SMSD Pro 8.1.0

SMSD Pro 8.1.0 adds Python data and machine-learning adapters, local MCP tools, packed fingerprint corpora, corpus threshold searches and SMARTS export of MCS results, and corrects the C++ RDKit adapter. The deprecated 8.x APIs remain available and supported throughout the 8.x line. The Java library and command-line tool stay at 8.0.3 on Maven Central. The CPU Docker images carry the 8.1.0 command-line tool; otherwise the Java changes below are available only from a source build. The changelog lists every change.

Changes

• Python adapters give fingerprints as NumPy, PyTorch, SciPy CSR, Arrow, pandas and Polars data, add scikit-learn fingerprint and similarity transformers, and build PyTorch Geometric graphs with native atom numbering and no coordinates or stereo. The PyTorch adapters refuse a conflicting bundled OpenMP runtime; use a shared runtime or build SMSD with SMSD_BUILD_OPENMP=OFF.
• smsd-mcp serves eight local molecular tools over MCP for MCS, substructure and SMARTS search, fingerprints, similarity, screening, standardisation and depiction; the registry also exports function definitions and optional LangChain and LlamaIndex tools. Results are JSON with search status and limits; no tool contacts a model provider.
• A consolidated API reference, an llms.txt index and a guide to connecting AI assistants make the interfaces and local tools easier to find. New Python examples cover NVIDIA screening checked against the CPU, an offline chain of the molecular tools and an MCP client.
• The MCP container recipe and Hugging Face demonstration source use supplied wheelhouses. They are source examples, not deployed services or published images. Native SVG helpers highlight existing MCS results.
• Packed fingerprint corpora are read-only memory-mapped files for threshold and exact top-k scans, Butina clustering and MaxMin selection. The CPU is the default; backend="cuda" computes exact integer intersections on the GPU, and "auto" does so when a supported GPU is available. Clustering decisions stay on the CPU, and cuda_calls counts completed CUDA calls. All-pairs work is quadratic.
• smsd-cuda adds smsd_cuda.preload(), retained device workspaces for prepared corpora, opt-in CUDA Graph replay (SMSD_CUDA_GRAPHS=1) and on-device top-k selection.
• stream_screen_smiles and stream_screen_many_smiles screen target iterables in bounded chunks, and warm_start prepares a corpus and returns a first screen, with the usual TargetCorpus GPU dispatch and CPU fallback.
• Substructure.match_corpus and MCS.match_corpus (C++ matchCorpus) decide, in target order, whether each target contains the query or shares at least a given number of atoms; a stopped search leaves the answer open. Matching runs on the CPU; gpu_prefilter=True lets an available GPU pre-screen targets.
• SMARTS export writes MCS matches in C++, Java and Python; matchWithSmarts (Python match_with_smarts) returns the result and SMARTS together, and Java SmartsWriter writes whole graphs. Stereo, tautomer, complete-ring and strict ring-fusion options are rejected. A pattern carries neither the induced MCIS objective nor the search status.
• Relevant cycles can be counted without listing them: computeRelevantCycleFamilies (Python compute_relevant_cycle_families) groups them into families, each with one member and its exact number of members, and two further functions find the family of a ring or list members up to a given limit. The grouping into families can depend on the input atom order.
• The C++ RDKit adapter (fromRDKit, fromSmiles) keeps each atom's hydrogen count and specified double-bond E/Z configuration. Aromatic N–H atoms, as in pyrrole, previously lost their hydrogen.
• SMILES writers and the C++ and Python SMARTS writer reuse ring-closure numbers and raise an error beyond 99 open closures; SMILES writers also write long chains without exhausting the stack. Compiled SMARTS evaluate distinct recursive predicates separately, and bond edits refresh cached rings and topology. Worker errors in TargetCorpus.prewarm, CPU corpus screening, C++ path-fingerprint batchFingerprint and batchFingerprintScreen now reach the caller instead of ending the process.
• Conda recipes build the CPU package and the Linux CUDA package; the conda guide gives the conda, mamba and Pixi install routes.
• NOTICE now carries licence notices for the RDKit MACCS key definitions and pybind11. The smsd-cuda package now ships the CUB and CCCL licences, and its NOTICE points to them.

Upgrading from 8.0.3

SMSD Pro 8.1.0 supports CPython 3.12–3.15. On Python 3.10 and 3.11, pip continues to install 8.0.3.

