Limits and Metrics

The demo exposes its live configuration at:

https://mrhisearch.com/api/demo/limits

Use that endpoint if you need the current deployed values at runtime.

Current defaults

These are the defaults in this repo right now.

Vector limits

Setting	Default	Meaning
`maxVectors`	`10000`	Max vectors per token
`maxDimensions`	`1024`	Max vector length
`maxAddBatch`	`512`	Max vectors per `add-many`
`maxIdBatch`	`1000`	Max ids per batch read/delete
`maxQueryBatch`	`50`	Max queries per `search-many`

Payload limits

Setting	Default	Meaning
`maxMetadataBytes`	`8192`	Max metadata size per vector
`maxFilterBytes`	`8192`	Max filter object size
`maxRequestBodyBytes`	`10485760`	Max request body size

Rate limits

Setting	Default	Meaning
`tokenIssuePerIp.limit`	`24`	Tokens per IP per hour
`requestsPerIp.limit`	`180`	General requests per IP per minute
`searchesPerIp.limit`	`60`	Search requests per IP per minute
`mutationsPerIp.limit`	`60`	Mutation requests per IP per minute

The corresponding windows are:

tokenIssuePerIp.windowMs = 3600000
requestsPerIp.windowMs = 60000
searchesPerIp.windowMs = 60000
mutationsPerIp.windowMs = 60000

Session defaults

Setting	Default	Meaning
`ttlMs`	`21600000`	Session lifetime after inactivity
`sweepCron`	`* * * * *`	Expired session cleanup schedule
`metric`	`cosine`	Default similarity metric
`vectorType`	`float32`	Stored vector type

Session behavior

One token owns one temporary session
The first successful add initializes the session’s dimensions and metric
Later adds and searches must match the same dimensions
clear removes vectors but keeps the session configuration
index/reset removes the whole index so the token can switch dimensions or metric
When the session expires, the token and its index are deleted

Metrics

Every response includes:

Metric	Meaning
`requestLatencyMs`	Total request time

Some routes also include operation-specific fields:

Metric	Meaning
`tokenIssueMs`	Token creation time
`dbOpenMs`	Time to open the index
`statsReadMs`	Time to read MRHI stats
`hasMs`	Time to test one id
`getMs`	Time to read one vector
`getManyMs`	Time to read multiple vectors
`getMetadataMs`	Time to read metadata
`setMetadataMs`	Time to write metadata
`deleteMetadataMs`	Time to remove metadata
`capacityCheckMs`	Time spent checking vector cap
`addMs`	Time to add one vector
`addManyMs`	Time to add a batch
`searchMs`	Time for one vector search
`searchManyMs`	Time for batched vector search
`resetMs`	Time to reset the index
`deleteMs`	Time to delete one vector
`deleteManyMs`	Time to delete many vectors
`clearMs`	Time to clear the index
`resultCount`	Number of hits returned
`queryCount`	Number of query vectors submitted

Not every endpoint returns every metric. See Endpoints for the exact per-route mapping.

Example limits response

{
  "success": true,
  "limits": {
    "vectors": {
      "maxVectors": 10000,
      "maxDimensions": 1024,
      "maxAddBatch": 512,
      "maxIdBatch": 1000,
      "maxQueryBatch": 50
    },
    "payloads": {
      "maxMetadataBytes": 8192,
      "maxFilterBytes": 8192,
      "maxRequestBodyBytes": 10485760
    },
    "rateLimits": {
      "tokenIssuePerIp": { "limit": 24, "windowMs": 3600000 },
      "requestsPerIp": { "limit": 180, "windowMs": 60000 },
      "searchesPerIp": { "limit": 60, "windowMs": 60000 },
      "mutationsPerIp": { "limit": 60, "windowMs": 60000 }
    },
    "session": {
      "ttlMs": 21600000,
      "sweepCron": "* * * * *"
    },
    "defaults": {
      "metric": "cosine",
      "vectorType": "float32"
    }
  }
}