Kramank Reference Data API (Product A) (0.1.0)

Public liveness check. Does not require an API key and is not metered.

Unlike the GSTIN/PAN endpoints, this is a reference-data lookup backed by an imported dataset (see scripts/import-ifsc.ts), so a malformed input and a well-formed-but-unknown input are distinct failure modes with distinct status codes.

Beta — data is from a community mirror, not the official current data.gov.in export, which is unreachable at present (see docs/PROJECT-PLAN.md decisions log). The mirror claims a 2019 vintage, but its GitHub repo's last commit is 2017-01-31, so the data is no better evidenced than January 2017 — every response's dataAsOf field reports that evidenced date. District names may predate district reorganizations since then — for example, Andhra Pradesh's 2022 district reorganization split/renamed several districts that this dataset still reports under their pre-2022 names. The official source is retried monthly; the importer's SOURCE_COLUMNS mapping is designed so swapping to it needs no schema or endpoint change.

Like /v1/ifsc/{code}, this is a reference-data lookup backed by an imported dataset (see scripts/import-pincode.ts), not a pure format validator — a malformed input and a well-formed-but-unknown input are distinct failure modes with distinct status codes. One pincode maps to many post offices, and district/state are reported per post office rather than once per pincode: a real minority of pincodes span more than one district (rarely, more than one state) across their post offices, so there is no single correct pincode-level district/state to return.

FORMAT validation only; see the top-level description. A malformed GSTIN (bad checksum, unrecognized state code, unrecognized structure, ...) is still a 200 with valid: false — validation failure is a successful answer, not a client error. 400 is reserved for input that isn't a candidate at all: empty, or far beyond any plausible GSTIN length.

FORMAT validation only; see the top-level description. A malformed PAN (wrong length, unrecognized structure, unrecognized entity-type character) is still a 200 with valid: false — validation failure is a successful answer, not a client error. 400 is reserved for input that isn't a candidate at all: empty, or far beyond any plausible PAN length.

Not yet implemented — deliberately deferred past the Week 5 Product A launch. This is the highest-risk dataset in the product (a wrong tax rate matters more than anything else), so the launch ships without it rather than with unverified rates; the hsn_codes code master is imported and the rate parser is built ahead of launch, but rate verification happens progressively afterward — see docs/PROJECT-PLAN.md decisions log (2026-09-29 entries) for the full design, sourcing, and verification plan. Unlike the other four endpoints, this is a reference-data lookup rather than a format validator, so its response shape isn't drawn from packages/validators. Returns every currently-in-force matching rate/exemption entry, not a single rate, since GST rates are often notified with conditions a single number would hide.