Data, coverage and licensing¶
bharat-judgements reads official court websites and public judgment datasets. It does not host a complete court database. Coverage depends on the source, lookup and records actually published.
Live sources¶
| Source | Client | Available workflows |
|---|---|---|
| High Court Services | HCServicesClient | Basic searches, CNR case details, advocates, orders and cause lists |
| District eCourts | DistrictCourtClient | Location discovery, searches, CNR case details, orders and listing entries |
| Judgment Search | JudgmentSearchClient | Text search, pages and session-aware PDFs for HC or SCR |
| Supreme Court | SCIClient | Recent homepage feed and associated PDFs |
| Calcutta High Court | CalcuttaHCClient | Case orders through the court's own website |
Live requests retrieve what the selected endpoint exposes at query time. They cannot guarantee that the court has published every record or updated every field. The static High Court registry is not a guarantee that every query will work across every court or bench.
Number and party searches provide basic CaseInfo. CNR lookups can provide fuller CaseDetail records. The SCI client does not implement year, party or case-number search on its current website. Use the documented alternative source for those research tasks.
Public judgment archive¶
| Dataset | Scope listed by publisher | Update frequency | Licence |
|---|---|---|---|
| Indian Supreme Court Judgments | Supreme Court judgments beginning in 1950 | Bi-monthly | CC-BY-4.0 |
| Indian High Court Judgments | Judgments from 25 High Courts | Quarterly | CC-BY-4.0 |
The AWS registry lists Dattam Labs as the manager of these public datasets. They are derived from court records; the archive is a publisher-maintained dataset rather than a live court endpoint. Both public buckets are in ap-south-1 and can be read without an AWS account.
The SDK does not include a district court judgment archive. High Court archive party matching uses titles, and citation filters are ignored for HC records. See archive filters.
Freshness has two layers¶
Dataset publication is periodic; no fixed maximum lag is guaranteed. The SDK also caches downloaded metadata with a default 30-day freshness interval. Refreshing that cache cannot produce records the dataset publisher has not added.
For a recent judgment absent from the archive, choose an appropriate live lookup explicitly. A failed request and a successful query with no matches are different outcomes. Keep the query date, identifiers and source with any research output.
Software and data are separate¶
The repository's software uses the MIT licence. The archive publishers list CC-BY-4.0, which requires appropriate attribution when sharing the licensed material. The software licence does not assign rights over all documents retrieved from live websites.
For archive material, retain the dataset name, publisher, registry link, access date and licence. Identify any transformations you made. Follow the publisher's citation instructions and keep document provenance alongside derived outputs.
Data flow and cost¶
Requests go from the execution environment to the chosen portal or archive host. Archive metadata and PDFs can be cached on that machine. Prompts, records or PDFs shared with an assistant are additionally handled by its provider.
There is no bharat-judgements query billing service. Assistant plans, infrastructure, storage, network usage and optional external services can still have costs. Public access does not imply unlimited portal capacity or guaranteed uptime.
Next: request architecture, archive retrieval, or contributing.