vs vocals transcript of the same boundary are distinct."""
     return derive_node_id("transcript", rendition_id, transcriber, config_hash)
 
-# %% ../nbs/schema.ipynb #f0e4f510
+
 def segment_node_id(
     rendition_id: str,     # Owning AudioRendition node id
     vad_config_hash: str,  # VAD capability config hash (skeleton identity input)
@@ -99,7 +91,7 @@ def segment_node_id(
     yield different VAD chunk boundaries than raw)."""
     return derive_node_id("segment", rendition_id, vad_config_hash, chunk_start, chunk_end)
 
-# %% ../nbs/schema.ipynb #d768584a
+
 @dataclass
 class SourceNode:
     """The provenance root: one ingested media file."""
@@ -128,7 +120,7 @@ class SourceNode:
                                   content_hash=self.content_hash).to_dict()],
         }
 
-# %% ../nbs/schema.ipynb #3d76c961
+
 @dataclass
 class AudioSegmentNode:
     """Coarse ~5-min spine member: a BOUNDARY range of the Source (an audio fact),
@@ -169,7 +161,7 @@ class AudioSegmentNode:
             "sources": [],
         }
 
-# %% ../nbs/schema.ipynb #c97ae502
+
 @dataclass
 class AudioRenditionNode:
     """A model-input rendition OF an AudioSegment — the materialized 16k-mono WAV
@@ -216,7 +208,7 @@ class AudioRenditionNode:
         """DERIVED_FROM edge: this rendition derives from its AudioSegment's audio."""
         return make_edge(self.id, self.audio_segment, OverlayRelations.DERIVED_FROM)
 
-# %% ../nbs/schema.ipynb #9948f2c5
+
 @dataclass
 class TranscriptNode:
     """One transcriber's text for one AudioRendition (per-transcriber variants at
@@ -259,7 +251,7 @@ class TranscriptNode:
         """DERIVED_FROM edge: this Transcript derives from its AudioRendition."""
         return make_edge(self.id, self.rendition, OverlayRelations.DERIVED_FROM)
 
-# %% ../nbs/schema.ipynb #0c830160
+
 @dataclass
 class TranscriptSliceRef:
     """One per-transcriber char-range reference for a fine Segment: where this
@@ -276,7 +268,7 @@ class TranscriptSliceRef:
                          content_hash=SourceRef.compute_hash(self.text.encode()),
                          slice=CharSlice(start=self.start_char, end=self.end_char)).to_dict()
 
