Molecular Identifier Resolver

Resolve molecular identifiers across official PubChem, ChEMBL, and ChEBI sources while keeping stereochemistry, charge, isotopes, components, salts, ambiguity, and provider disagreements visible.

This is an identifier-evidence tool, not medicinal, clinical, or drug advice.

What this does not do

• No medicinal chemistry recommendation.
• No drug or clinical advice.
• No target prediction or activity inference.
• No name-only silent identity selection.
• No stripping salts or stereochemistry without explicit separate representation.
• No proprietary database scraping.

Sources and authorization

Version 0.1 queries the official public PubChem PUG REST, ChEMBL Data API, and ChEBI 2 public API endpoints. RDKit 2025.9.6 performs local parsing and reports canonical forms separately; it is never treated as an identity authority. See SOURCE_POLICY.md and the versioned files under source_contracts/.

Input

Supply 1-1,000 strict {type, value} entries. Supported types are pubchemCid, chemblId, chebiId, inchiKey, inchi, smiles, and name. standardization is none or report_only; raw input is always retained. Name searches return up to maxCandidates (1-25). Concurrency is capped at 8.

{
  "identifiers": [{"type": "pubchemCid", "value": "2519"}],
  "sources": ["pubchem", "chembl"],
  "standardization": "report_only",
  "maxCandidates": 5,
  "concurrency": 1
}

Output

One result row is written per input identifier, followed by an uncharged run summary. Each result keeps raw input, local parse output, provider candidates, exact/related relations, evidence locators, source-response hashes, conflicts, comparability notes, and a canonical candidate only when a deterministic full-InChIKey rule is unique.

The live-verified default resolves to this identity (timestamps and response hashes vary):

{
  "status": "exact",
  "canonicalCandidate": {
    "sourceId": "CHEMBL113",
    "preferredName": "CAFFEINE",
    "inchiKey": "RYYVLZVUVIJVGH-UHFFFAOYSA-N"
  },
  "charge": {"event": "identifier-resolution", "chargeable": true}
}

Optional bounded canonical JSON and SDF evidence artifacts are stored in the default key-value store only after the paid event is accepted and before the result row references them. Keys use Apify's portable character set and stay below 256 characters. A rejected charge writes no artifacts; a partial storage failure removes staged records and emits no dangling references. Every retained artifact has a content type, byte length, SHA-256, schema version, and generation rule.

Pricing

The only visible business event is identifier-resolution at $0.01 per identifier resolution. Exact resolutions, evidence-backed ambiguity reports, and definitive not-found results are chargeable. Invalid input, local parse failure, exhausted rate limits, and provider outages that prevent a terminal result are not charged. There is no per-source event. The Actor preflights the platform event budget and never launches more unique identifiers than it can admit; skipped identifiers receive explicit uncharged charge_limit_reached rows. Maximum event cost is the number of unique admitted identifiers multiplied by $0.01; platform start-event charges may apply separately.

Limits and support

The run timeout is 600 seconds. Memory is 512 MB minimum, 2,048 MB default, and 4,096 MB maximum. Responses are rejected while streaming as soon as they exceed 5 MB, before full-body buffering, and JSON depth is bounded to 32; each request has at most three attempts. For help, see SUPPORT.md or contact support@flintglade.com.