Ring perception works within one fixed, deterministic work limit, the same in C++, Java and Python and independent of time, thread count and hardware. Very large fused-ring systems, such as synthetic cages, can reach it. Some ring systems that 8.0.3 left incomplete now complete, and their ring-dependent results can change. Where rings stay incomplete, values computed without the missing rings stay exact; where 8.0.3 computed the rest from incomplete rings, searches now report LimitReached and other calls raise an error naming the cause, with the zero-based input index in C++ and Python batch calls. That error is the new RingBasisUnavailable (Java RingBasisUnavailableException), which existing std::runtime_error, RuntimeError and IllegalStateException handlers still catch. A longer timeout does not change this; see Results and budgets.

Complete-ring matching uses all relevant cycles, and the new relevantCyclesTruncated (Python relevant_cycles_truncated) reports when their enumeration is incomplete. A relevant cycle outside the smallest set of smallest rings now counts as a complete ring, so a completeRingsOnly result that 8.0.3 reported as proved, sometimes empty, can now be larger. With completeRingsOnly, Java substructure matching now rejects matches that cover only part of a ring, as C++ and Python do. ringsTruncated keeps its meaning. Per-atom ring counts, and with them strict ring-fusion results, can differ from 8.0.3.

SMARTS R<n> (n > 0) uses the same minimum cycle basis in C++ and Java. If ring perception cannot complete it, an R<n> search that cannot be decided without it returns LimitReached without matches, which does not prove absence; other functions that need the basis raise an error, and computeSSSR and computeRings may return incomplete rings with truncation flags. R and R0 are unaffected. Ring-size and double-bond stereo fingerprint terms, aromatic-ring keys (as in MACCS) and SMARTS keys whose search needs the basis also raise an error then. Java now uses the same basis as C++ and Python for R<n> and these terms, so a Java result from a different minimum basis of equal size can change.

SMARTS parsing in C++ and Python now rejects a pattern ending in a bond, such as CC-, as Java already did; 8.0.3 C++ and Python accepted it after dropping the bond.

C++ and Python MCIS search (the default MCS objective) has changed: when 8.0.3 or 8.1.0 stops a search at its node or time limit, including the default time limit, the two results can differ, and a result of the same size can come with a different atom mapping. Java does not have this change, so a stopped search can return a different result in Java than in C++ and Python.

MCS.consensus always reports optimal as false; complete means the computation finished, not that no larger common graph exists. Searches on graphs from the C++ RDKit adapter can now differ. MCS searches reject graphs with inconsistent bond properties.

Downloads

SMSD Pro 8.1.0 is available from PyPI and GitHub. Install it with python -m pip install "smsd==8.1.0", adding "smsd-cuda==8.1.0" (matching smsd) for GPU screening on Linux x86_64 with an NVIDIA GPU. Python wheels cover CPython 3.12–3.15 on Linux x86_64 and aarch64, Windows x86_64 and macOS arm64 (macOS 14 or later). For Java, use 8.0.3: the library com.bioinceptionlabs:smsd:8.0.3 and the command-line tool com.bioinceptionlabs:smsd-cli:8.0.3 on Maven Central, and the CLI files and installers of the 8.0.3 release. The CPU Docker images below carry the 8.1.0 command-line tool. Multi-GPU screening and native Windows CUDA wheels are not provided. See the installation guide for requirements, its Docker section for the Docker images and the GitHub releases page for the files and SHA256SUMS:

File	Contents
smsd-8.1.0.tar.gz	Python source distribution
smsd-8.1.0-cp312-cp312-*.whl to smsd-8.1.0-cp315-cp315-*.whl	Python CPU wheels, one per CPython version on each platform: Linux x86_64 (manylinux_2_27_x86_64.manylinux_2_28_x86_64), Linux aarch64 (manylinux_2_26_aarch64.manylinux_2_28_aarch64), macOS arm64 (macosx_14_0_arm64) and Windows x86_64 (win_amd64)
smsd_cuda-8.1.0-py3-none-manylinux_2_34_x86_64.whl	Optional CUDA package for Linux x86_64
smsd-cpp-8.1.0-headers.tar.gz	Header-only C++17 library
smsd-8.1.0-cpu-linux-amd64.tar.xz and smsd-8.1.0-cpu-linux-arm64.tar.xz	Docker CPU images for Linux amd64 and arm64 with smsd and the 8.1.0 Java command-line tool
smsd-8.1.0-cuda-linux-amd64.tar.xz	Docker GPU image for Linux amd64 with smsd-cuda and the CUDA runtime
SOURCE_MANIFEST.json, WHEEL_MANIFEST.json and CUDA_WHEEL_MANIFEST.json	Source commit and checksums of the source distribution and wheels
RELEASE_NOTES.md	These release notes
SHA256SUMS	SHA-256 checksums of the release files