-# %% ../nbs/schema.ipynb #fecb183e
+
 @dataclass
 class SegmentNode:
     """Fine spine member: one VAD chunk — IMMUTABLE audio range + CORRECTABLE
@@ -331,3 +323,88 @@ class SegmentNode:
             "properties": props,
             "sources": sources,
         }
+
+
+def normalize_collection_title(
+    title: str,  # Raw title (folder name, user input, ...)
+) -> str:  # Normalized identity form
+    """Normalize a collection title for identity derivation.
+
+    Casefold + underscores-to-spaces + whitespace collapse, so a folder named
+    ``Hardcore_History`` and a hand-typed ``Hardcore History`` converge on the
+    SAME collection node (ae3464fc: select-existing and retype-the-title are
+    the same act by construction). Display keeps the raw title; only identity
+    normalizes."""
+    return " ".join(title.replace("_", " ").split()).casefold()
+
+
+def collection_node_id(
+    title: str,  # Collection title (normalized internally)
+) -> str:  # Deterministic Collection node id
+    """Collection identity = the NORMALIZED TITLE (ratified fork, hub-session
+    2026-07-21): path-based identity breaks workspace relocatability, and
+    member-hash identity breaks late-added members — late binding is a feature.
+    The same book ingested twice (different paths, different machines, members
+    accreted over weeks) is the same collection; rare title collisions are
+    propose/confirm reconciler territory."""
+    return derive_node_id("collection", normalize_collection_title(title))
+
+
+@dataclass
+class CollectionNode:
+    """An intrinsic collection (book, podcast series, lecture course) — the
+    fractal spine layer ABOVE Source (the grammar's "series -> episode
+    tomorrow", arrived).
+
+    Status rides the node (ae3464fc actor criterion): a human who accepted,
+    renamed, or selected the collection at run launch has CONFIRMED it;
+    untouched tool defaults (headless folder run) land PROPOSED — visible and
+    grouping-usable everywhere, flagged as awaiting a human. A folder is
+    EVIDENCE of a grouping, not proof: proposed collections are cheap to
+    rename, refile out of, or dissolve. root_kind=asserted — a collection is
+    a grouping ASSERTION about ingested content, not ingested content itself.
+    Collection PART_OF Collection nesting is grammar-free (no v0 tooling)."""
+    title: str                     # Display title (identity input via normalize_collection_title)
+    status: str = "proposed"       # "proposed" (tool inference) | "confirmed" (human act)
+    actor: str = "human"           # Who declared it (attribution; e.g. "human", "tui:transcription", "cli:transcribe")
+    asserted_at: Optional[float] = None  # Declaration timestamp; None = now
+
+    @property
+    def id(self) -> str:  # Deterministic node id
+        """Deterministic node id."""
+        return collection_node_id(self.title)
+
+    def to_graph_node(self) -> Dict[str, Any]:  # Node wire dict
+        """Build the Collection node wire dict (asserted root; attribution included)."""
+        props: Dict[str, Any] = {
+            "title": self.title,
+            "status": self.status,
+            "root_kind": "asserted",
+        }
+        props.update(attribution(self.actor, method="collection-declaration",
+                                 asserted_at=self.asserted_at))
+        return {
+            "id": self.id,
+            "label": TranscriptGraphLabels.COLLECTION,
+            "properties": props,
+            "sources": [],
+        }
+
+
+def collection_edges(
+    collection_id: str,     # Collection node id
+    member_ids: List[str],  # Member Source node ids (declaration order when ordered=True)
+    ordered: bool = False,  # True = materialize the order (NEXT chain + STARTS_WITH)
+) -> List[Dict[str, Any]]:  # Edge wire dicts
+    """Membership edges for a collection (ae3464fc: ORDER IS OPTIONAL).
+
+    Unordered membership is just PART_OF member->collection — individually
+    accreted members (chapters run one at a time) attach without fabricating a
+    sequence. When an order IS known (a folder run's sorted expansion, or a
+    curation act), the full fractal spine pattern materializes via
+    spine_edges: PART_OF + NEXT chain + STARTS_WITH — identical grammar to
+    every layer below. Reordering and late insertion are curation ops, never
+    capture-time guesses."""
+    if ordered:
+        return spine_edges(collection_id, member_ids)
+    return [make_edge(m, collection_id, SpineRelations.PART_OF) for m in member_ids]

