Cardano Updates

Plumbs tip_hash through the stats pipeline (telemetry → cluster →
UI) so each node in the topology graph displays its block number
plus hash suffix, making it easy to verify all nodes agree on the
same block.

Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>

Charge strength -300 → -800 so the topology graph doesn't bunch up.

Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>

- Sample cluster: 5 nodes → 25 nodes, degree 3
- RB generation probability: 0.05 → 0.5 (~1 block per 2s network-wide)
- Remove ib_generation_probability (no IBs in this Leios spec)

Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>

Two issues fixed:
- KeepAlive RTT now includes the configured inbound_delay so reported
  latency reflects the simulated link delay (real localhost RTT is ~0)
- Accepted (inbound) peers skip RTT storage since they can't match a
  configured delay; the outbound side of each connection tracks RTT
- LatencyMeasured events are exempt from the delay buffer (they're
  measurements, not data)

Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>

Default latency range changed from 10-200ms to 5-300ms, sample config
from 10-100ms to 5-250ms. This better reflects real intercontinental
Cardano network latencies (local ~5ms, cross-continent up to ~300ms).

Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>

Adds @mui/icons-material and a pause/play toggle to the event log
panel. When paused, polling and auto-scroll stop so events can be
inspected; on resume the timestamp cursor catches up automatically.

Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>

- HttpEventSink sends each event immediately (no batching) — reqwest
  uses persistent connections so overhead is minimal
- Server pushes events to EventWindow on ingestion, bypassing the
  aggregator watermark delay (aggregator still writes ordered JSONL)
- Increase server→aggregator channel capacity 256→4096
- Reduce ordering_window_secs 5→1s, stats_interval_secs 5→1s
- Bump UI MAX_SERIES 60→300 to keep 5 min history at 1s stats interval

Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>

Cluster --set only applies to cluster config. The new --node-set flag
forwards key=value overrides as --set to each spawned net-node process,
e.g. --node-set 'production.rb_generation_probability=0.2'.

Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>

ChainStore block_no was a running append counter used as Tip.block_no in
ChainSync. Different nodes diverged because they saw blocks in different
orders (especially fork blocks from same-slot producers), causing block
propagation to stall when two nodes had the same counter for different
chains.

Fix: consensus passes the actual chain height via InjectBlock instead of
ChainStore maintaining its own counter. Self-produced blocks update
local_tip immediately in register_self_produced rather than waiting for
peer re-announcement.

Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>

Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>

- Fix duplex peer reconnection: outbound peers now reconnect on failure
  (was only reconnecting InitiatorOnly, missing Duplex mode)
- Accept loop creates duplex peers instead of responder-only, so accepted
  connections run both client and server protocols (ChainSync client etc.)
- Make topology edges directional (from→to only) since duplex handles both
  directions — halves connection count per node pair
- Remove coordinator-side append_block on BlockFetched — blocks only enter
  ChainStore after app validation via InjectBlock (was double-counting)
- Add dedup to ChainStore::append_block as safety net
- Fix ResponderOnly address format to include port (SocketAddr not IpAddr)
- Remove responder_task module (replaced by accepted duplex task)
- Move server_protocol_configs to peer_task module

Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>

- Nodes flash green on block production, amber on block receipt (600ms fade)
- Dark blue-grey default node background, thick border on selection
- Show full node ID and tip in topology nodes
- Show resolved node ID in peer list (mapped from topology addresses)
- Larger event log font

Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>

- Full-height right sidebar with inspector (3:2) and event log
- Edge handles centered on nodes (no more dangling connection points)
- Node/edge click always selects (no toggle deselect), pane click deselects
- Fix infinite update loop on node selection (filter onNodesChange to dragging only)
- Inspector charts padded to fixed 60-point window like aggregate charts
- Larger title, graph labels, and event log font
- Lighter borders between chart panels

Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>

- Charts show per-second bandwidth rate and delta counts for messages/blocks
- Rate calculation uses time since last actual change, not last poll
- Only emit data points when cumulative values change (no zero-spike pattern)
- Fixed-width 60-point chart window with null padding during fill-in
- Bandwidth formatted as KB/MB with units on axes and tooltips
- Disable animation on inspector panel charts

Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>

The 60s timeout on StBusy was killing peer connections when no Leios
data was produced within the window. The server legitimately waits
indefinitely for new data (same pattern as ChainSync MsgAwaitReply).

Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>

Vite + React 19 + TypeScript + Zustand + MUI + ReactFlow + d3-force + recharts.
Connects to net-cluster REST API (topology, stats, events) with 1s polling.

Features:
- Force-directed topology graph with latency-labelled edges
- Rolling aggregate charts (bandwidth, messages, blocks produced)
- Per-node inspector panel with details, peers, and per-node charts
- Streaming event log with color-coded event types

Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>

New net-cluster crate that orchestrates local multi-node test networks:
- Random topology generation with configurable degree and latency
- Per-node TOML overlay generation with HTTP telemetry sinks
- Child process spawning with log capture
- Axum HTTP server receives event batches and stats from nodes
- Watermark-based time-ordered event merge to JSONL output
- 27 unit tests covering topology, overlay, aggregator, server, config

Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>

block-extracted headers were missing #6.24 wrapping, header_hash was
hashing the era-tagged wrapper instead of just the inner header_cbor,
and consensus was injecting entire block bodies as "headers". These
combined to break block fetch routing (point mismatch) after the first
block when following a real Cardano relay.

- try_extract_header: wrap in #6.24 to match ChainSync wire format
- header_hash: hash inner header_cbor bytes (skip era tag + #6.24 framing)
- consensus: use body.header() instead of WrappedHeader::opaque(body.raw)
- production: use header.point() instead of manual hash, store raw header
  in block body (no #6.24 wrapping, matching real Cardano blocks)
- test fixtures: match real block structure (no #6.24 on in-block headers)
- add follow-real-relay.toml config for passive mainnet relay testing

Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>

Sample configs in net-node/configs/:
- mainnet.toml: base config (mainnet magic, Leios enabled, production
  params, validation timing)
- node0.toml: producer with 500/1000 stake, tx generation, file+log
  telemetry, peers to node1
- node1.toml: producer with 500/1000 stake, 50ms simulated delay on
  peer, file+log telemetry

CLAUDE.md: add net-node testing section with usage examples, add
net-node to workspace structure with module descriptions.

README.md: add test node feature summary, add net-node to workspace
structure, add test node usage section with two-node example, config
layering explanation, and telemetry output format.

Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>

Add longest-chain consensus and configurable fake validation:

- consensus.rs: tracks local_tip, fetches blocks for longer chains,
  skips self-produced blocks, deduplicates in-flight fetches, handles
  rollbacks. 4 unit tests.
- validation.rs: spawns background sleep tasks (constant + per_byte ms),
  sends completion on mpsc channel. 2 unit tests.
- config.rs: ValidationConfig (rb_head_ms, rb_body_ms_constant/per_byte,
  tx_validation_ms/per_byte)
- main.rs: consensus handles TipAdvanced/BlockReceived/RolledBack events,
  validation_rx arm in select loop

Fix coordinator bug: PeerEvent::BlockFetched now carries the original
fetch range (from, to) so the coordinator uses the authoritative point
instead of falling back to Point::Origin when body.point() fails on
opaque/fake blocks. This fixes pending_fetches cleanup, fragment
pruning, and NetworkEvent::BlockReceived point accuracy.

317 tests (293 net-core + 24 net-node), 0 clippy warnings.

Marks Stage 3 complete in test-node-plan.md.

Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>

Add VRF-based block production (lottery ported from sim-rs) and Poisson
transaction generation to the test node:

- production.rs: stake-weighted VRF lottery, fake block building, deterministic seeding
- mempool.rs: background Poisson tx generator, random body sizes
- config.rs: ProductionConfig (stake, total_stake, rb_probability) and TxConfig (tx_rate, size range)
- main.rs: slot tick triggers try_produce_block, spawns tx generator

18 tests, 0 clippy warnings. Live-tested: producer (full stake, 200ms
slots) generates blocks seen by follower via ChainSync; txs broadcast
at configured rate.

Marks Stages 1 and 2 complete in test-node-plan.md.

Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>

New net-node binary crate implementing Stage 1 of the test node plan:
- TOML config with figment layering (repeatable --config, --set overrides)
- Slot clock aligned to wall-clock with configurable slot duration
- Coordinator wrapper (duplex mode, auto peer addition)
- Main event loop: slot ticks + network events + graceful shutdown
- 10 unit tests, 0 clippy warnings

Also adds the full test-node-plan.md with 5-stage implementation plan
covering production, consensus, validation, Leios, and telemetry.

Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>