> ## Documentation Index > Fetch the complete documentation index at: https://docs.onera.app/llms.txt > Use this file to discover all available pages before exploring further. # Coverage > Understand the verified source ranges, identifiers, processing state, and known gaps for US, India, and China datasets. Corpus publishes separate US, India, and China datasets under one API contract. Coverage is stated conservatively: a source range or metadata row does not automatically mean every document, extracted section, or normalized record is ready to serve. China A-share prices are served under `/china`. See [China](#china) below. Dated counts on this page describe a validated source or audit snapshot. Use the discovery endpoints and per-record availability fields for the current serving state. ## Coverage at a glance | Dataset | United States | India | China | | ----------------------------------- | ------------------------------------------------------------------ | ------------------------------------------------------------------- | ----------------------------------------------------------- | | Companies or securities | SEC CIK and ticker | ISIN, NSE symbol, and BSE code | Exchange-qualified A-share code, exchange, and board | | Historical prices | Daily historical EOD panel | Daily historical NSE and BSE panel; governed exports available | Daily unadjusted A-share panel with explicit trading status | | Normalized financial statements | SEC XBRL-era income, balance-sheet, and cash-flow data | Not published | Not published | | As-reported and fundamentals detail | Accession-level statements and an ingested XBRL concept projection | Not published | Not published | | Regulatory or issuer filings | SEC filing metadata, documents, XBRL, and supported sections | Annual reports with PDF and extracted-text availability | Not published | | Ownership and filing events | Form 4, Form 13F, and supported 8-K or 6-K items | Not published as equivalent datasets | Not published | | News | Not published | Publisher-syndicated headlines, links, summaries, and categories | Not published | | Court judgments | Not published | Court metadata, PDF and text availability, court facets, and search | Not published | | Document search | SEC filing text and items | Annual reports and provisioned judgment records | Not published | ## Live serving snapshot The following production counts were measured on **July 23, 2026**. The US summary was last refreshed July 11 and the India summary July 22; source dates can be earlier. | Dataset | Records currently indexed | Identity or source range | | -------------------------- | ------------------------: | ------------------------------------------------------------ | | US daily prices | 33,859,056 bars | 16,311 priced instruments; January 2, 1962-June 16, 2026 | | US fundamentals | 4,373,869 statement rows | 6,774 companies; reported periods May 25, 1967-June 16, 2026 | | India daily prices | 22,234,653 bars | 5,024 ISINs; January 3, 2000-October 24, 2025 | | India annual reports | 117,261 artifacts | 14,641 distinct company-name values; source years 1997-2022 | | India news | 138,083 items | 19 categories; latest syndicated item July 22, 2026 | | India High Court judgments | 19,930,172 metadata rows | 25 High Courts | | India district-court cases | 80,835,983 metadata rows | Anonymized 2010-2018 source panel | The earliest fundamentals period is not evidence that every company has continuous history from that date. Court-row counts describe indexed metadata, not universal judgment-PDF or extracted-text availability. ## United States ### Companies and identifiers US company identity is anchored on the SEC CIK. Tickers are convenient lookup identifiers but can change or be reused; accession numbers identify filings. Exchange, SIC, active status, and similar attributes are returned only when the source or reconciled company record supports them. Corpus does not currently claim complete active-and-delisted ticker parity. Price-universe membership and SEC-filer identity are different concepts and should not be treated as interchangeable. ### SEC filings The authoritative annual-filing denominator is built from SEC quarterly master indexes rather than only the current ticker universe. The separate active **all-form source index** spans **1993 Q1-2026 Q3**, contains **135 quarterly shards**, and retains **27,040,422 source association rows** through **July 24, 2026**. Association rows are not unique filings: EDGAR labels, co-registrants, and issuer relationships are intentionally preserved. The validated annual-filing snapshot through **July 24, 2026** contains: * **286,772** 10-K or 10-K/A accession/CIK associations * **276,686** unique 10-K or 10-K/A accessions * **39,763** associated CIKs The independently validated **10-K and 10-K/A denominator** begins in **1993 Q1** and covers available quarters through the current quarter, matching the lower bound of the all-form source index above. These counts describe annual filing metadata, including co-registrant associations. They do not assert that every accession already has archived bytes, normalized text, parsed sections, XBRL components, or as-reported statements. Each filing exposes `content_status` as one of `pending`, `available`, `not_provided`, `unsupported`, or `permanent_source_error`. A content endpoint may return `503` with `Retry-After` while a pending accession is processed. As of July 23, 2026, the serving table contains **9,406,107 canonical filing rows**: **829,015** are `available`, **8,577,089** are `pending`, and **3** are `unsupported`. Corpus therefore has broad EDGAR metadata coverage, but the all-form document-to-text pipeline is not yet complete end to end. ### Financial statements Normalized statements are derived from SEC XBRL and therefore have stronger coverage in the XBRL era. The live serving snapshot contains 4,373,869 statement rows across 6,774 companies, but Corpus has not established complete legacy, delisted-issuer, operational-KPI, or segment-taxonomy parity. Supported concepts map into `corpus-financials/v1`. Concepts outside that mapping remain available through the `/us/fundamentals/ledger` serving projection. This endpoint is not a lossless accession, context, and unit ledger: the current loader deduplicates its serving grain and can coalesce same-day duplicates. ### Ownership and events Corpus exposes normalized Form 4 transactions, Form 13F holdings, and supported 8-K or 6-K filing items. Historical availability is checkpointed and should be inspected through `/us/edgar-research/coverage` before using a range in a backtest. Accessions still in `needs_review` are excluded from normalized ownership responses. ### Prices The live US price panel contains 33,859,056 daily bars across 16,311 priced instruments from January 2, 1962 through June 16, 2026. It uses available Stooq symbols plus a frozen 2017 Stooq-derived supplement