cjm_transcript_graph_schema-0.0.8/cjm_transcript_graph_schema.egg-info/PKG-INFO

@@ -0,0 +1,52 @@
+Metadata-Version: 2.4
+Name: cjm-transcript-graph-schema
+Version: 0.0.8
+Summary: Audio-transcript layer schema for context graphs: Source, AudioSegment, Transcript, and Segment nodes with deterministic identity tuples and graph-node mapping, shared by the transcription, decomposition, and correction workflow cores.
+Author-email: "Christian J. Mills" <9126128+cj-mills@users.noreply.github.com>
+License: Apache-2.0
+Project-URL: Repository, https://github.com/cj-mills/cjm-transcript-graph-schema
+Keywords: python
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3 :: Only
+Requires-Python: >=3.12
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: cjm-context-graph-layer>=0.0.7
+Requires-Dist: cjm-context-graph-primitives>=0.0.12
+Dynamic: license-file
+
+# cjm-transcript-graph-schema
+
+<!-- generated from the context graph by `cjm-context-graph readme` — do not edit by hand; edit the graph (the urge to hand-edit = move it on-graph) -->
+
+Audio-transcript layer schema for context graphs: Source, AudioSegment, Transcript, and Segment nodes with deterministic identity tuples and graph-node mapping, shared by the transcription, decomposition, and correction workflow cores.
+
+## Modules
+
+- **`cjm_transcript_graph_schema.schema`** — The audio-transcript layer schema (where-graph-begins locked layer schema): Source -> AudioSegment(coarse boundary) -> AudioRendition(model-input: raw | vocals | ...) -> Transcript(per-transcriber variants) emitted by transcription, extended with the fine Segment spine (per-rendition) by decomposition. Deterministic identity tuples per the stage-5 ratified rule; the AudioRendition node lets raw + preprocessed model-inputs of one boundary coexist in one graph. Document from the pre-CR-18 era dissolves into Source.
+
+## API
+
+### `cjm_transcript_graph_schema.schema`
+
+- `AudioRenditionNode` _class_ — A model-input rendition OF an AudioSegment — the materialized 16k-mono WAV
+- `AudioSegmentNode` _class_ — Coarse ~5-min spine member: a BOUNDARY range of the Source (an audio fact),
+- `CollectionNode` _class_ — An intrinsic collection (book, podcast series, lecture course) — the
+- `SegmentNode` _class_ — Fine spine member: one VAD chunk — IMMUTABLE audio range + CORRECTABLE
+- `SourceNode` _class_ — The provenance root: one ingested media file.
+- `TranscriptGraphLabels` _class_ — Node labels of the audio-transcript layer schema.
+- `TranscriptNode` _class_ — One transcriber's text for one AudioRendition (per-transcriber variants at
+- `TranscriptSliceRef` _class_ — One per-transcriber char-range reference for a fine Segment: where this
+- `audio_rendition_node_id` _function_ — AudioRendition identity = (audio segment, preprocessing chain).
+- `audio_segment_node_id` _function_ — AudioSegment identity = (source, boundary range).
+- `collection_edges` _function_ — Membership edges for a collection (ae3464fc: ORDER IS OPTIONAL).
+- `collection_node_id` _function_ — Collection identity = the NORMALIZED TITLE (ratified fork, hub-session
+- `normalize_collection_title` _function_ — Normalize a collection title for identity derivation.
+- `segment_node_id` _function_ — Fine Segment identity = audio-side only (audio rendition, VAD config, chunk
+- `source_node_id` _function_ — Source identity = the ingested file's content hash (Thread-1 ingested-root identity).
+- `transcript_node_id` _function_ — Transcript identity = (audio rendition, transcriber, config) — MIRRORS the
+
+## Dependencies
+
+**Depends on:** `cjm-context-graph-layer`, `cjm-context-graph-primitives`
+**Used by:** `cjm-context-graph-layer`, `cjm-transcript-correction-core`, `cjm-transcript-correction-tui`, `cjm-transcript-decomp-core`, `cjm-transcription-core`

{cjm_transcript_graph_schema-0.0.6 → cjm_transcript_graph_schema-0.0.8}/cjm_transcript_graph_schema.egg-info/SOURCES.txt

@@ -3,11 +3,10 @@ MANIFEST.in
 README.md
 pyproject.toml
 cjm_transcript_graph_schema/__init__.py
-cjm_transcript_graph_schema/_modidx.py
 cjm_transcript_graph_schema/schema.py
 cjm_transcript_graph_schema.egg-info/PKG-INFO
 cjm_transcript_graph_schema.egg-info/SOURCES.txt
 cjm_transcript_graph_schema.egg-info/dependency_links.txt
-cjm_transcript_graph_schema.egg-info/entry_points.txt
 cjm_transcript_graph_schema.egg-info/requires.txt
-cjm_transcript_graph_schema.egg-info/top_level.txt
+cjm_transcript_graph_schema.egg-info/top_level.txt
+tests/test_schema.py

cjm_transcript_graph_schema-0.0.8/cjm_transcript_graph_schema.egg-info/requires.txt

@@ -0,0 +1,2 @@
+cjm-context-graph-layer>=0.0.7
+cjm-context-graph-primitives>=0.0.12

{cjm_transcript_graph_schema-0.0.6 → cjm_transcript_graph_schema-0.0.8}/pyproject.toml

@@ -10,19 +10,15 @@ readme = "README.md"
 requires-python = ">=3.12"
 license = {text = "Apache-2.0"}
 authors = [{name = "Christian J. Mills", email = "9126128+cj-mills@users.noreply.github.com"}]
-keywords = ['nbdev']
+keywords = ['python']
 classifiers = [
     "Programming Language :: Python :: 3",
     "Programming Language :: Python :: 3 :: Only",
 ]
-dependencies = ['cjm-context-graph-layer>=0.0.5', 'cjm-context-graph-primitives>=0.0.10']
+dependencies = ['cjm-context-graph-layer>=0.0.7', 'cjm-context-graph-primitives>=0.0.12']
 
