aria-json-telemetry 0.2.2

Aria JSON telemetry — closed operator envelopes, hosted command list, work gateway. 560 catalog identities; one transform.
Documentation

aria-json-telemetry

Published crate for Aria JSON telemetry (aria_operator rustc name). 0.2.2.

aria-json-telemetry = "0.2.2"
cargo add aria-json-telemetry
cargo install aria-json-telemetry --bin work

The node is a stateless IPO: workers bring organized JSON, one ingest + optional Φ projects 560 closed verticals, the callback returns working data only. Neo4j is memory. Nothing is stored here.

Deployed crate Version Registry
aria-json-telemetry (work) 0.2.2 crates.io
aria-engine-core 0.2.1 crates.io (already)
aria-engine-backends 0.2.1 crates.io (already)
aria-engine / aria-engine-wasm 0.2.1 crates.io (already)
560 crates/operators/* workspace 0.2.1 publish = false

Engine crates stay at 0.2.1; this cut does not republish Φ. Workspace will not move to 0.3.0 until winning-condition W6.

Call it

# Identify funnel (production): one binary, stdin JSON, working vertical or nothing
printf '{"nodes":[{"id":1,"type":"Person","label":"Ada","notes":"founder"}]}' \
  | work --binary BIN.PEOPLE --steps 0

# PCVC Mode 4 harness lane (stderr empty, bindings echoed, ≤ 64 KiB)
work --harness < request.json          # or auto-detected by schemaVersion
work --dispatch                        # aria-dispatch-v1: 560 binaries + exe sha256
work --serve 0.0.0.0:8080              # /health /commands /dispatch /work /harness

Container (scratch, static MUSL, non-root 65534):

docker build --target work -t aria-work:0.2.2 .           # 2.04 MB image
docker run -i --network=none aria-work:0.2.2 --harness < request.json      # PCVC lane
docker run -i --network=none aria-work:0.2.2 --binary BIN.PEOPLE --steps 0 < payload.json
docker run -p 8080:8080 aria-work:0.2.2                    # hosted shell
docker compose -f compose.telemetry.yaml up --build --scale telemetry=4

Hosted shell under load

--serve is a fixed worker pool (4× cores) over a bounded queue (1024). Past the queue the node answers 503 Retry-After: 1 at once instead of adding latency; sockets carry 10 s read/write deadlines so a stalled client cannot pin a worker; /health /commands /dispatch are serialized once per process. There is no shared mutable state — nothing to lock, nothing to deadlock. Lock: tests/serve_load.rs (32 clients × 8 harness calls → one distinct body, 0 errors; ~950 ops/s debug build on 12 cores).

What it is

Every catalog binary (crates/operators/*) is a distinct crate. It owns a frozen spec.json (unique binary_id, operator, declared node/rel/tag types). It links this library, which:

  1. Runs aria_engine_backends::telemetry::transform — the Aria transformer, Init + Φ + projection. No sixth action. No Trust field.
  2. Projects a closed operator JSON (sheet 09_JSON_OPERATOR_SHAPE).
  3. Nests the full aria-telemetry-query-v1 document under telemetry only when --telemetry is set.

B0: the operator body contains only the types this crate declared. B1: crates do not share mutable calculation state. B7: Aria is never a judge.

A worker is pointed at one crate. See WORKER.md. What remains: PLAN.md. Measured speed / virality / wire shape: OPTIMIZATION.md. Every identity: catalog/INDEX.md. The 25 sealed map mixers: maps/MAPS.md.

Dump (garbage collection + scores):

cargo run -p aria-json-telemetry --example dump -- dump

Regenerate crates from the sheet dump:

python3 crates/aria-operator/generate.py