Layout message v4 flag bit 0 (H5D_CHUNK_DONT_FILTER_PARTIAL_CHUNKS, set
with H5Pset_chunk_opts) makes libhdf5 store every chunk that extends past
the dataset's extent without the filter pipeline, while its filter mask
still reads 0. The parser ignored the flag, so readers tried to inflate
raw bytes: libhdf5's own h5fc_edge_v3.h5 failed with "deflate: ...
unknown compression method".
DataLayout::Chunked gains dont_filter_partial_edge_chunks (always false
for v3), and list_chunks — the one place every read path gets its chunk
list from — marks such partial chunks as having skipped every filter, so
the full, cached, indexed, parallel and selection readers all copy them
as-is. chunked_write.rs gets `..` in one exhaustive test pattern for the
new field.
Regression: libhdf5_edge_chunk_fixture_reads (h5fc_edge_v3.h5 from the
HDF5 tools test files, committed as a 2.5 KB fixture), and
h5py_unfiltered_partial_edge_chunks_read (the flag set through h5py's
bundled libhdf5 via ctypes, as h5py has no binding for it: fixed array,
extensible array and B-tree v2 indexes, 1-D and 2-D, plus a hyperslab
of the last chunk), and v4_chunked_dont_filter_partial_edge_chunks_flag.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
The Fixed Array writer always packed every element into one data block
behind one checksum. Past 2^10 elements libhdf5 (and our reader) expect a
paged block: a page-init bitmap after the prefix, then one checksummed page
per 1024 elements. Any dataset with more than 1024 chunks and no unlimited
dimension failed with "incorrect metadata checksum" in h5py, h5dump and
our own reader.
build_fixed_array_at now takes one Option<WrittenChunk> per array slot so
later fixes can leave unallocated slots.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
Datatype::serialize returned an empty message for these four classes, so
any dataset or attribute of them (including a Raw attribute copied from
another file) was unreadable by libhdf5 ("ran off end of input buffer
while decoding"). They now encode exactly as libhdf5 does: legacy object
and region references as datatype version 1, H5T_STD_REF kinds as version
4 with their encoding version, opaque tags NUL-padded to 8 bytes.
Parsing an opaque tag now stops at its first NUL, so libhdf5's padding
no longer becomes part of the tag. Datatype::check_encodable rejects
what has no encoding (an opaque tag over 248 bytes); FileWriter::finish
calls it for every dataset and attribute type.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
Every decode stage was capped at the chunk's decoded size. That holds
only when every filter ahead of the codec preserves size; Fletcher32
does not (it appends a 4-byte checksum), so a pipeline with Fletcher32
before deflate (NetCDF-4's fletcher32 -> shuffle -> deflate ordering,
h5repack's "all filters") failed with "deflate: output exceeds size
limit" on every chunk.
decompress_chunk_masked now computes each stage's bound by running the
chunk size forward through the filters that precede it in write order
(and that the chunk's mask did not skip): shuffle keeps the size,
Fletcher32 adds 4, any codec adds at most n/8 + 64. The cap is still a
small constant factor of the chunk, so a decompression bomb is rejected
as before (tested).
Shuffle also had to learn libhdf5's handling of a length that is not a
whole number of elements (chunk + checksum): shuffle the whole elements
and leave the trailing bytes in place, in both directions. It used to
refuse such data.
Regression: h5py_fletcher32_before_deflate_reads (fletcher->shuffle->
gzip, fletcher->gzip, shuffle->fletcher->gzip, and a 2-D i32 grid),
fletcher32_ahead_of_deflate_stays_bounded and
shuffle_leaves_a_partial_trailing_element_in_place.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
Pcodec chunks were written as filter 32023, which the HDF Group registry
assigns to Granular BitRound (GBR). Pcodec has no registered ID (checked
2026-09-25 against hdf5_plugins/docs/RegisteredFilterPlugins.md, which
ends at 32033 with no pcodec entry). GBR's decode is a pass-through, so
libhdf5 with that plugin loaded would have returned the compressed bytes
as the dataset's values.
Write pcodec as 480, from the registry's testing/private range (256-511),
named "pcodec (clawhdf5 private)", and document it as non-interoperable:
only clawhdf5 with the `pcodec` feature reads it. Chunks under 32023 are
still read as pcodec when the filter is named exactly "pcodec" (what
clawhdf5 <= 2.7.0 wrote); any other 32023 is UnsupportedFilter.
