Limits & Constraints
Request Limits
Section titled “Request Limits”| Limit | Value | Notes |
|---|---|---|
| Search results | Max 50 (GraphQL), max 25 (Core tools) | Default 10 |
| Discover results | Max 50 per query | Default 10 |
| Recent results | Max 50 per query | Default 15 |
| Documents query | Max 100 per page | Default 25 |
| Sync runs query | Max 100 per page | Default 20 |
| Batch ingest | Max 50 documents | Per ingestDocuments call |
| Source nested queries | Default 5 | recentDocuments, syncRuns (topAuthors defaults to 20) |
Session & Timeout Limits
Section titled “Session & Timeout Limits”| Limit | Value |
|---|---|
| MCP session TTL (HTTP transport) | 1 hour |
| Worker execution timeout | ~30 seconds |
| WebSocket idle | Hibernates (no timeout; sleeps until next message) |
Pagination
Section titled “Pagination”Documents and sync runs use offset-based pagination with limit and offset parameters.
- No cursor-based pagination is available.
- There is no enforced maximum offset, but very large offsets produce slower queries.
- Use the
hasMorefield in paginated responses to determine if more results exist. - The
totalCountfield returns the total number of matching records.
Data Constraints
Section titled “Data Constraints”| Constraint | Details |
|---|---|
| Document ID format | 16-character hex string |
| External ID | Must be unique per feed (deduplicated by the (feedId, externalId) pair) |
| Content types | text, transcript, highlight, bookmark |
| Tags | JSON string array, no enforced limit on count |
| Preview text | First ~300 words of document content |
| Vector dimensions | 1024 (bge-m3 model) |
Rate Limiting
Section titled “Rate Limiting”Trove does not enforce per-user rate limits at the application level. Cloudflare’s platform DDoS protection applies at the infrastructure level.
For high-volume ingestion, batch documents into groups of 50 (the maximum per request) and allow reasonable delays between requests to avoid overwhelming the ingest pipeline.