feat: v1.0.0 SPEC v2.0 implementation

126/133 conformance fixtures passing (7 skipped). 40M property-based
round-trips with zero failures.

New: scalar.py, test_conformance_v2.py, test_roundtrip_v2.py.
Rewritten: generic.py, decode_generic.py.
Updated: encode.py, decode.py, session.py, delta.py, stream.py,
stream_generic.py (profile=graph, ##! summary, v2.0 error categories).

Author: blackwell-systems

.github/workflows/ci.yml

@@ -14,8 +14,17 @@ jobs:
         python-version: ['3.9', '3.10', '3.11', '3.12', '3.13']
     steps:
       - uses: actions/checkout@v4
+      - uses: actions/checkout@v4
+        with:
+          repository: blackwell-systems/gcf
+          path: gcf
       - uses: actions/setup-python@v5
         with:
           python-version: ${{ matrix.python-version }}
       - run: pip install -e ".[dev]" || pip install -e . && pip install pytest
-      - run: pytest tests/ -v
+      - name: Unit tests
+        run: pytest tests/ -v --ignore=tests/test_conformance_v2.py --ignore=tests/test_roundtrip_v2.py
+      - name: Conformance tests (133 fixtures)
+        run: pytest tests/test_conformance_v2.py -v
+      - name: Property-based round-trip (100K random + 100K adversarial)
+        run: pytest tests/test_roundtrip_v2.py -v

CHANGELOG.md

@@ -1,5 +1,25 @@
 # Changelog
 
+## v1.0.0 (2026-06-10)
+
+SPEC v2.0 implementation. 126/133 conformance fixtures passing (7 skipped: session, delta, binary UTF-8, negative zero, graph encode). 40M property-based round-trips with zero failures.
+
+### Breaking changes from v0.5.0
+
+- `encode_generic` emits `GCF profile=generic` header
+- `decode_generic` requires `GCF profile=` header
+- Strings colliding with typed literals are quoted
+- Full JSON string escaping and number grammar
+- `-` for null, `~` for absent, `^` for nested attachments
+- `##! summary` trailer replaces `## _summary`
+- Graph encoder emits `profile=graph`
+
+### New
+
+- `scalar.py`: common scalar grammar (quoting, escaping, parsing, number formatting)
+- Conformance test runner (133 fixtures)
+- Property-based round-trip tests (40M verified, configurable via `GCF_ITERATIONS`)
+
 ## v0.5.0 (2026-06-06)
 
 - `GenericStreamEncoder`: zero-buffering tabular streaming encode (begin_array/write_row/end_array/write_kv/write_section/write_inline_array)

src/gcf/decode.py

@@ -35,7 +35,11 @@ def decode(input_text: str) -> Payload:
     _parse_header(header[4:], p)
 
     if not p.tool:
-        raise DecodeError("header missing required 'tool' field")
+        raise DecodeError("missing_tool: header missing required 'tool' field")
+
+    # Detect delta mode.
+    is_delta = "delta=true" in header
+    valid_delta_sections = {"removed", "added", "edges_removed", "edges_added"}
 
     # Parse body: symbols and edges.
     symbols: list[Symbol] = []
@@ -48,13 +52,19 @@ def decode(input_text: str) -> Payload:
         if not line:
             continue
 
+        # Skip ##! summary trailer.
+        if line.startswith("##! "):
+            continue
+
         # Group header.
         if line.startswith("## "):
             group = line[3:]
             # Strip bracket suffix: "edges [200]" -> "edges"
             bracket_idx = group.find(" [")
             if bracket_idx >= 0:
                 group = group[:bracket_idx]
+            if is_delta and group not in valid_delta_sections:
+                raise DecodeError(f"malformed_delta: invalid delta section {group!r}")
             in_edges = group == "edges"
             if not in_edges:
                 if group == "targets":
@@ -113,19 +123,19 @@ def _parse_header(fields: str, p: Payload) -> None:
 def _parse_symbol_line(line: str, distance: int) -> tuple[Symbol, int]:
     """Parse a symbol line into a Symbol and its local ID."""
     if not line.startswith("@"):
-        raise DecodeError(f"expected symbol line starting with @, got {line!r}")
+        raise DecodeError(f"invalid_node_line: expected symbol line starting with @, got {line!r}")
 
     parts = line.split()
     if len(parts) < 5:
         raise DecodeError(
-            f"symbol line needs at least 5 fields, got {len(parts)} in {line!r}"
+            f"invalid_node_line: symbol line needs at least 5 fields, got {len(parts)} in {line!r}"
         )
 
     id_str = parts[0][1:]  # strip @
     try:
         sym_id = int(id_str)
     except ValueError as e:
-        raise DecodeError(f"invalid symbol id {id_str!r}: {e}") from e
+        raise DecodeError(f"invalid_symbol_id: invalid symbol id {id_str!r}: {e}") from e
 
     kind = parts[1]
     kind = KIND_EXPAND.get(kind, kind)
@@ -135,7 +145,7 @@ def _parse_symbol_line(line: str, distance: int) -> tuple[Symbol, int]:
     try:
         score = float(parts[3])
     except ValueError as e:
-        raise DecodeError(f"invalid score {parts[3]!r}: {e}") from e
+        raise DecodeError(f"invalid_score: invalid score {parts[3]!r}: {e}") from e
 
     provenance = parts[4]
 
@@ -157,7 +167,7 @@ def _parse_edge_line(line: str, sym_by_id: dict[int, Symbol]) -> Edge:
     ref = parts[0]
     lt_idx = ref.find("<")
     if lt_idx < 0:
-        raise DecodeError(f"edge line missing '<' separator in {ref!r}")
+        raise DecodeError(f"invalid_edge_syntax: edge line missing '<' separator in {ref!r}")
 
     target_id_str = ref[1:lt_idx]  # strip leading @
     source_id_str = ref[lt_idx + 2:]  # strip <@
@@ -176,7 +186,7 @@ def _parse_edge_line(line: str, sym_by_id: dict[int, Symbol]) -> Edge:
     source_sym = sym_by_id.get(source_id)
     if target_sym is None or source_sym is None:
         raise DecodeError(
-            f"edge references unknown symbol id(s): target={target_id} source={source_id}"
+            f"unknown_edge_reference: edge references unknown symbol id(s): target={target_id} source={source_id}"
         )
 
     edge_type = parts[1]