Test: pcodec_uses_private_id_and_reads_legacy_32023.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
A chunk's filter mask has one bit per pipeline filter; bit i set means
filter i was not applied to that chunk (an optional filter that
declined, or a direct chunk write). Every read path treated any nonzero
mask as "no filters applied" and returned the stored bytes, so a chunk
that skipped only gzip in a shuffle+gzip pipeline came back still
shuffled (h5py write_direct_chunk with filter_mask=0b10: 8 of 32 values
wrong).
decompress_chunk_masked undoes the filters the mask leaves set and skips
the rest; an unsupported filter is no longer an error when the chunk
skipped it. The full, cached, sweep, indexed, parallel and selection
(partial_read) paths all use it, and a chunk is copied straight from the
file only when every filter was skipped. decompress_chunk is the mask-0
case.
Regression: h5py_partial_filter_mask_skips_only_masked_filters (1-D
shuffle+gzip with masks 0, 0b10 and 0b11; 2-D with 0b01; full and
hyperslab reads) and filter_mask_skips_only_the_masked_filters.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
Filter 32015 chunks were written with the streaming encoder
(zstd::encode_all), whose frames carry no content size. The registered
HDF5 Zstandard filter (H5Zzstd.c, libhdf5 + hdf5plugin) sizes its output
from ZSTD_getFrameContentSize and fails on such frames, so h5py could not
read our zstd datasets ("filter returned failure during read"). Compress
with the one-shot API, which records the size.
Tests: zstd_frames_record_content_size (content size was None before),
hdf5plugin_reads_our_zstd (ignored interop test; failed before).
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
Both indexes place each chunk at a linear index computed from the
dataset's maximum dimensions (libhdf5's max_down_chunks), and the
Extensible Array first swizzles its unlimited dimension to the slowest
position. We linearised by the current dimensions, so any dataset whose
shape was smaller than its maxshape, or whose unlimited dimension was not
the first, read back scrambled without an error: h5py libver="latest"
files with maxshape (10, None) or (20, 10), and the libhdf5 test files
h5fc_ext*.h5 and test_ld.h5.
The linearisation now lives in chunk_grid (shared with the writers), and
slots beyond the current extent are ignored as the library does.
read_fixed_array_chunks / read_extensible_array_chunks take the
dataspace's max dimensions.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
A v2 object header message has a 2-byte size field. The writer truncated
larger sizes to 16 bits, so an attribute over ~64 KiB (or a compact
dataset of 65532-65535 bytes, whose layout message adds 4 bytes) produced
a file libhdf5 rejects ("message of unshareable class flagged as
shareable", "bad flag combination").
ObjectHeaderWriter::serialize now returns a Result and fails on any message
over MAX_MESSAGE_SIZE; FileWriter::finish propagates it. Compact storage
falls back to contiguous above 65531 bytes, the real limit. Dense storage
for large attributes remains future work.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
Filter 32004 chunks were framed as a 4-byte little-endian size plus one
LZ4 block. That is not the registered HDF5 LZ4 format (H5Zlz4.c: 8-byte
big-endian total size, 4-byte big-endian block size, then per block a
4-byte big-endian compressed length and the block, stored raw when the
length equals the block size), so libhdf5 + hdf5plugin could not read
our LZ4 datasets and we could not read theirs (h5ex_d_lz4.h5:
"lz4: 0 is not a valid match offset").
Write the registered format (cd_values[0] is honoured as the block size,
default 1 GiB like the plugin) and read it, multi-block and raw blocks
included. Chunks in the old framing stay readable: an HDF5 chunk is under
4 GiB, so a registered chunk always starts with four zero bytes and is at
least 12 bytes long, while an old one starts with four zero bytes only
when empty (5 bytes).
Tests: lz4_reads_registered_hdf5_format (chunk of the HDF Group's
h5ex_d_lz4.h5, block size 3), lz4_writes_registered_hdf5_format,
lz4_reads_legacy_clawhdf5_format, and hdf5plugin_reads_our_lz4 (ignored
interop test; failed before with "filter returned failure during read").
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
A type-1 (raw data chunk) B-tree key holds the chunk size, the filter
mask and one offset per dimension, and those offsets are always 8 bytes:
they are dataset coordinates, not file addresses. The reader used the
superblock's size-of-offsets for them, so in a file with 4-byte offsets
every key was misparsed. Unfiltered chunked datasets read as zeros (with
stray bytes where a misread address landed on data) and filtered ones
failed with "deflate: truncated stream".
Only the sibling and child addresses follow size-of-offsets now. The
unit-test B-tree builder wrote keys the same wrong way, which is why its
tests passed; it now matches the format.