 [project.urls]
 Repository = "https://github.com/cj-mills/cjm-transcript-graph-schema"
-Documentation = "https://cj-mills.github.io/cjm-transcript-graph-schema/"
-
-[project.entry-points.nbdev]
-cjm_transcript_graph_schema = "cjm_transcript_graph_schema._modidx:d"
 
 [tool.setuptools.dynamic]
 version = {attr = "cjm_transcript_graph_schema.__version__"}
@@ -30,4 +26,3 @@ version = {attr = "cjm_transcript_graph_schema.__version__"}
 [tool.setuptools.packages.find]
 include = ["cjm_transcript_graph_schema*"]
 
-[tool.nbdev]

cjm_transcript_graph_schema-0.0.8/tests/test_schema.py

@@ -0,0 +1,176 @@
+"""Tests for cjm_transcript_graph_schema.schema — identity tuples + wire shapes.
+
+Projected from the schema notebook's four test cells at the c25780e8 flip."""
+from cjm_context_graph_primitives.provenance import SourceRef
+from cjm_transcript_graph_schema.schema import (AudioRenditionNode, AudioSegmentNode,
+                                                CollectionNode, SegmentNode, SourceNode,
+                                                TranscriptNode, TranscriptSliceRef,
+                                                audio_rendition_node_id,
+                                                audio_segment_node_id,
+                                                collection_edges, collection_node_id,
+                                                normalize_collection_title,
+                                                segment_node_id, source_node_id,
+                                                transcript_node_id)
+
+SID = source_node_id("sha256:abc")
+A1 = audio_segment_node_id(SID, 0.0, 300.2)
+R_RAW = audio_rendition_node_id(A1, [])
+R_VOX = audio_rendition_node_id(A1, ["source_separation:demucs@cfg"])
+
+
+def test_identity_determinism_and_cross_type_distinctness():
+    assert SID == source_node_id("sha256:abc")
+    assert A1 == audio_segment_node_id(SID, 0.0, 300.2)
+    assert A1 != audio_segment_node_id(SID, 0.0, 300.3)
+    # renditions: empty chain (raw) is stable + distinct from any preprocessing chain
+    assert R_RAW == audio_rendition_node_id(A1, [])
+    assert R_VOX != R_RAW, "vocals rendition is a distinct node from raw"
+    assert R_RAW != A1, "the raw rendition is its own node, not the audio segment"
+    assert audio_rendition_node_id(A1, ["a", "b"]) != audio_rendition_node_id(A1, ["b", "a"]), \
+        "chain order matters"
+    # transcript / segment key on the RENDITION now
+    t1 = transcript_node_id(R_RAW, "whisper", "cfg1")
+    assert t1 != transcript_node_id(R_RAW, "voxtral", "cfg1")
+    assert t1 != transcript_node_id(R_RAW, "whisper", "cfg2")
+    assert t1 != transcript_node_id(R_VOX, "whisper", "cfg1"), "raw vs vocals transcript distinct"
+    s1 = segment_node_id(R_RAW, "vadcfg", 1.5, 4.25)
+    assert s1 == segment_node_id(R_RAW, "vadcfg", 1.5, 4.25)
+    assert s1 != segment_node_id(R_VOX, "vadcfg", 1.5, 4.25), "raw vs vocals fine spine distinct"
+    assert len({SID, A1, R_RAW, R_VOX, t1, s1}) == 6, "kinds never collide"
+
+
+def test_source_audio_segment_rendition_wire_shapes():
+    src = SourceNode(content_hash="sha256:abc", path="/media/ep1.mp3")
+    n = src.to_graph_node()
+    assert n["label"] == "Source" and n["id"] == source_node_id("sha256:abc")
+    assert n["properties"]["title"] == "ep1" and n["properties"]["root_kind"] == "ingested"
+    assert n["sources"][0]["content_hash"] == "sha256:abc"
+    assert SourceNode(content_hash="sha256:abc", path="/x.mp3",
+                      title="Custom").to_graph_node()["properties"]["title"] == "Custom"
+
+    # AudioSegment is a hashless boundary now (model-input moved to the rendition)
