Open Training Corpora
Unmetered, no-key-required training corpora for developers and researchers. Every dataset is free to query over HTTP, connect an MCP client to, or download in bulk — no registration, no payment, no vendor lock-in.
Catalog
Each dataset is designed for a specific training purpose — coding agents, security evals, protocol grounding. Access them via REST API, MCP server, or bulk download.
Issue-Patch Corpus
PolyPatch-11k
Real-world GitHub issue-and-patch instances paired with test changes that define the task. 11,611 gold instances across 7 languages and 3,886 repositories — repository-disjoint splits, static test attribution, and full provenance. Free API, MCP server, and bulk download available.
Access
Query via REST API, connect an MCP client, or download the full corpus. No API key required for any access method.
REST Endpoints
PolyPatch-11k API
Full-text search, instance sampling by language/repo/CI-annotation, single-instance retrieval, schema contract, and corpus totals. Every endpoint accepts GET or POST with the same filters. Rate limited to 120 requests per minute per IP.
Agent Server
Registered MCP Server
Streamable HTTP MCP server at https://polypatch.lateos.ai/mcp — tools for fetching instances, searching the corpus, retrieving individual records, and getting corpus stats. No auth headers needed.
Bulk Corpus
Full Dataset Dump
One request gets the entire 27.5 MB compressed JSONL corpus — all 11,611 instances at full field length. Prefer this over paging through the API endpoints.
HuggingFace
Public Preview
Deterministic public preview subset on HuggingFace — same schemas, same quality, parquet format. Browse before you download.
| Endpoint | Purpose |
|---|---|
| POST /v1/instances/fetch | Sample instances, filtered by language, repo, ci_annotated, min_tests |
| POST /v1/instances/search | BM25 full-text search across problem statements |
| GET /v1/instances/{id} | Retrieve one instance |
| GET /v1/schema | Field contract, withheld columns, caveats, licence |
| GET /v1/count | Corpus totals and per-language breakdown |
| GET /v1/bulk | Redirect to the full 27.5 MB compressed dump |
| POST /mcp | MCP server — fetch_instances, search_instances, get_instance, get_corpus_stats |
Machine-readable contracts: OpenAPI · plugin manifest · llms.txt · MCP server.json · Croissant metadata
Quickstart
Search the corpus
curl -s -X POST https://polypatch.lateos.ai/v1/instances/search \
-H 'content-type: application/json' \
-d '{"q": "race condition", "language": "go", "limit": 5}'Every endpoint also accepts GET with the same filters as query parameters — e.g. https://polypatch.lateos.ai/v1/instances/search?q=race+condition&language=go&limit=5 — for GET-only fetch tools (browsers, web fetchers, agents).
Want more?
The free datasets above are just the beginning. Explore our complete collection including golden LLM security training datasets with complete legal provenance — SFT corpora, DPO preference pairs, and domain-grounded evaluation sets.
partnerships@lateos.ai · response within 2 business days