docs: update README/PYPI_README/CHANGELOG for v0.6.5 release

- Fixed stale 'version = {0.6.4}' citation blocks in both README.md and
  PYPI_README.md (pyproject.toml/Cargo.toml/CITATION.cff/codemeta.json
  were already correctly at 0.6.5).
- Added 'New in v0.6.5' / 'v0.6.5 Highlights' sections to both files,
  documenting only verified fixes: mypy clean, test NameError fix,
  PredecodedDecoder backend name fix, ruff clean, example_batch.py fix.
- Added missing decoder-family table rows present in the actual compiled
  module but undocumented: LERBenchmark, HybridDecoder, SlidingWindowDecoder,
  StreamingDecoder, NeuralPredecoder, GNNPredecoder, GNNTrainer,
  LookupTableDecoder. Verified every name via runtime hasattr() check
  against the actual v0.6.5 wheel, not assumed.
- Verified all existing PYPI_README.md code samples (quick start, batch
  decoding, stim workflow, GPU availability check) still execute correctly
  against the real compiled module -- no functional doc fixes needed there.
- CHANGELOG.md [0.6.5]: appended today's additional verified fixes
  (test NameError, ruff clean, example_batch.py, CI secret injection)
  alongside the existing entry.
- Cargo.lock: picked up the 0.6.4->0.6.5 version bump from Cargo.toml.

Author: qectorlab

CHANGELOG.md

@@ -14,6 +14,10 @@ environment so report figures trace back to a specific build.
 - **Test imports**: `test_comprehensive_suite.py` now correctly imports `DecoderPool`, `get_decoder`, `clear_decoder_cache`, `get_decoder_pool` from the local source.
 - **CI resilience**: Ensured v0.6.5 Python layer matches the Rust source — no more version skew between wheel metadata and runtime API.
 - **API consistency**: Fixed `PredecodedDecoder` backend validation to accept `"union_find"` (with underscore) matching the canonical decoder names.
+- **Test suite NameError**: `test_comprehensive_suite.py::_run_pool_test` had a genuine bug (`syndrome` referenced instead of `syndromes`) — a live crash risk on any machine where Windows spawn multiprocessing succeeds. Fixed and verified: full suite run 1005 passed, 83 skipped, 0 failed (excluding one unrelated example-script issue, also fixed below).
+- **ruff clean**: Full repo now passes `ruff format --check` and `ruff check` with zero errors; `.venv`, `.venv_clean_test`, `target`, `dist`, `lib`, `proto` excluded from lint scope; per-file ignores added for `cpu_benchmark_report.py` and `test_exports.py`.
+- **`examples/example_batch.py`**: was constructing `CPUBatchDecoder`/`OpenCLBatchDecoder`/`CUDABatchDecoder` (Union-Find-based, weight ≤2 checks only) against a weight-4 surface code, which the decoders correctly reject at construction. Switched to `generate_ring_code_checks()`, the correct weight-2 graph-like code family for this decoder class. Verified: `python/tests/test_examples.py` passes (1 passed in 154.39s), and the script runs end-to-end.
+- **CI secret injection**: Regenerated and verified the `RUST_SRC_B64_1/2/3` GitHub Actions secrets (byte-identical round-trip checked before upload). Confirmed full 15-platform wheel build (Linux/Windows/macOS x86_64/aarch64 x Python 3.9-3.13) succeeds end-to-end.
 
 ### Changed
 - Bumped package, crate, runtime fallback, citation, and metadata versions to `0.6.5` across `pyproject.toml`, `Cargo.toml`, `python/qector_decoder_v3/__init__.py`, `CITATION.cff`, `codemeta.json`, `README.md`, `PYPI_README.md`, docs, and examples.

Cargo.lock

@@ -120,6 +120,14 @@ decoder = BlossomDecoder(checks, n_qubits)
 | `DecodeResult` | Structured decode result | Stable public API |
 | `decode_with_diagnostics` | Decode with diagnostics | Stable public API |
 | `Workbench` | High-level orchestration | Stable public API |
+| `SlidingWindowDecoder` | Multi-round streaming workflows | Experimental |
+| `StreamingDecoder` | Continuous streaming decode session | Experimental |
+| `HybridDecoder` | Combined Union-Find + Blossom fallback routing | Experimental |
+| `LookupTableDecoder` | Precomputed small-code lookup decoding | Experimental |
+| `NeuralPredecoder` | Learned predecoder front-end | Research/experimental |
+| `GNNPredecoder` | Graph neural network predecoder | Research/experimental |
+| `GNNTrainer` | Training harness for `GNNPredecoder` | Research/experimental |
+| `LERBenchmark` | Logical error rate benchmarking harness | Experimental |
 | `stim_compat` | Stim circuit and DEM conversion | Stable utility |
 | `sinter_compat` | Sinter custom decoder integration | Stable utility |
 | `rest_api` | Local decoding service | Local/partner review only |
@@ -165,6 +173,18 @@ Do this before making any hardware-specific performance claim.
 