Regression: h5py_four_byte_offsets_chunked_reads (h5py, set_sizes(4, 4)
and (4, 8); 1-D and 2-D, unfiltered and gzip) and the unit test
collect_chunks_with_four_byte_addresses.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
An 82-byte fuzz_btree_v2 crash input from 2026-09-20 was left untracked
in fuzz/artifacts. Replayed today it runs cleanly: the depth cap and
record budget added to B-tree v2 traversal that day fixed it. It is now
in the committed fuzz corpus, and a robustness test replays the fuzz
target's exact code path on it so a regression fails CI rather than
waiting for someone to run the fuzzer.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
The previous commit said clawhdf5 has no integration at all. ClawBrainHub
(clawverse/clawbrainhub) does use it: cbh-core reads and writes .brain
files through the facade, cbh-scanner uses the facade, and cbh-cli uses
clawhdf5_agent::bm25::BM25Index, all via path dependencies on this repo.
Checked on 2026-09-25 against main: it builds on its pinned toolchain and
its 204 tests pass. CLAUDE.md now records that, and that path
dependencies mean API changes here reach it directly.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
CLAUDE.md said ZeroClaw "imports this as a Cargo feature (clawhdf5
feature flag)" and uses clawhdf5 as its memory backend; the agent crate
called itself the "ZeroClaw agent memory HDF5 backend"; the migrator
claimed to read "the ZeroClaw layout". Checked on 2026-09-25 against
ZeroClaw v0.8.5 (its latest release), the osobh/zeroclaw fork (on
v0.8.5) and both histories back to February 2026:
- no `clawhdf5` feature, dependency or memory backend has ever existed
in ZeroClaw; its backends are sqlite, lucid, postgres, qdrant,
markdown and none, behind its own `Memory` trait;
- ZeroClaw's SQLite schema is a single `memories` table (id, key,
content, category, embedding, created_at, updated_at); the
migrator's memory_chunks/sessions/entities/relations layout never
existed in ZeroClaw, so it cannot read a ZeroClaw database.
Decision: withdraw the claims (as with OpenClaw); clawhdf5 is a
standalone library with no framework integration. The migrator's
default layout is documented as its own. ZEROCLAW_VERSION keeps its name
and value (it is the persisted `edgehdf5_version` writer tag) with a
doc comment saying it is unrelated to ZeroClaw.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
The docs described a "drop-in" OpenClaw memory backend enabled with
`memory.backend = "clawhdf5"`. Checked against OpenClaw's source and
docs (v2026.2.26 through v2026.9.6): that config was never valid —
v2026.2-v2026.7 accepted only "builtin"/"qmd" and rejected unknown
keys, so a Gateway given it refuses to start, and v2026.8.1 (OpenClaw
2.0) removed the key. No plugin was ever built (no manifest, no
registration, no tools), nothing was tested against OpenClaw, the
linked github.com/redclawsystems/openclaw is a 404, and
@redclaw/clawhdf5 was never published.
Decision (2026-09-25): not pursuing an OpenClaw plugin for now; ZeroClaw
is the integration target.
- Remove openclaw-integration.md, openclaw-config.md and
migration-guide.md; add docs/openclaw.md: the status, what a memory
plugin needs against v2026.9.6 (plugins.slots.memory, manifest with
kind "memory", registerMemoryCapability / MemorySearchManager,
prebuilt native packages), and what this repo has as building blocks.
- README, QUICKSTART, USE_CASES, ROADMAP (Track 7 withdrawn), CLAUDE.md
and the `openclaw` module docs describe ClawhdfBackend as what it is:
a Markdown-oriented library backend, not an OpenClaw plugin. The
QUICKSTART example is corrected (the old one called a three-argument
create that does not exist) and states its limits.
- packages/clawhdf5-node: marked unpublished and broken, "private": true
so it cannot be published by accident; its bugs (snake_case vs
camelCase fields, wrong addon path, no way to store an embedding,
wrong WAL name) are recorded in docs/known-issues.md.
- Two broken rustdoc links fixed along the way.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
Makes the README's "cryptographically verifiable memory" true.
With HDF5Memory::set_signing_key(key), every checkpoint stores a signed
manifest of the store: a SHA-256 per memory record (text, embedding as
stored, channel, timestamp, session, tags, deleted flag, activation) in
a Merkle tree, plus hashes of the settings (and WAL mark), sessions and
knowledge graph. The signature, public key and manifest hashes go in
/meta; the per-record hashes in /integrity/record_hashes, so
HDF5Memory::verify(path, &public_key) can say which records changed, not
just that something did. A forged manifest fails the signature.
Decisions, as agreed:
- the key is set on the open store and never persisted;
- a signed store refuses to checkpoint without its key
(MemoryError::SigningKeyRequired); remove_signature() is the
deliberate way back to unsigned;
- checkpoints only: saves still in the WAL are not covered, and verify
reports how many there are.
