Short answer
Legalcode's indexed LV_ALL_COURTS collection contains 384,037 anonymised decisions from Latvijas tiesu portāls. It reaches across Latvia's general, administrative and specialized courts and across trial, appellate and Supreme Court levels. Retrieved examples include civil, criminal and administrative proceedings, ordinary judgments, procedural rulings, default judgments and abbreviated criminal judgments.
The collection is broad, but its boundaries matter. It is not the same as Legalcode's separate selected Supreme Court archive or Constitutional Court collection. The 384,037 figure describes the indexed courts-portal corpus reviewed on 13 September 2026, not a claim that every decision ever issued by every Latvian court is available.
Anonymisation uses placeholders such as /pers. A/, /personas kods/, /konta numurs/, /reģistrācijas numurs/ and /Nosaukums/. Courts, judges, lawyers, public bodies and some legal persons may remain named. The text can therefore explain the dispute and outcome without reliably identifying a private party. The audit also found portal download footers embedded in indexed text; that ingestion defect is tracked in Legalcode issue #590.
The correct source boundary
Production MCP Discovery identifies LV_ALL_COURTS as the Latvijas tiesu portāls collection of anonymised court decisions. A source-scoped Search returned an exact total of 384,037 records. Two other Latvian case-law sources must be kept separate: LV_SUPREME_COURT is a selected Supreme Court jurisprudence archive, while LV_CONSTITUTIONAL_COURT contains Constitutional Court material.
That distinction affects the answer. A Supreme Court decision may appear in the all-courts portal, but it does not make the portal collection identical to the curated Supreme Court archive. Likewise, Constitutional Court cases are not part of the 384,037-record total used here.
Court levels represented
Every LV_ALL_COURTS record has a normalized court-level value. Production Analyze returned:
- 265,273 trial-level decisions;
- 34,815 appellate decisions;
- 17,765 Supreme Court decisions; and
- 66,184 records in the
OTHERlevel bucket.
The OTHER value needs care. Court-name analysis shows many historical or former court labels that are not mapped into the current hierarchy. It is a metadata bucket, not a separate type of Latvian jurisdiction and not proof that the record is non-judicial.
The latest source-scoped results illustrate the span. They include trial judgments from district and city courts, administrative trial decisions, an appellate administrative judgment and decisions from the Senate's civil, criminal and administrative departments.
General, administrative and specialized courts
The court-type field also has complete coverage. Of the indexed decisions, 302,145 are classified as GENERAL, 81,471 as ADMINISTRATIVE, and 421 as SPECIALIZED.
General-court results include both civil and criminal matters. Docket prefixes and fetched text make that distinction visible: C records concern civil proceedings, while K records in the samples concern criminal proceedings. Administrative records include first-instance matters from Administratīvā rajona tiesa and appeals from Administratīvā apgabaltiesa. The specialized sample comes from Ekonomisko lietu tiesa, Latvia's Economic Affairs Court.
The court-type value describes the institution, not necessarily the entire subject matter of the dispute. A general court may decide a commercially significant debt case, and an administrative court may review a public register decision affecting a business.
Judgments, rulings and other document forms
The portal corpus is not limited to one procedural form. Fetched Latvian headings show at least these forms:
SPRIEDUMS— a judgment;LĒMUMS— a ruling or order;AIZMUGURISKS SPRIEDUMS— a default judgment; andSaīsinātais spriedums— an abbreviated judgment.
For example, the begins:
Civillietu departaments — Latvijas Republikas Senāts — SPRIEDUMS.
—
An instead identifies itself as:
Lietas arhīva Nr. AA43-0962-16/11 — LĒMUMS.
—
Legalcode currently normalizes the sampled portal records under the broad COURT_DECISION decision kind. The narrower form must be read from the document heading and, where available, the ECLI. An ECLI ending in .S appears on sampled judgments; .L appears on sampled rulings. Researchers should not infer that every generic COURT_DECISION is a final merits judgment.
What the anonymisation preserves and removes
The decisions retain enough structure to understand the proceeding: court, date, docket or archive number, ECLI where present, procedural history, legal provisions, reasoning and operative result. The text also keeps roles distinct even when names are removed.
In the Senate civil sample, the parties and identifying vehicle data are replaced:
/pers. C/ … /pers. D/ … /pers. E/ … /TL marka/ … /reģistrācijas numurs/.
—
The similarly replaces the defendant, personal code and account number while preserving the offence, sentence, confiscation and appeal information. This makes legal analysis possible but blocks a reliable name-based match to a private debtor.
Anonymisation is selective. Judges and other professional participants may be named. Public authorities and corporate names can remain where disclosure is part of the published judgment. Consequently, “anonymised” does not mean that the document contains no identifiable person or organization; it means that protected party and identifier data have been redacted according to the portal's publication process.
Search and citation identifiers
The most reliable retrieval keys are the exact Legalcode sourceRef, the official portal record ID, case number and ECLI. A Search result for SKC-0080-25, for example, returned exactly one LV_ALL_COURTS record with sourceRef case/LV/lvcourt_db8d9e94e78833f52bdd5d44 and the official PDF above.
Older records may lack an ECLI and use a source-record identity instead. They remain retrievable by court, date, case number and full-text terms, but cross-source matching is less certain. Names hidden behind /pers. A/ are not valid join keys, and a placeholder has no stable identity across unrelated decisions.
MCP source-quality finding
The audit found a distinct normalization defect. Fetched documents contain repeated page footers such as an IP address, a 2026 extraction timestamp and the words Anonīms sistēmas lietotājs. Searches for both the IP and the footer label matched at least 50 indexed decisions; the production response capped the count, so this is only a lower bound.
The footer is portal acquisition material, not judicial reasoning. It can create irrelevant keyword matches, inflate document length and make an extraction date look like a decision date. records exact Search and Fetch reproductions and calls for normalization, corpus audit and reindexing. The underlying decisions remain usable if the footer is disregarded, but quoted passages should exclude it.
Conclusion
Latvijas tiesu portāls supplies a large, cross-court body of anonymised civil, criminal and administrative material. The indexed collection includes trial, appellate and Supreme Court decisions from general, administrative and specialized courts, with judgments, rulings, default judgments and abbreviated judgments among the fetched forms. It preserves legal reasoning and procedural identifiers while replacing many private-party names and direct identifiers. That supports legal research, but not automatic attribution to a particular debtor or company. Researchers should cite the exact sourceRef and official PDF, distinguish the broad COURT_DECISION class from the document's actual form, and disregard the tracked portal-footer contamination.
Sources
This article provides general legal information, not advice for a specific matter. Laws, facts and procedural deadlines may change the result. Follow the linked primary sources and obtain local advice before acting.