Tutorials

1. Open the browser demo.
2. Enter CCO, the SMILES for ethanol, then run the analysis.
3. Compare the structure drawing with the molecular formula and molecular weight.

CheckYou should see a 2D structure and descriptor values for ethanol.

If it does not workIf the structure is blank, check that the input is valid SMILES and try CCO again.

1. Create a virtual environment and install chematic.
2. Import the package and parse CCO.
3. Print the molecule or one descriptor to confirm the call completed.

CheckThe notebook returns a molecule object without sending the structure to a server.

If it does not workActivate the environment before installing, and check the package name if the import is not found.

1. Parse a query molecule and two candidates.
2. Generate the same fingerprint type for each molecule.
3. Calculate Tanimoto similarity and sort the candidates.

CheckThe identical molecule scores 1.0; less similar structures score lower.

If it does not workUse the same fingerprint settings for every molecule before comparing scores.

1. Prepare a file with one SMILES value per row.
2. Parse each row and calculate the selected properties.
3. Write results together with the original input and any parse error.

CheckThe output has one traceable result per input row, including invalid rows.

If it does not workDo not discard the original string; it makes failed rows easy to inspect and retry.

1. Create a small binary crate and add the required features.
2. Parse one known SMILES in main.
3. Return a clear error for invalid input before adding batch behavior.

CheckThe binary builds locally and produces a deterministic result for the same input.

If it does not workStart with only the features you need if compile time or binary size matters.

1. Install the versioned MCP package locally.
2. Register the stdio command in your AI client.
3. Ask the client to parse a test SMILES and inspect the returned structured data.

CheckThe client can call a chemistry tool and receive a structured response without a hosted service.

If it does not workUse the loopback HTTP adapter only when the client cannot start a local process.

1. Open the browser guide or Playground.
2. Paste the aspirin SMILES shown here.
3. Run depiction and inspect the generated SVG structure.

CheckYou should see a 2D aspirin structure that remains crisp when enlarged.

If it does not workUse a valid, complete SMILES string; SVG depiction does not require a backend.

1. Parse a known SMILES into a molecule object.
2. Calculate molecular weight, LogP, and TPSA from that same object.
3. Keep the input SMILES next to the calculated values in your output.

CheckThe three values are returned locally for the same parsed molecule.

If it does not workTreat descriptor values as calculated properties, not experimental measurements or clinical predictions.

1. Open the homepage feature examples.
2. Keep aspirin as the molecule and enter C(=O)O as the SMARTS pattern.
3. Run the search and inspect the highlighted match in the SVG.

CheckThe result reports a match and highlights the matching atoms.

If it does not workSMARTS describes a pattern, not necessarily a complete molecule; start with a small pattern.

1. Open Local Compound Explorer.
2. Load an SDF, CSV, or .smi file from your computer.
3. Filter or sort the local results, then export the result you need.

CheckThe file is handled in the browser without a dedicated analysis backend.

If it does not workKeep a source copy of the file and use a small sample first to confirm the expected SMILES column or SDF fields.

1. Analyze CCO as a valid reference input.
2. Replace it with incomplete C1CC.
3. Confirm that the page shows an error rather than a made-up result.

CheckValid input produces a structure; incomplete syntax produces a visible error.

If it does not workA parse error is about syntax, not a chemical safety or experimental-validity decision.

1. Open the interactive feature examples.
2. Compare aspirin with salicylic acid using ECFP4.
3. Record both inputs, settings, and the displayed score.

CheckThe same input pair and fingerprint setting returns the same local score.

If it does not workA similarity score does not establish activity, safety, or synthetic feasibility.

1. Analyze one complete SMILES in the browser.
2. Read molecular weight, LogP, TPSA, HBD, HBA, and Lipinski.
3. Keep the exact input beside the calculated values.

CheckThe result panel presents the property indicators for the parsed molecule.

If it does not workThese are calculated indicators, not a conclusion about efficacy, safety, or clinical suitability.

1. Read a CSV with a smiles column.
2. Record values or a ValueError for each row.
3. Write one report before filtering valid rows.

CheckEach non-empty input row has an explicit status.

If it does not workbulk.descriptors skips invalid SMILES, so validate before using it for a source-table report.