The hashes cover exactly what the file persists, in the form the loader
returns it (strings lose trailing NULs; an empty WAL mark is not
written), so untouched stores verify across any number of reopen and
checkpoint cycles. MemoryError becomes #[non_exhaustive] (it already
gains variants in this unreleased version).
CLI: keygen (owner-only key file), --signing-key / CLAWHDF5_SIGNING_KEY
on writing commands (create signs immediately), verify --public-key
(JSON; exit 2 if not valid), `signed` in create/stats output.
Tests: reopen/checkpoint cycles with awkward strings (f16 and f32),
refusal without the key, wrong and rotated keys, eight kinds of edit
each detected and located, a forged manifest, unsigned stores, NULs in
text, and an edit made in place with h5py that verify pinpoints.
Cost on tank (search_harness --signing-study --full, 3 runs): ~20% of a
checkpoint (+9 ms at 10K, +89-112 ms at 100K), verify 18.6 ms / 247 ms,
32 bytes per record in the file. New deps ed25519-dalek, sha2,
rand_core: pure Rust, the no-C check passes.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
It enabled nothing — the agent layer is always built — yet the README,
QUICKSTART and USE_CASES told people to pass it. Removed, with those
snippets fixed: they now depend on the git repository (nothing is on
crates.io, so `version = "2.0"` never resolved) and USE_CASES no longer
presents the `float16` feature as half-precision storage (that is
MemoryConfig::float16, on by default for new stores).
Breaking for anyone passing `features = ["agent"]`: drop it.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
consolidation_efficiency never finished: stopped after 19 minutes on one
core while building its 100K case. Not the consolidation cycle (linear:
17 us at 100 records, 2.16 ms at 10K) but the setup — every add_memory
scores the new record's novelty against the whole working tier, the
benchmark lets that tier reach 50K, and each comparison recomputed both
norms: ~5e9 comparisons of three passes each.
ImportanceScorer::score_surprise now computes the new record's norm
once, takes each comparison in one fused, 8-lane pass (dot product and
the other norm together), and splits a working tier of 4096+ records
across threads with the `parallel` feature. Same results: tested against
the old cosine formula, including shorter, empty and zero vectors and
the parallel path. The work stays quadratic in the working-tier size by
design; with regular consolidation the tier stays near
working_capacity (100) and inserts are cheap.
The complete run takes 8 min 10 s on tank and fills in the 100K cycle
row (46.66 ms) and the memory-reduction table, which had never been
published. The binary no longer prints a record-count ratio as a
"BM25 Speedup" (never measured; Part 1 measures search latency) or
claims sub-linear cycle scaling (its own numbers grow slightly faster
than linearly).
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
Every undated or pre-September section re-run on one machine on one day
(tank, AMD Ryzen 7 7800X3D, 2026-09-24, commit 5c8323c), 24 commands run
serially with the load average checked before each, with the command
recorded for each section. A separate check traced every changed number
back to the raw output; its corrections are applied (e.g. the on-disk
~820 B/record is float16 plus always-deflated text on a synthetic corpus
of 40 distinct texts, not float16 alone).
Two apparent regressions were isolated rather than published:
- knowledge-graph traversal: a real bug, fixed in the previous commit;
- the write path: v2.3.0 built and run on the same machine measures the
same as today, so the old 18 us / 6.17 ms figures (undated, other
hardware) are not reproducible; float16 adds ~2 us per save and the
int8 index nothing (both isolated by switching the bench's config).
Also:
- new multimodal_bench: cross-modal search at 1K/10K records, which the
README claimed but nothing measured;
- footprint_bench reports whether it built float16 or f32 stores and
takes --f32 (it kept printing "f32" after the default changed);
- README: performance tables, the "Why" table figures and the SQLite
migration section (from the previous migrate commit);
- CHANGELOG for this branch.
Not re-run: consolidation_efficiency's 100K row and its memory-reduction
part (stopped for time), and cross_platform.sh.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
bfs_neighbors and spreading_activation built an adjacency index over the
whole graph on every call (1efd82c), so a 2-hop BFS over 1K entities
paid to index every entity and relation first: 155 us, 6.5x the 24 us
the README quoted. Found by the dated benchmark re-run.
