proof-server:proof-server-architecture

What problem does it solve?

Provides a focused, practical reference for operating and reasoning about the midnight-proof-server so engineers can reliably generate zero-knowledge proofs, manage concurrency, and handle cryptographic key material without surprises. It reduces time spent debugging capacity, timeout, serialization, and versioning issues by explaining the server's components, job lifecycle, and runtime behaviors.

Core Features & Use Cases

• Worker pool & job queue: Channel-based async job queue with configurable parallelism and queue capacity to control concurrency and backpressure.
• Capacity limiting & backpressure: HTTP 429 responses for full queues and TTL-based garbage collection to prevent unbounded memory growth.
• Key material management: Startup pre-fetching, on-demand proving key fetches, and use of MidnightDataProvider for circuit parameters.
• ZKIR versioning & dispatch: Versioned IR routing so V2/V3 proofs are dispatched to appropriate handlers and reported via /proof-versions.
• Binary serialization & deployment: Custom tagged binary format for proofs, musl static builds, and multi-arch Docker images for portable production deployments.
• Use Case: Tune worker counts, queue capacity, and job timeouts for a production environment; diagnose 429s, long-tail proving times, or missing keys during first requests.

Quick Start

Ask the proof-server skill to explain the worker pool design and recommend --num-workers, --job-capacity, and --job-timeout settings for a production 16-core Linux host.