docs/china-corporate-disclosures.mdx
The China corporate-disclosure lane monitors a small reviewed issuer basket against official exchange metadata. It is not a general-purpose exchange scraper, a full-market announcement archive, or a document redistribution service.
| Exchange | Launch status | Admission decision | Collection contract |
|---|---|---|---|
SSE | launched | admitted_metadata_only | Public announcement metadata, bounded first page, document links retained but bodies not fetched |
SZSE | launched | admitted_metadata_only | Public announcement metadata, bounded first page, document links retained but bodies not fetched |
HKEX | blocked | rejected | No automated collection; TERMS_PROHIBIT_AUTOMATED_ACCESS |
The reviewed terms records are:
SSE and SZSE are metadata-only because commercial reuse of document bodies can
require written permission. documentRetrieval is lazy-link-only: consumers
receive the official link but WorldMonitor does not ingest or redistribute the
linked body. HKEX terms prohibit programmatic systematic retrieval without
permission, so the source remains blocked even when its website is reachable.
| Exchange | Code | Symbol | Issuer |
|---|---|---|---|
| SSE | 600519 | 600519.SS | Kweichow Moutai |
| SSE | 601318 | 601318.SS | Ping An Insurance |
| SSE | 600900 | 600900.SS | China Yangtze Power |
| SSE | 688981 | 688981.SS | SMIC |
| SZSE | 300750 | 300750.SZ | CATL |
| HKEX, blocked | 0700 | 0700.HK | Tencent |
| HKEX, blocked | 1211 | 1211.HK | BYD |
| HKEX, blocked | 0939 | 0939.HK | China Construction Bank |
| HKEX, blocked | 0857 | 0857.HK | PetroChina |
HKEX issuers remain in the reviewed registry so the source decision is visible; they do not cause HKEX network requests.
Only reviewed title patterns become typed events:
Each event retains official exchange identity, issuer mapping, original title, publication time, metadata URL, document link, extraction/classification confidence, content hash, transport/content freshness, and a stable lineage. Corrections, revisions, withdrawals, and cancellations attach to that lineage instead of appearing as unrelated new events. Unclassified revisions are kept in a bounded diagnostic collection rather than promoted to decision signals.
SSE requests at most 100 rows from one reviewed 90-day page. SZSE requests at
most 50 rows from one reviewed page. Neither adapter silently follows
additional pages. A saturated page reports PAGE_LIMIT_REACHED; three
consecutive empty results report COVERAGE_GAP.
The combined snapshot retains at most 100 classified events, at most 20
revisions per event, and at most 100 unclassified revisions. The Railway
market-backup member runs every 30 minutes and publishes
market:china:corporate-disclosures:v1.
SSE and SZSE first try direct Railway egress. At process start, each source selects the first non-empty proxy setting in this precedence order:
SSE: SSE_PROXY_URL → SZSE_PROXY_URL → PROXY_URL
SZSE: SZSE_PROXY_URL → PROXY_URL
This is configuration precedence, not sequential URL failover. After a proxy URL is selected, a failed attempt does not advance to the next environment variable. Distinct SZSE proxy attempts rotate eligible Decodo sticky gateway ports within the selected proxy configuration.
SSE uses up to four direct and four proxy requests with bounded concurrency. SZSE uses one direct attempt and up to two proxy attempts on distinct eligible gateway ports.
The fallback order applies only to transport failures. HTTP or schema results that prove an application-level response are recorded and do not trigger unbounded route guessing. SZSE requires two consecutive successful runs to clear its durable transport-failure state.
SSE_PROXY_URL is an optional SSE-specific override. SZSE_PROXY_URL is an
optional shared China exit. PROXY_URL is selected only when the corresponding
source-specific setting is absent and is also required by other market-bundle
members.
The snapshot can be:
healthy when every launched exchange has usable reviewed coverage;degraded when SSE or SZSE is missing, stale, saturated, repeatedly empty,
or in transport error while another launched exchange remains usable; orA fresh seed heartbeat does not override a degraded embedded source state. Failed refreshes preserve valid last-good events with their original publication times and current transport failure. The blocked HKEX decision is reported separately and does not masquerade as a runtime outage.
The public China decision-signal surface includes at most four valid classified events. Proxy credentials, raw per-attempt source diagnostics, and unclassified revisions are operator-only.