The index is now cached on KnowledgeCache and checked against a
fingerprint of the graph on each use — one pass over entity ids and
relation endpoints, no allocation — so any change, including direct
edits of the public entities/relations Vecs (schema.rs's load path
pushes to them), still triggers a rebuild. A test edits the graph
directly in every way (push, in-place rewire, pop + push at equal
length) between traversals.
tank, 2026-09-24: BFS 1K entities 155.1 -> 23.1 us, 100 entities
17.5 -> 5.23 us, spreading activation 100 22.8 -> 10.1 us.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
clawhdf5-migrate wrote a layout of its own (/chunks, /sessions,
/entities, /relations, root attributes, no /meta or schema_version) that
HDF5Memory::open rejects, so a "migrated" SQLite database could not be
used as agent memory — contrary to the README.
It now writes through the agent's own API (HDF5Memory::create/open,
save_batch, the session cache and the knowledge graph), so there is no
second copy of the schema:
- sessions and entities/relations carry over; deleted rows become
deleted records (or are left out with --skip-deleted);
- embeddings follow the library default (float16), --f32 opts out and
--float16 is a hidden no-op, as in clawhdf5-cli; the `half`-based
conversion is gone;
- every source row is checked before the output is created: a wrong
embedding length, an empty embedding, a dimension that differs from
an existing store's, or a float16 value beyond +-65504 is an error
naming the chunk id, and an existing store is left untouched;
- --incremental opens the existing store, adds only rows it does not
hold (matched by content) and follows the source's deleted flags;
- a source with no memory rows needs --embedding-dim;
- validation reads the result back with HDF5Memory::open_read_only,
compares every field (embeddings bit for bit, round_to_f16 of the
source for float16) and checks a migrated record is found by search.
clawhdf5-agent gains HDF5Memory::sessions()/sessions_mut(),
HDF5Memory::delete_batch (one save, all-or-nothing, no auto-compact),
SessionCache::add_at, and re-exports SessionCache/SessionEntry.
The old layout's per-dataset SHA-256 provenance attributes have no place
in the agent schema and are gone. An adversarial review found two
blockers (silent truncation of long embeddings; an --incremental
dimension check that could never fire) and four majors (a failed run
wiping the existing store, dim-0 stores, deleted-flag drift); all are
fixed with regression tests. 42 migrate tests, incl. h5py opening a
migrated store.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
MemoryConfig::float16 now defaults to true for new stores, on
measurement: on the full LongMemEval haystack with real MiniLM
embeddings every retrieval metric matched f32 (previous commit), and at
100K the file is 48% smaller with faster checkpoints and opens.
Existing stores are unaffected: every agent store has recorded
`float16 = false` in /meta and keeps it. A test opens the v2.5.0
fixture, saves and checkpoints, and checks the embeddings are still f32
with the old rows bit-identical; another checks a new store is float16.
CLI: `create --f32` opts out; like `--f32-index` it only ever switches
the default off. `--float16` is still accepted and now a no-op.
Values beyond +-65504 are refused, so f32 remains the choice for
unnormalised vectors — the upgrade note says so.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
`longmemeval_bench --float16` builds every per-question store with
MemoryConfig::float16, so the vector stage searches half-rounded
embeddings exactly as such a store holds them.
Full longmemeval_s (500 questions, ~494 turns each) with real
all-MiniLM-L6-v2 embeddings, f32 vs float16, on tank (CUDA): identical
at every Hit@k and MRR, turn and session level, in all eight modes —
bar RRF session MRR 0.9253 vs 0.9254 and one or two flips out of ~320
in which gold session ranks first. The f32 run reproduces the published
hybrid numbers exactly. The earlier float16 evidence was synthetic
clustered data only; this is the real-embedding check.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
`HDF5Memory::search(query_embedding, query_text, &SearchOptions)` is the
store's full search path. `SearchOptions::new(k)` is plain hybrid search
with the tuned default fusion; each further stage is opt-in:
- `with_sources([..])`: only records from these source channels. The
filter applies before ranking, so a filtered search still returns up
to k results, normalised over what it can return. The HNSW pool is
over-fetched in proportion to what the filter removes, and the allowed
records are scanned exactly whenever that costs fewer distance
evaluations than the index would (~pool x M) — and as the fallback if
the pool comes back short. Keyword matches are filtered too.
- `with_rerank(ReRankConfig)` re-ranks a max(3k, 10) candidate pool by
relevance, recency, source authority and activation;
`with_confidence(ConfidenceConfig)` drops low-confidence results;
`at_time(now)` pins the recency clock.
These were reachable only through the OpenClaw backend, which is now
`search` with both on. Its Hebbian boost now goes to the k results it
returns rather than the whole 3k candidate pool. `hybrid_search` and
`hybrid_search_with` are wrappers and unchanged (tested bit for bit).