1. Validate candidate SMILES.
2. Score one query against the local list.
3. Retain inputs, method, and ranked scores.

CheckThe same valid inputs return the same ranking.

If it does not workInvalid entries are skipped by the bulk API and can break raw-row alignment.

1. Choose a SMARTS pattern.
2. Screen the local list in parallel.
3. Save the pattern and matching SMILES together.

CheckMatching strings remain in source order.

If it does not workThis API treats invalid SMILES as non-matches, so validate data when a negative needs an explanation.

1. Create a binary crate.
2. Enable the three required features.
3. Parse CCO and print its basic descriptors.

CheckCCO reports atoms=3, mw=46.07, and tpsa=20.23.

If it does not workUse the smallest documented feature set for the workflow you actually need.

1. Clone the source repository.
2. Send a local JSON-RPC request over stdio.
3. Confirm the structured benzene response.

CheckThe result has valid=true, atoms=6, bonds=6, and mol_weight=78.114.

If it does not workThis is a local stdio process, not a hosted MCP URL.

1. Parse one complete SMILES such as CCO.
2. Print the formula and molecular weight.
3. Keep the input and package version with the output.

CheckCCO returns C2H6O and a molecular weight of about 46.07 g/mol.

If it does not workA formula does not uniquely identify a structure, and molecular weight here is a calculation rather than an experimental measurement.

1. Serve a small HTML page from HTTPS or a local server.
2. Paste the version-pinned module script.
3. Confirm 180.16 in the browser console, then release the molecule object.

CheckAspirin prints 180.16 after the pinned module initializes.

If it does not workDo not use @latest, and do not rely on a file URL for ES-module loading.

1. Run the local MCP process.
2. Send the name_to_smiles request.
3. Retain the source and returned SMILES before local analysis.

CheckAspirin returns a PubChem-backed isomeric SMILES.

If it does not workThis is the only MCP tool that sends input to a network service.

1. Run the local MCP process.
2. Send the canonical_smiles request.
3. Keep the input, canonical output, and version together.

CheckThe response contains a canonical SMILES string without a network request.

If it does not workDo not treat canonical strings from different libraries or releases as interchangeable without validation.

1. Read a CSV with a smiles column.
2. Write canonical SMILES and a row status.
3. Review duplicate candidates before deleting data.

CheckCCO and OCC share C(C)O; the later row is flagged as a duplicate candidate.

If it does not workCanonical strings are not universal identity proof; retain original rows and version information.

1. Keep the source SMILES unchanged.
2. Standardize a separate review value locally.
3. Compare only under the recorded policy.

CheckSodium acetate and acetic acid both yield CC(O)=O for a largest-fragment review.

If it does not workNever replace the submitted salt form or stereochemical record with the review representation.

1. Load source SMILES in row order.
2. Record accepted and rejected results.
3. Map input_index back to the source before the next workflow.

CheckCCO and OCC are accepted; incomplete C1CC remains rejected at input_index 1.

If it does not workSyntax acceptance does not prove chemical identity, salt policy, safety, or experimental validity.

1. Parse one complete SMILES locally.
2. Write the SVG depiction as UTF-8 text.
3. Keep input and package version with the artifact.

Checkaspirin.svg opens as a scalable 2D structure drawing.

If it does not workAn SVG drawing is not chemical identity, stereochemical, safety, or experimental proof.

1. Set a reaction SMILES in reactants>>products form.
2. Write the SVG diagram as UTF-8 text.
3. Keep the reaction SMILES and package version with the artifact.

Checkesterification.svg opens with reactants, an arrow, and products.

If it does not workA reaction SVG is not proof of atom mapping, balance, conditions, yield, identity, safety, or experimental validity.

1. Parse one complete SMILES locally.
2. Set the record name and source SMILES as properties.
3. Write and inspect one SDF record.

Checkethanol.sdf contains the ethanol record, source_smiles: CCO, and a $$$$ terminator.

If it does not workAn SDF record and generated 2D coordinates do not prove chemical identity, material provenance, safety, or experimental validity.

1. Read name and smiles from a CSV.
2. Write valid records with source-row properties.
3. Keep invalid rows in a CSV report.

CheckValid rows create SDF records; C1CC remains as invalid_smiles in the report.

If it does not workSDF output does not prove identity, provenance, safety, or experimental validity.