SMSD Pro 8.0.3

SMSD Pro 8.0.3 corrects complete-ring MCS matching, the status of MCS results that depend on truncated ring perception, and the built-in named SMARTS predicates. It adds opt-in Metal fingerprint screening for C++ source builds on macOS and a Python call that fingerprints many molecules at once. The changelog lists every change.

Changes

• C++ and Java MCIS searches with stereochemistry find the largest mapping that keeps the configuration of mapped stereocentres and double bonds.
• C++ MCIS searches return the maximum mapping whatever the atom and bond order of the target.
• Complete-ring MCS matching recognises rings in graphs built without ring flags.
• With complete rings or strict ring fusion, MCS keeps its mappings but no longer reports them as proved when it depends on truncated ring perception.
• GPU similarity-bound screening returns the same hits as the CPU screen, including scores at the threshold and molecules with very large atom counts. The float-valued Metal similarity API is unchanged.
• Metal fingerprint screening for C++ source builds on macOS returns the same hits, in the same order, as the CPU screen. It runs only when requested explicitly. The C++ guide gives its requirements and CPU fallback. The Python wheels use the CPU for screening on macOS.
• The built-in named SMARTS predicates have corrected definitions, shared by C++ and Java and listed in the Java guide. isKetone and isEster now mark the carbonyl carbon.
• Python fingerprints computes the fingerprints of many molecules or SMILES strings in one call, each identical to fingerprint and in input order.

Upgrading from 8.0.2

Patterns that use the built-in named predicates can match different atoms. Java expands names made of ASCII letters, digits and underscores, including inside recursive SMARTS; a registered name containing other characters is no longer expanded.

With complete rings or strict ring fusion, a search that depends on truncated ring perception now reports LimitReached, with optimal and complete false, instead of Completed. isMappingMaximal (Python is_mapping_maximal) then raises its proof-unavailable error rather than returning a value.

In C++ and Python, a single-result connected MCIS search without stereochemistry or ring options can return a different mapping of the same size when every atom of the target is bonded to every other.

Downloads

The installation guide lists the supported platforms and runtime requirements. Python CPU wheels target CPython 3.10–3.14 on macOS arm64, Linux x86_64 and aarch64, and Windows x86_64. The optional smsd-cuda package targets Linux x86_64 and must match the smsd version. The Java CLI is supplied as a portable distribution and as Windows x86_64 MSI, macOS arm64 DMG and Linux amd64 DEB installers that include Java 25. CPU Docker image archives target Linux amd64 and arm64; the GPU Docker image archive targets Linux amd64 and needs a compatible NVIDIA GPU and driver.

See the GitHub releases page for the files and SHA256SUMS. Install with python -m pip install "smsd==8.0.3", and add "smsd-cuda==8.0.3" for GPU screening with an NVIDIA GPU.

SMSD Pro 8.0.2

SMSD Pro 8.0.2 adds the optional smsd-cuda package and ordered batches of corpus screening queries. Matching runs on the CPU. The changelog lists the changes in full.

Highlights

• TargetCorpus.screen_many screens ECFP4 Tanimoto similarity for each query; screen_targets_many screens the atom-count MCS similarity bound. Each result keeps query and target order.
• The optional CUDA backend keeps corpus fingerprints on the device for reuse. Scores use the CPU expressions and hits keep target order.
• CPU wheels cover CPython 3.10–3.14 on four platforms. CPU Docker images cover Linux amd64 and arm64; the separate GPU image targets Linux amd64.

Downloads and installation

See the supported-platforms table for the Python, Java and C++ requirements, the Docker section for the Docker images and the GitHub releases page for files and SHA256SUMS. Install the Python package with python -m pip install "smsd==8.0.2"; the optional CUDA package must also be version 8.0.2.

The Python guide describes CUDA dispatch and CPU fallback behaviour. Native Windows CUDA wheels and macOS GPU support are not included. Windows GPU use is through the GPU Docker image under WSL2.