+    aseg = AudioSegmentNode(source=src.id, index=0, start=0.0, end=300.2,
+                            segment_path="/cuts/seg0.mp3")
+    an = aseg.to_graph_node()
+    assert an["id"] == audio_segment_node_id(src.id, 0.0, 300.2)
+    assert an["label"] == "AudioSegment"
+    assert an["properties"]["source_id"] == src.id
+    assert an["properties"]["segment_path"] == "/cuts/seg0.mp3"
+    assert "model_input_path" not in an["properties"], "model-input lives on the rendition now"
+    assert an["sources"] == [], "boundary node is hashless"
+
+    # AudioRendition owns the model-input; raw vs vocals coexist under one AudioSegment
+    raw = AudioRenditionNode(audio_segment=aseg.id, model_input_path="/cache/seg0.wav",
+                             model_input_hash="sha256:wav0")
+    rn = raw.to_graph_node()
+    assert rn["id"] == audio_rendition_node_id(aseg.id, []) and rn["label"] == "AudioRendition"
+    assert rn["properties"]["audio_segment_id"] == aseg.id
+    assert rn["properties"]["is_raw"] is True
+    assert rn["properties"]["chain"] == [] and "preprocessing" not in rn["properties"]
+    assert rn["sources"][0]["content_hash"] == "sha256:wav0"
+    re = raw.derived_edge()
+    assert re["relation_type"] == "DERIVED_FROM"
+    assert re["source_id"] == rn["id"] and re["target_id"] == aseg.id
+
+    vox = AudioRenditionNode(audio_segment=aseg.id, model_input_path="/cache/seg0_vocals.wav",
+                             model_input_hash="sha256:wav0vox",
+                             chain=["source_separation:demucs@cfg"])
+    vn = vox.to_graph_node()
+    assert vn["id"] != rn["id"], "vocals rendition coexists as a distinct node"
+    assert vn["properties"]["is_raw"] is False
+    assert vn["properties"]["preprocessing"] == "source_separation:demucs@cfg"
+    assert vn["sources"][0]["content_hash"] == "sha256:wav0vox"
+
+
+def test_transcript_shape_attribution_and_derived_edge():
+    t = TranscriptNode(rendition=R_RAW, transcriber="whisper", config_hash="cfg1",
+                       text="hello world", audio_hash="sha256:wav0",
+                       metadata={"model": "base"}, asserted_at=123.0)
+    tn = t.to_graph_node()
+    assert tn["id"] == transcript_node_id(R_RAW, "whisper", "cfg1")
+    assert tn["properties"]["actor"] == "capability:whisper"
+    assert tn["properties"]["method"] == "transcribe"
+    assert tn["properties"]["asserted_at"] == 123.0
+    assert tn["properties"]["text"] == "hello world"
+    assert tn["properties"]["metadata"] == {"model": "base"}
+    assert tn["properties"]["rendition_id"] == R_RAW
+    assert tn["sources"][0]["slice"]["kind"] == "full"
+    e = t.derived_edge()
+    assert e["relation_type"] == "DERIVED_FROM"
+    assert e["source_id"] == tn["id"] and e["target_id"] == R_RAW
+    assert e["id"] == t.derived_edge()["id"], "deterministic edge id"
+
+
+def test_segment_shape_slices_identity_and_empty():
+    acc = TranscriptSliceRef(transcript="t-acc", start_char=0, end_char=11, text="hello world")
+    var = TranscriptSliceRef(transcript="t-lw", start_char=0, end_char=10, text="helo world")
+    seg = SegmentNode(rendition=R_RAW, vad_config_hash="vadcfg", chunk_start=1.5, chunk_end=4.25,
+                      index=7, start_time=101.5, end_time=104.25, text="hello world",
+                      audio_hash="sha256:wav0", source=SID, text_from="t-acc",
+                      text_slices=[acc, var])
+    sn = seg.to_graph_node()
+    assert sn["id"] == segment_node_id(R_RAW, "vadcfg", 1.5, 4.25)
+    assert sn["properties"]["text"] == "hello world" and sn["properties"]["text_from"] == "t-acc"
+    assert sn["properties"]["source_id"] == SID and sn["properties"]["rendition_id"] == R_RAW
+    assert len(sn["sources"]) == 3
