Skip to content

Questions and answers

Use this page to check what bharat-judgements can retrieve, what a result means, and where your data goes.

It is an open-source Python SDK and CLI that reads court portals and public archive datasets. It does not create a complete, editorially curated database or perform legal analysis by itself.

What does it cost?

The software has no bharat-judgements subscription or metered query fee. Your assistant provider, hosting, storage, network access and optional services may have their own charges.

Does it find every case?

No. Results depend on court, source coverage, published metadata, query parameters and portal availability. The archive includes Supreme Court and High Court judgments; it does not include a district judgment dataset. Empty results do not prove that a matter is absent.

Can it show the next hearing?

High Court and district CNR lookups can return CaseDetail, including the next hearing date when the portal publishes it. Number or party searches return CaseInfo and can omit status and hearing dates. Check the actual source page for time-sensitive listing information.

Is the archive current?

It is historical. The dataset publishers list bi-monthly Supreme Court updates and quarterly High Court updates. Actual coverage can lag, and the tool's metadata cache has its own refresh interval. See source notes.

Does it connect automatically to every assistant?

No. The bundle contains instructions, not a server connection. Local Claude Code can load the project skill and execute installed tools when permitted. Desktop, Cowork and hosted environments require their own skill and execution setup. The repository now includes a local Docker MCP server. Hosted environments need a supported connection; no hosted connector is provided.

Are searches private?

The SDK runs wherever you execute Python. Requests go to the court portals or archive hosts. Records and prompts shared with an assistant also go to that provider. Archive files are cached locally in the execution environment. Local execution alone does not make an assistant conversation confidential.

Does automatic CAPTCHA solving always work?

No. OCR can fail and the live clients have finite retries. Without OCR, the default solver falls back to manual input. See the CAPTCHA guide.

Can it download every judgment PDF?

Only where the selected source exposes a working download. Archive PDFs can be fetched through Judgments.fetch_pdf. Live facade results need the original result from JudgmentSearchClient for a session-aware download. Supreme Court archive retrieval may download a whole year's tar file first.

Can I redistribute records?

The software and archive data have different licences. The software uses MIT; the public archive datasets list CC-BY-4.0. Read data and licensing before republishing. A software licence does not grant rights over every portal document.

What should an assistant report when something fails?

The failed operation, source, error and any records it actually retrieved. It should not convert a connection failure into “no results” or invent missing case details.

Next: setup, example prompts, or source coverage.