Measured on tank (search_harness --options-study --full, 3 runs): at
100K every filter — 50%, 10%, 1% of the store, and records far from the
query — returns the exact filtered top 10, and none is slower than an
unfiltered search (1%: 2.3 ms vs 4.6 ms). Re-rank + confidence costs
about 3%. A first version decided between index and exact scan by pool
size vs store size; it measured 0.976 recall at 12.3 ms on the
far-from-query filter, which is why the rule compares costs instead.
Tests: tests/search_options.rs (filter correctness and full pages via
both paths, far-from-query fallback, edge cases, equality with
hybrid_search_with, re-rank recency, confidence, boost scope).
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
The setting was persisted in /meta and otherwise ignored: embeddings
were always written as f32. It now does what it says.
clawhdf5-format:
- `DatasetBuilder::with_f16_data` writes IEEE binary16 (numpy float16),
rounding to nearest-even, and `make_f16_type`.
- `clawhdf5_format::float16` holds the f32 <-> f16 conversions, the one
implementation the writer, the reader and the agent all use. Checked
against the `half` crate on 16.7M f32 values and round-trips all 65536
half values; the h5py interop tests confirm the rounding matches
numpy's bit for bit (4020 values incl. ties, subnormals, overflow).
- Reading little-endian float16 as f32 has a fast path.
clawhdf5-agent:
- A float16 store writes /memory/embeddings as half precision, and
`MemoryCache::half_precision` rounds each embedding as it enters the
cache (save, update, WAL replay, and on load of a store still f32 on
disk), so memory and file agree bit for bit and a store searches the
same before and after a reopen (tested).
- Values beyond +-65504 are refused with the new
`MemoryError::InvalidEntry` rather than stored as infinity, on every
save path; batches are all or nothing, and a rejected ephemeral entry
stays in the ephemeral tier. Breaking for exhaustive matches.
- CLI: `create --float16`. Off by default.
Measured on tank, 384-dim, six runs alternating order, medians
(search_harness --float16-study --full): at 100K the file goes from
154.0 to 80.8 MiB (-48%), checkpoint 752 -> 512 ms, open 300 -> 252 ms;
vector recall@10 against an exact scan and hybrid_search latency do not
change. At 10K open is 3 ms slower. Also a test that h5py opens a whole
agent store, f32 and float16, and decodes every dataset.
Docs: README, BENCHMARKS.md ("float16 embedding storage"), CHANGELOG
(including the h5py interop fixes in the previous commit), CLAUDE.md.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
Two write-side bugs, both present in every release (the first at least
since v2.1.0), made libhdf5 refuse files written by clawhdf5. Our own
reader ignores both fields, and the interop suites only ever wrote f64
from our side, so nothing here caught them.
- Every f32 dataset: "sign bit position out of bounds". The float
datatype encoder hard-coded the sign bit's position (bits 8-15 of the
class bit field) to 63, which is right only for f64. It is now derived
from the type: bit_offset + bit_precision - 1. This covered every
agent store's embeddings, norms and activation weights.
- Every empty dataset: "invalid dataset size, likely file corruption".
It was written with a real address and size 0, which trips libhdf5's
`addr + size <= addr` overflow check. An empty contiguous dataset now
gets the undefined address, as libhdf5 writes it. This covered every
agent store without sessions or a knowledge graph.
Agent stores are rewritten in full at each checkpoint, so they become
readable at their next checkpoint on a fixed build; other files with f32
or empty datasets need rewriting. Both are recorded in
docs/known-issues.md.
Tests: the sign position byte for f32/f64, and h5py reading our f32
datasets (plain and chunked + deflate) bit for bit.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
rust-version = "1.92" in [workspace.package], inherited by every crate.
1.92 is the floor: wgpu (clawhdf5-gpu) requires it, and the whole
workspace, Python bindings included, checks cleanly on it. ci-test.sh
reads the version from Cargo.toml and checks the workspace on exactly
that toolchain, so the manifests and the README badge cannot drift from
what actually builds. The badge said 1.75, below edition 2024's own
floor.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
The core crates (clawhdf5, -agent, -format, -io, -filters, -ann, -accel,
-netcdf4, -cli) now build no C by default: deflate defaults to zlib-rs,
a pure-Rust port of zlib-ng, and zlib-ng becomes the opt-in
`fast-deflate`, which overrides zlib-rs wherever it is enabled. A default
build no longer needs cmake or a C compiler.
Measured on tank, both builds run alternately, three rounds, medians:
zlib-rs is within 6% of zlib-ng on every HDF5 read and write (512x512
deflate-6 chunked write 1.458 vs 1.484 ms; 64 MB compressed read 64.4
vs 65.2 ms), and compressed output is byte-identical. Details in
BENCHMARKS.md, "Deflate backend".