+    audio_ref, acc_ref, var_ref = sn["sources"]
+    assert audio_ref["slice"]["kind"] == "time" and audio_ref["slice"]["start"] == 1.5
+    assert audio_ref["locator"]["node_id"] == R_RAW, "audio ref points at the rendition"
+    assert acc_ref["slice"]["kind"] == "char"
+    assert acc_ref["content_hash"] == SourceRef.compute_hash(b"hello world")
+    assert var_ref["content_hash"] == SourceRef.compute_hash(b"helo world")
+
+    # identity is audio-side: text changes do NOT change the id
+    seg2 = SegmentNode(rendition=R_RAW, vad_config_hash="vadcfg", chunk_start=1.5, chunk_end=4.25,
+                       index=7, start_time=101.5, end_time=104.25, text="different text")
+    assert seg2.id == seg.id
+
+    # empty (D14-class) segment: audio ref only, no text_from
+    empty = SegmentNode(rendition=R_RAW, vad_config_hash="vadcfg", chunk_start=9.0, chunk_end=9.5,
+                        index=8, start_time=109.0, end_time=109.5, audio_hash="sha256:wav0")
+    en = empty.to_graph_node()
+    assert en["properties"]["text"] == "" and "text_from" not in en["properties"]
+    assert len(en["sources"]) == 1
+
+
+def test_collection_identity_wire_shape_and_edges():
+    # identity: normalized-title — folder-name and hand-typed forms converge
+    assert normalize_collection_title("Hardcore_History") == "hardcore history"
+    cid = collection_node_id("Hardcore_History")
+    assert cid == collection_node_id("hardcore  history")
+    assert cid != collection_node_id("Supernova in the East")
+    assert cid != SID, "kinds never collide"
+
+    # wire shape: proposed tool default vs confirmed human act (ae3464fc actor criterion)
+    proposed = CollectionNode(title="Hardcore_History", actor="cli:transcribe",
+                              asserted_at=123.0)
+    pn = proposed.to_graph_node()
+    assert pn["id"] == cid and pn["label"] == "Collection"
+    assert pn["properties"]["status"] == "proposed"
+    assert pn["properties"]["root_kind"] == "asserted"
+    assert pn["properties"]["actor"] == "cli:transcribe"
+    assert pn["properties"]["method"] == "collection-declaration"
+    assert pn["properties"]["asserted_at"] == 123.0
+    assert pn["properties"]["title"] == "Hardcore_History", "display keeps the raw title"
+    assert pn["sources"] == [], "a collection asserts grouping; it ingests nothing"
+    confirmed = CollectionNode(title="Hardcore History", status="confirmed").to_graph_node()
+    assert confirmed["id"] == cid, "confirming/renaming to an equivalent title is the same node"
+    assert confirmed["properties"]["status"] == "confirmed"
+
+    # membership edges: unordered = PART_OF only; ordered = full fractal spine
+    m1 = source_node_id("sha256:ch1")
+    m2 = source_node_id("sha256:ch2")
+    m3 = source_node_id("sha256:ch3")
+    loose = collection_edges(cid, [m1, m3])
+    assert [e["relation_type"] for e in loose] == ["PART_OF", "PART_OF"]
+    assert all(e["target_id"] == cid for e in loose)
+    ordered = collection_edges(cid, [m1, m2, m3], ordered=True)
+    rels = sorted(e["relation_type"] for e in ordered)
+    assert rels == ["NEXT", "NEXT", "PART_OF", "PART_OF", "PART_OF", "STARTS_WITH"]
+    starts = [e for e in ordered if e["relation_type"] == "STARTS_WITH"][0]
+    assert starts["source_id"] == cid and starts["target_id"] == m1
+    nexts = {(e["source_id"], e["target_id"]) for e in ordered if e["relation_type"] == "NEXT"}
+    assert nexts == {(m1, m2), (m2, m3)}
+    # late-added member attaches to the SAME node without touching the chain
+    late = collection_edges(cid, [source_node_id("sha256:ch2.5")])
+    assert late[0]["relation_type"] == "PART_OF" and late[0]["target_id"] == cid