for some symbols that later delisted. This is a split-adjusted research panel, not an official exchange feed, not intraday data, and not a guarantee of complete survivorship-bias-free US coverage. ## India ### Securities India security identity is anchored on ISIN. NSE symbols and BSE codes are reconciled from public exchange security masters where available. A missing mapping remains `null`; it is not inferred from a company name. ### Historical prices The live India price panel contains 22,234,653 daily bars across 5,024 ISINs from January 3, 2000 through October 24, 2025. It is built from licensed historical NSE and BSE source files. These figures do not guarantee that every instrument has a complete daily series or that every account has redistribution rights. The public price contract is daily OHLCV. Corpus does not currently publish a real-time or intraday India price feed. ### Annual reports The registered BSE/NSE Annual Reports Library covers **1997-2022** and has a known source gap: **2017 is absent**. The collection is roughly 400 GB upstream. The live index contains 117,261 artifacts across 14,641 distinct company-name values. All indexed artifacts have a source PDF pointer; 117,220 have extracted character content, leaving 41 without extracted text at the measured cutoff. Individual records report PDF, text, page, OCR, scan, and character-count availability. Do not treat the library year range as proof that every issuer filed a report in every year. ### News India news is an allow-listed RSS and Atom signal layer. The live index contains 138,083 items in 19 categories, with the latest measured publication timestamp on July 22, 2026. Coverage depends on each publisher's feed history and availability. Corpus stores syndicated headline, link, timestamp, category, and available summary fields; it does not provide full scraped article text under this dataset. ### Court judgments The upstream High Court collection spans **1950-2026**, approximately **17.8 million judgments**, and **25 High Courts**. The live serving index contains 19,930,172 High Court metadata rows across 25 courts after source reconciliation and supplemental public datasets. Source-corpus and serving counts differ because they describe different stages and source combinations. The live serving index also contains 80,835,983 rows from an anonymized Development Data Lab district-court panel. That panel contains structured eCourts/NJDG case metadata but no party names and is not a complete national archive. A separate general district-court source remains inactive. Use `/india/judgments/courts` to discover the courts currently present, and use `has_pdf` or `has_text` for per-record artifact availability. ## China ### Securities and identifiers China identity is anchored on the **exchange-qualified code** (`sh.600519`). A bare six-digit code is not unique across mainland exchanges and is rejected rather than resolved by guesswork. The registered universe contains **5,540** instruments across the Shanghai (**2,457**) and Shenzhen (**3,083**) exchanges, of which **337** are delisted and carry a delisting date. By board: **3,485** main, **1,441** ChiNext, **614** STAR. The upstream source returns no Beijing Stock Exchange codes, so **BSE is not covered**. `BSE` remains a valid enum value in the contract but no instrument currently carries it. The 337 delisted names reflect what the source retains; this is not asserted to be a complete record of every A-share delisting since 1990. ### Historical prices Prices are **unadjusted** (不复权). Corpus does not store the 前复权 forward-adjusted convention Chinese feeds default to, because forward adjustment retroactively restates earlier bars and would defeat point-in-time reads. The China panel contains **18,318,169 daily bars across all 5,540 instruments**, spanning **December 19, 1990 – July 31, 2026**. Every registered instrument has bars; the initial backfill is complete. Of those bars, **586,769 are suspended sessions**, retained as rows with null prices and zero volume so a halt is distinguishable from a data gap. Suspended sessions are excluded from `/china/prices` unless `include_suspended=true`. **Zero** suspended rows carry a price — this is enforced by a database CHECK constraint, not only by the loader. Adjustment factors are held separately from the bars: **64,392** factor rows across **5,540** codes. They use a return-based (涨跌幅) method and will not reconcile against a naive cumulative price ratio. The exchange trading calendar covers **8,694 trading days** from December 19, 1990. A bar count is not a claim of continuous history per instrument: a name listed in 2021 has a short series, and delisted names stop at their delisting date. ### Rights China price data is classified `scraped_prohibited` and is internal-only. It is excluded from every resale surface, and `china:read` is a distinct grant that a US or India key does not inherit. See [China API](/china/overview) for suspension, adjustment, and price-limit semantics. ## Discover current availability * `GET /catalog` lists the datasets, methods, and scopes enabled under `/us` and `/india`. * `GET /us/companies` discovers US company records; `GET /india/securities` discovers Indian instrument records. * `GET /us/edgar-research/coverage` reports the safe public completion ledger for Form 4 and Form 13F research data. * `GET /india/news/categories` and `GET /india/judgments/courts` expose current facets instead of relying on a static list. * Filing `content_status`, annual-report `pdf_available` and `text_available`, and judgment `has_pdf` and `has_text` describe per-record readiness. * `meta.as_of` records the point-in-time cutoff applied to a read. ## Current limitations * China covers Shanghai and Shenzhen only; the upstream source returns no Beijing Stock Exchange instruments. * China prices are unadjusted. Corpus does not publish a China total-return or corporate-action-adjusted series. * China publishes prices and securities only. Company fundamentals, filings, and search are not published for China. * India normalized fundamentals, ownership datasets, and SEC-style filing-item models are not published. * US news and court judgments are not published. * Prices are historical daily data, not real-time snapshots or intraday trades and quotes. * Complete legacy and delisted-issuer parity for normalized US financial statements is not yet proven. * EDGAR document, section, XBRL-component, and as-reported backfills continue independently of filing metadata. * Options, implied volatility, analyst estimates, exchange short interest, securities lending, and complete beneficial ownership are outside the current public contract. * Access to a dataset does not itself grant redistribution rights; scopes and source-specific rights decisions still apply. For the origin and transformation path behind each dataset, see [Data Provenance](/data-provenance).