Upgrading from 8.0.1

Python now requires version 3.10 or later. Existing scalar screening calls and prepared search calls keep their signatures; the two corpus batch screens are additional methods.

SMSD Pro 8.0.1

SMSD Pro 8.0.1 corrects parsing, aromaticity, CIP descriptors and stereochemistry, gives fingerprint quality one definition in C++, Java and Python, and screens targets on every core in Python. The prepared-query API is unchanged. The changelog lists the changes in full.

Highlights

• SMILES, SMARTS and V3000 MOL input outside the supported characters or ranges is rejected with the same error on every platform, and SMILES atoms joined by : ring bonds are aromatic, as RDKit reads them.
• CIP R/S and E/Z labels that only the later sequence rules decide, and C++ SMARTS chirality, ring closures and MCES stereochemistry that match Java.
• Python screen_targets takes num_threads and a prepared TargetCorpus, and rings_truncated reports when a bounded ring search is cut short.

Downloads

File	Requirements
C++ headers: smsd-cpp-8.0.1-headers.tar.gz	A C++17 compiler
Java library: smsd-8.0.1.jar, with smsd-8.0.1-sources.jar and smsd-8.0.1-javadoc.jar	Java 8 or later and CDK 2.13
Command-line tool: smsd-cli-8.0.1-jar-with-dependencies.jar and smsd-8.0.1-cli.tar.gz	Java 8 or later
Python source distribution: smsd-8.0.1.tar.gz	Python 3.9 or later, a C++17 compiler and CMake 3.18 or later
Python wheels for macOS arm64, Linux x86_64 and Windows x86_64	CPython 3.14, as for 8.0.0
SHA256SUMS	Lists the SHA-256 checksum of each file

Upgrading from 8.0.0

• expected_collision_rate (Python analyze_fp_quality, Java analyzeFpQuality) is now the estimated share of hashed features that fell on a bit already set, and Java suggestFingerprintSize returns larger sizes; review thresholds based on either.
• In C++ and Python, rings written aromatic that are not, such as cyclobutadiene, are no longer aromatic.
• A long fragment MCES search for one best mapping can return a different mapping of the same size.

SMSD Pro 8.0.0

SMSD Pro 8.0.0 provides substructure and maximum common substructure (MCS) search for C++, Java and Python through prepared MCS and Substructure queries. A query is prepared once and searches one or more targets; each result holds its matches with their atom and bond mappings, the search status and whether the optimum was proved. The changelog lists the changes in full.

Highlights

• One search API in three languages: connected and disconnected maximum common induced subgraphs (MCIS), maximum common edge subgraphs (MCES), atom-weighted MCIS and substructure search, through match, matchAll and matchMany (Python match, match_all and match_many).
• Constrained, consensus and R-group searches: MCS.matchConstrained, MCS.consensus and Substructure.decomposeRGroups (Python match_constrained, consensus and decompose_r_groups).
• Results that report status, proved optimality and completeness separately from the matches, under timeouts, node limits and cancellation.
• Daylight SMARTS semantics in C++ and Java through Substructure.smarts, documented SMARTS size limits in every language, and the OpenSMILES limits for SMILES bracket atoms in C++ and Python.
• The same valence and tetrahedral-centre rules in C++, Java and Python, and CIP descriptors for R/S that follow the IUPAC 2013 Sequence Rules 1 to 5, with Rule 6 as proposed by Hanson et al. (2018).
• Tautomer classes that do not depend on the atom order, pH-dependent tautomer weights and a tautomer confidence for the selected bonds of a match.
• FingerprintOptions, one fingerprint specification for circular, path, torsion and structural-key fingerprints, including the 166 MACCS keys, with identical bits in C++, Java and Python.
• The Java library com.bioinceptionlabs:smsd, which depends only on the CDK modules it uses, and the separate command-line tool com.bioinceptionlabs:smsd-cli.