Getting there took two fixes the first measurement exposed:
- zlib-rs needs `std` to detect SIMD at runtime. flate2 enables it via
its default `runtime_detection`, which `default-features = false` had
switched off, leaving zlib-rs 3.5x slower on inflate. The `zlib-rs`
features now enable it.
- Both deflate paths streamed through flate2's 32 KiB read/write
wrappers. They now hand the codec the whole chunk in one call, into a
buffer sized up front (~5% on chunked writes). This also fixes a
silent short read: the streaming reader returned a truncated stream's
bytes without an error; a truncated chunk is now DecompressionError.
In clawhdf5-filters, output longer than the stated size is now an
error rather than silently cut off.
CI: ci-test.sh lints and tests the zlib-ng path, and fails if a
C-building crate (*-sys, cc, cmake) enters a core crate's default
dependency tree. The arm64 job no longer installs cmake.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
The README had fallen behind v2.3.0-v2.7.0, and parts of it were not
true. Checked every claim against the code and BENCHMARKS.md:
- Three of the six Quick Start snippets no longer compiled (Agent
Memory, Consolidation, OpenClaw); all six now do.
- Hybrid search was described as RRF throughout. The default has been
weighted 0.4/0.6 fusion since v2.5.0; re-ranking and confidence
rejection run only in the OpenClaw backend.
- The `float16` feature does not halve embedding storage (the store
always writes f32), `--features agent` enables nothing, "Source
Isolation" is not wired in, and nothing backs "billion-scale" IVF-PQ.
- "Cryptographically verifiable" overstated an unkeyed, session-scoped
FNV-1a ledger; "Zero C dependencies" was false while zlib-ng was the
default deflate backend.
- Stale numbers: tests (1,650 -> 1,868), Rust badge (1.75 is below
edition 2024's floor), 6.5 KB/record on disk (BENCHMARKS.md: 1.7 KB),
consolidation and hybrid-search latency, and a feature-flag table
broken by a paragraph pasted into it.
- The file schema, module table and crate map now match the code.
Adds a "What's new (v2.2 -> v2.7)" section for collaborators, leading
with the silent Extensible Array read bug fixed in v2.7.0. Footer links
point at git.redclaw.dev. CLAUDE.md: clawhdf5-migrate is the SQLite
migration tool, and MemoryConfig::compression is off by default.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
Two jobs, which runners serve them, and the two constraints that kept
the x86 job failing on every push until today: no JavaScript actions
(`rust:latest` has no `node`, and GitHub is not reachable from every
runner) and `cmake` for libz-ng-sys. Also notes that the Docker Hub
`latest` tag for the runner is frozen at 0.6.1, so it is not a way to
stay current.
Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
Read from the job logs of the first run with the arm64 job, rather than
guessed:
- The x86 `test` job has been failing on every push, in about three
seconds. It runs in `rust:latest` and starts with actions/checkout, a
JavaScript action, and `rust:latest` has no `node`: exit 127 before a
line of code was built. It is replaced with a plain git checkout (the
image has git), and actions/cache, JavaScript for the same reason, is
dropped.
- `test-arm64` got through checkout, toolchain, the aarch64 check and
clippy, then failed building libz-ng-sys — pulled in by
clawhdf5-format's default `fast-deflate` — because the host runner had
no cmake. The x86 image lacks it too, so the x86 job would have hit the
same wall one step later.
Both jobs now install cmake where they can (the Docker job and the x86
container run as root) and say plainly when they cannot (a host runner),
instead of failing inside a build script. vision-01 now has cmake.
Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
The NEON kernels in clawhdf5-accel — `dot_i8` including its SDOT path,
and the f32 NEON kernels that predate it — are cfg'd out on x86, so the
existing job has never compiled, linted or tested a line of them. They
were verified once, by hand, on a Raspberry Pi 5.
`test-arm64` runs on `linux_arm64`, which two runners serve in different
ways: vision-01 executes steps on the host with Rust preinstalled, and
vision-02 executes them in docker.gitea.com/runner-images. The job is
written to work in both: no `container:`, no JavaScript actions (those
are fetched from GitHub, which not every runner reliably reaches), and
an explicit `+stable` toolchain so a host's default — vision-01's is a
January nightly — is neither relied on nor changed. Fetches retry, since
one runner's outbound network was seen failing intermittently.
It lints the accel crate and tests accel, ann and format. It reports
rather than requires the dot-product extension: on a core without it the
plain-NEON kernel is the one that runs, and the tests cover whichever is
present.
Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
quantized_index defaults to true for new stores — smaller and faster at
equal recall on every configuration measured. Existing stores keep
their setting, and stores predating it stay f32, guarded by a real
v2.5.0 store committed as a test fixture.
Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
`MemoryConfig::quantized_index` now defaults to `true`. It holds a
quarter of the index memory and, with the exact re-score, is faster at
equal recall on every configuration measured: 1.63x the queries per
second on x86-64 (AVX2) and 1.18x on a Raspberry Pi 5 (NEON SDOT), with
builds 1.8x and 2.3x faster. The one argument for keeping it off — that
int8 search was slower on ARM — did not survive being measured.
Existing stores do not change. A store written with v2.6.0 or later
keeps its persisted setting. One written before the setting existed has
no stored value, and it loads as `false` rather than as the new default,
so reopening it never changes how its index is held. That case is
guarded by a real store written with the v2.5.0 CLI, committed as
`tests/fixtures/store_v2_5_0.h5` (6.8 KB): the test asserts it reopens
with an f32 index and still searches, and it fails if the load default
is changed to `true`.
The CLI needed more than a new default. `create --quantized-index`
assigned its value straight into the config, so under the new default
every CLI-created store would have been forced back to f32 unless the
caller knew to ask for int8. It is replaced by `--f32-index`, which only
ever switches the default off; `--quantized-index` is still accepted,
hidden, as a no-op, and the two conflict.
The whole agent suite passes under the new default, including the
brute-force recall oracle, now running on int8 plus re-score without
being asked to.
Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
SDOT and plain-NEON kernels for dot_i8, tested bit-exact against scalar
on real ARM. At equal recall the quantised index is 1.18x f32 on a Pi 5
and builds 2.3x faster. Also corrects an unmeasured claim that it was
slower than f32 on ARM.
Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
On a Raspberry Pi 5 at N = 100 000 and equal recall (0.9940 vs 0.9945),
medians of three runs:
f32 33 413 ms build 6 164 QPS
int8 scalar 18 950 ms ~6 190 QPS (what v2.7.0 shipped)
int8 NEON ~17 000 ms 6 640 QPS
int8 SDOT 14 464 ms 7 267 QPS 1.18x f32, 2.3x build
The docs said quantised search stayed off by default because aarch64
"falls back to the scalar loop, where the original trade still
applies" — that it was ~13% slower than f32 there, as on x86. That was
extrapolated rather than measured, and it was wrong: x86's portable
baseline is SSE2 against hand-written AVX2 f32 kernels, but on aarch64
NEON is the baseline and the scalar loop vectorises well, so it already
matched f32. Corrected in BENCHMARKS.md, README.md and CLAUDE.md; the
released v2.7.0 changelog entry is left as it was and the correction is
recorded in a new one.
Labelled as Pi 5 figures throughout — a Pi's memory bandwidth and cache
are far below an M-series or flagship phone, so the ratios will move.
Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
`dot_i8` had an AVX2 kernel and a scalar fallback, so on aarch64 the
quantised HNSW index ran the scalar loop. It now dispatches to one of
two NEON kernels:
- `dot_i8_dotprod`: the ARMv8.2 dot-product instruction, `SDOT`, which
multiplies and accumulates sixteen i8 pairs into four i32 lanes per
instruction. Present on Cortex-A76 and later (Raspberry Pi 5, current
Android phones), Neoverse-N1 (Graviton2, Ampere Altra) and every Apple
Silicon generation. Issued as inline assembly because the `vdotq_s32`
intrinsic is still behind the unstable `stdarch_neon_dotprod` feature;
inline asm is stable on aarch64.
- `dot_i8`: plain NEON for cores without the extension — `vmull_s8`
widens to i16 (even -128 * -128 fits) and `vpadalq_s16` folds adjacent
pairs into i32 accumulators, so nothing overflows.
Selected at runtime with `is_aarch64_feature_detected!("dotprod")`.
Verified on a Raspberry Pi 5 (Cortex-A76, `asimddp` present), not just
compiled — the aarch64 code is cfg'd out on x86, so x86 CI never builds
or lints it:
- both kernels bit-exact against scalar at every length, tails and
extremes included. Each is tested directly rather than through
dispatch, because dispatch only takes one path on a given CPU: on the
Pi, testing through it alone would never have run the plain-NEON
fallback at all.
- mutation-checked: dropping the SDOT kernel's second accumulator fails
at length 32, and using the low half twice in the NEON kernel fails at
length 16 — the first lengths that exercise each.
- the ANN suite passes, including int8 recall against ground truth.
- clippy clean with -D warnings on aarch64.
Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
osobh