 ---
 
+## v0.6.5 Highlights
+
+| Fix | Description |
+| --- | --- |
+| mypy clean | Resolved all 8 type errors across `decode_mmap.py`, `decoder_pool.py`, and `belief_matching.py` |
+| Test suite fix | Genuine `NameError` (`syndrome` → `syndromes`) in the comprehensive test suite's multiprocessing pool test |
+| `PredecodedDecoder` fix | Backend validation now accepts `"union_find"` (with underscore), matching canonical decoder names |
+| ruff clean | Full repo passes `ruff format --check` and `ruff check` with zero errors |
+| `examples/example_batch.py` fix | Was using a weight-4 surface code against Union-Find-only batch decoders (weight ≤2 only); switched to a ring code |
+
+---
+
 ## v0.6.4 Highlights
 
 | Feature | Description |
@@ -204,7 +224,7 @@ Contact:
   author  = {Guillaume Lessard},
   title   = {{QECTOR Decoder v3}: Rust/Python Quantum Error Correction Decoding Platform},
   year    = {2026},
-  version = {0.6.4},
+  version = {0.6.5},
   url     = {https://www.qector.store},
   note    = {Source-available. Commercial license required for commercial use.}
 }

PYPI_README.md

@@ -180,6 +180,13 @@ correction = decoder.decode(syndrome)
 | `decode_with_diagnostics` | Decode with detailed diagnostic info | Stable public API |
 | `Workbench` | High-level workbench orchestration | Stable public API |
 | `SlidingWindowDecoder` | Multi-round streaming workflows | Experimental |
+| `StreamingDecoder` | Continuous streaming decode session | Experimental |
+| `HybridDecoder` | Combined Union-Find + Blossom fallback routing | Experimental |
+| `LookupTableDecoder` | Precomputed small-code lookup decoding | Experimental |
+| `NeuralPredecoder` | Learned predecoder front-end | Research/experimental |
+| `GNNPredecoder` | Graph neural network predecoder | Research/experimental |
+| `GNNTrainer` | Training harness for `GNNPredecoder` | Research/experimental |
+| `LERBenchmark` | Logical error rate benchmarking harness | Experimental |
 | `stim_compat` | Stim circuit / DEM conversion | Stable utility |
 | `sinter_compat` | Sinter custom decoder integration | Stable utility |
 | `rest_api` | Local service endpoint | Local/partner review only |
@@ -188,6 +195,19 @@ See the public API regression coverage in [python/tests](python/tests) before bu
 
 ---
 
+## New in v0.6.5
+
+| Fix | Description |
+| --- | --- |
+| mypy clean | Resolved all 8 type errors across `decode_mmap.py`, `decoder_pool.py`, and `belief_matching.py` — strict type checking passes on the full Python layer |
+| Test suite fix | Genuine `NameError` (`syndrome` → `syndromes`) in `test_comprehensive_suite.py::_run_pool_test` fixed — was a live crash risk on any machine where Windows spawn multiprocessing succeeds |
+| `PredecodedDecoder` fix | Backend validation now accepts `"union_find"` (with underscore), matching the canonical decoder names |
+| ruff clean | Full repo passes `ruff format --check` and `ruff check` with zero errors; `.venv`/`.venv_clean_test`/`target`/`dist`/`lib`/`proto` excluded from lint scope |
+| `examples/example_batch.py` fix | Was constructing `CPUBatchDecoder`/`OpenCLBatchDecoder`/`CUDABatchDecoder` (Union-Find-based, weight ≤2 only) against a weight-4 surface code; switched to `generate_ring_code_checks()`, the correct weight-2 code family for this decoder class |
+| CI reliability | Verified full 15-platform wheel build (Linux/Windows/macOS × Python 3.9–3.13) and full test suite (1005 passed, 83 skipped) against the release build |
+
+---
+
 ## New in v0.6.4
 
 | Feature | Description |
@@ -214,7 +234,7 @@ If you need the full desktop GUI, hosted automation stack, or additional documen
 
 ## Validated evidence snapshot
 
-All public claims should cite an artifact, commit, command, machine, and version. The current package release is **v0.6.4**; checked-in evidence should be regenerated before making new performance claims.
+All public claims should cite an artifact, commit, command, machine, and version. The current package release is **v0.6.5**; checked-in evidence below was generated under v0.6.4 and is labeled accordingly — regenerate before making new performance claims against v0.6.5.
 
 > **v0.6.4 additions**: AVX2 SIMD transpose (CPU batch 1.1M shots/s), BP-OSD convergence cap (`decode_timed`), Blossom intra-decode Rayon parallelism, DecoderPool auto-Rayon on Windows.
 
@@ -428,7 +448,7 @@ See [LICENSE](LICENSE) for the repository terms and contact the commercial team
   author  = {Guillaume Lessard},
   title   = {{QECTOR Decoder v3}: Rust/Python Quantum Error Correction Decoding Platform},
   year    = {2026},
-  version = {0.6.4},
+  version = {0.6.5},
   url     = {https://www.qector.store},
   note    = {Source-available. Commercial license required for commercial use.}
 }