Downloads

File	Requirements
C++ headers: smsd-cpp-8.0.0-headers.tar.gz	A C++17 compiler
Java library: smsd-8.0.0.jar, with smsd-8.0.0-sources.jar and smsd-8.0.0-javadoc.jar	Java 8 or later (Java 25 LTS preferred) and CDK 2.13
Self-contained CLI: smsd-cli-8.0.0-jar-with-dependencies.jar	Java 8 or later
Portable CLI with Unix and Windows launchers: smsd-8.0.0-cli.tar.gz	Java 8 or later
Python source distribution: smsd-8.0.0.tar.gz	Python 3.9 or later, a C++17 compiler and CMake 3.18 or later
Python wheel for macOS arm64	macOS 26 or later and CPython 3.14
Python wheels for Linux x86_64 and Windows x86_64	glibc 2.28 or later, or Windows 10 or later, and CPython 3.14
SHA256SUMS	Lists the SHA-256 checksum of each file

The 8.0.0 wheels run batch operations on the CPU with OpenMP. The current installation guide explains how to verify the downloads and install each package, and the validation record lists the release checks.

Upgrading from 7.x

The 7.x search functions are removed. Prepare a query with find (or Substructure.smarts for SMARTS), pass search options with withOptions (Python with_options), whose chemistry field holds the chemistry settings, then search with match, matchAll or matchMany. Java matchMany takes a List<MolGraph>; new MolGraph(container) converts a CDK molecule.

7.x	8.0.0
Python smsd.find_mcs(a, b), smsd.mcs_result(a, b), smsd.mcs_rdkit(a, b)	smsd.MCS.find(a).match(b), which accepts SMILES, MolGraph objects and RDKit molecules
Python smsd.find_substructure(q, t), smsd.is_substructure(q, t)	smsd.Substructure.find(q).match(t), then .found or .best
Python smsd.batch_mcs(q, targets), smsd.batch_substructure(q, targets)	smsd.MCS.find(q).match_many(targets), smsd.Substructure.find(q).match_many(targets)
Python smsd.smarts_match(smarts, t), smsd.smarts_find_all(smarts, t), smsd.compile_smarts(smarts)	smsd.Substructure.smarts(smarts), then .match(t).found or .match_all(t, limit)
Java SearchEngine.findMCS(a, b), new SMSD(q, t, chem).findMCS()	MCS.find(a).match(b)
Java SearchEngine.findSubstructure(q, t), SearchEngine.isSubstructure(q, t, chem, ms), new SMSD(q, t, chem).isSubstructure()	Substructure.find(q).match(t), then found()
Java SearchEngine.batchSubstructure(q, targets, chem, ms), SearchEngine.findAllMCS, SearchEngine.findAllSubstructures	matchMany(targets, numThreads) or matchAll(target, limit) on the prepared query
Java SearchEngine.findMCSSmarts(smarts, target, chem, ms), new SMSD(smarts, target, chem)	Substructure.smarts(pattern).match(target)
C++ smsd::findMCS(a, b, ...), smsd::isSubstructure(q, t, ...), smsd::findSubstructure(q, t, ...)	smsd::MCS::find(a).match(b), smsd::Substructure::find(q).match(t)
C++ smsd::batch::batchMCS, smsd::batch::batchSubstructure, smsd::findAllMCS, smsd::findAllSubstructures	matchMany(targets, numThreads) or matchAll(target, limit) on the prepared query
Python timeout_ms= keyword arguments, Java SMSD.setMCSTimeoutMs(ms) and setSubstructureTimeoutMs(ms)	withTimeout(ms) (Python with_timeout(ms)) on the prepared query
Java artefact com.bioinceptionlabs:smsd with the CLI	com.bioinceptionlabs:smsd for the library and com.bioinceptionlabs:smsd-cli for the command-line tool

• Status and optimality: each result reports status (Completed, TimedOut, Cancelled or LimitReached; in Python SearchStatus.COMPLETED, TIMED_OUT, CANCELLED or LIMIT_REACHED), optimal and complete. A stopped search may return a valid mapping without proving the optimum, and a capped matchAll (Python match_all) may prove it without listing every match, so check optimal and complete separately.
• Java library and CLI: the library no longer brings picocli, jackson-databind or CDK modules such as cdk-smiles, cdk-data and the CDK file readers; projects that used them through smsd declare them. The self-contained CLI is smsd-cli-8.0.0-jar-with-dependencies.jar.
• SMARTS: in C++, as in Java, uppercase bracket atoms such as [C] match aliphatic atoms only and lowercase ones aromatic atoms only; use [#6] or [C,c] for both. C++ reads [H], [2H] and [H+] as hydrogen atoms, and SMARTS beyond the size limits is rejected in every language.
• Deprecated names keep working in 8.0.0, and Python names warn with DeprecationWarning. The warning, or the Java and C++ API documentation of the name, gives its replacement where one exists.