About

What InferWeave is

InferWeave is a platform for deploying and serving inference workloads across a shared cluster. The public dashboard is its read-only, pseudonymous view: aggregate capacity and health, model and deployment identities, and benchmark metadata.

What the public dashboard shows and does not show

The dashboard is metadata only: aggregate counts and identifiers, never the text a user wrote or a model produced.

Published fields

• snapshotTotals: aggregate counts only — total/healthy nodes and deployments, active requests
• nodeId: pseudonymous, rotatable node identifier (no private address)
• nodeStatus: health classification (healthy / stale) from metadata telemetry
• nodeCapabilities: capability tags such as gpu model and cpu size
• modelId: model repository identifier
• deploymentFingerprint: model repository@revision, KV precision, max context tokens and serving state
• benchmarkRecords: numeric benchmark results (ttft, decode throughput) with measurement source

Deliberately withheld

• promptText: request text a user wrote — never published
• responseText: text a model produced — never published
• credentials: API keys, tokens or other secrets
• userIdentity: user emails or account details
• privateNodeAddress: stable private node addresses or wallet mappings
• locationData: home or geolocation data
• perRequestDetail: per-request or per-user telemetry beyond aggregate counts