// cmd/node — full validator node with stats HTTP API.
//
// Flags:
//
//	--db           BadgerDB directory           (default: ./chaindata)
//	--listen       libp2p listen multiaddr      (default: /ip4/0.0.0.0/tcp/4001)
//	--announce     comma-separated multiaddrs to advertise to peers
//	               Required for internet deployment (VPS, Docker with fixed IP).
//	               Example: /ip4/1.2.3.4/tcp/4001
//	               Without this flag libp2p tries UPnP/NAT-PMP auto-detection.
//	--peers        comma-separated bootstrap peer multiaddrs
//	--validators   comma-separated validator pub keys
//	--genesis      create genesis block on first start
//	--key          path to identity JSON file   (default: ./node.json)
//	--stats-addr   HTTP stats server address    (default: :8080)
//	--wallet       path to payout wallet JSON   (optional)
//	--wallet-pass  passphrase for wallet file   (optional)
//	--heartbeat    enable periodic heartbeats   (default: true)
package main

import (
	"context"
	"encoding/json"
	"flag"
	"fmt"
	"log"
	"log/slog"
	"net/http"
	"os"
	"os/signal"
	"path/filepath"
	"strings"
	"sync"
	"syscall"
	"time"

	libp2ppeer "github.com/libp2p/go-libp2p/core/peer"
	"github.com/multiformats/go-multiaddr"

	"go-blockchain/blockchain"
	"go-blockchain/consensus"
	"go-blockchain/economy"
	"go-blockchain/identity"
	"go-blockchain/node"
	"go-blockchain/node/version"
	"go-blockchain/p2p"
	"go-blockchain/relay"
	"go-blockchain/vm"
	"go-blockchain/wallet"
)

const heartbeatFeeUT = 50_000 // 0.05 T

// expectedGenesisHash is populated from the --join seed's /api/network-info
// response. After sync completes we verify our local block 0 hashes to the
// same value; mismatch aborts the node (unless --allow-genesis-mismatch).
// Empty string means no --join: skip verification.
var expectedGenesisHash string

func main() {
	// Every flag below also reads a DCHAIN_* env var fallback (see envOr /
	// envBoolOr / envUint64Or). Flag on CLI wins; env on container deploys;
	// hard-coded default as last resort. Lets docker-compose drive the node
	// from an env_file without bake-in CLI strings.
	dbPath        := flag.String("db", envOr("DCHAIN_DB", "./chaindata"), "BadgerDB directory (env: DCHAIN_DB)")
	listenAddr    := flag.String("listen", envOr("DCHAIN_LISTEN", "/ip4/0.0.0.0/tcp/4001"), "libp2p listen multiaddr (env: DCHAIN_LISTEN)")
	announceFlag  := flag.String("announce", envOr("DCHAIN_ANNOUNCE", ""), "comma-separated multiaddrs to advertise to peers (env: DCHAIN_ANNOUNCE)")
	peersFlag     := flag.String("peers", envOr("DCHAIN_PEERS", ""), "comma-separated bootstrap peer multiaddrs (env: DCHAIN_PEERS)")
	validators := flag.String("validators", envOr("DCHAIN_VALIDATORS", ""), "comma-separated validator pub keys (env: DCHAIN_VALIDATORS)")
	genesisFlag := flag.Bool("genesis", envBoolOr("DCHAIN_GENESIS", false), "create genesis block (env: DCHAIN_GENESIS)")
	keyFile := flag.String("key", envOr("DCHAIN_KEY", "./node.json"), "path to identity JSON file (env: DCHAIN_KEY)")
	relayKeyFile := flag.String("relay-key", envOr("DCHAIN_RELAY_KEY", "./relay.json"), "path to relay X25519 keypair JSON (env: DCHAIN_RELAY_KEY)")
	statsAddr := flag.String("stats-addr", envOr("DCHAIN_STATS_ADDR", ":8080"), "HTTP stats server address (env: DCHAIN_STATS_ADDR)")
	walletFile := flag.String("wallet", envOr("DCHAIN_WALLET", ""), "payout wallet JSON (env: DCHAIN_WALLET)")
	walletPass := flag.String("wallet-pass", envOr("DCHAIN_WALLET_PASS", ""), "passphrase for wallet file (env: DCHAIN_WALLET_PASS)")
	heartbeatEnabled := flag.Bool("heartbeat", envBoolOr("DCHAIN_HEARTBEAT", true), "enable periodic heartbeat transactions (env: DCHAIN_HEARTBEAT)")
	registerRelay    := flag.Bool("register-relay", envBoolOr("DCHAIN_REGISTER_RELAY", false), "submit REGISTER_RELAY tx on startup (env: DCHAIN_REGISTER_RELAY)")
	relayFee         := flag.Uint64("relay-fee", envUint64Or("DCHAIN_RELAY_FEE", 1_000), "relay fee per message in µT (env: DCHAIN_RELAY_FEE)")
	mailboxDB        := flag.String("mailbox-db", envOr("DCHAIN_MAILBOX_DB", "./mailboxdata"), "BadgerDB directory for relay mailbox (env: DCHAIN_MAILBOX_DB)")
	govContractID    := flag.String("governance-contract", envOr("DCHAIN_GOVERNANCE_CONTRACT", ""), "governance contract ID for dynamic chain parameters (env: DCHAIN_GOVERNANCE_CONTRACT)")
	joinSeedURL      := flag.String("join", envOr("DCHAIN_JOIN", ""), "bootstrap from a running node: comma-separated HTTP URLs (env: DCHAIN_JOIN)")
	// Observer mode: the node participates in the P2P network, applies
	// gossiped blocks, serves HTTP/WS, and forwards submitted txs — but
	// never proposes blocks or votes in PBFT. Use this for read-heavy
	// deployments: put N observers behind a load balancer so clients can
	// hammer them without hitting validators directly. Observers also
	// never need to be in the validator set.
	observerMode := flag.Bool("observer", envBoolOr("DCHAIN_OBSERVER", false), "observer-only mode: apply blocks + serve HTTP but don't propose or vote (env: DCHAIN_OBSERVER)")
	// Log format: `text` (default, human-readable) or `json` (production,
	// machine-parsable for Loki/ELK). Applies to both the global slog
	// handler and Go's std log package — existing log.Printf calls are
	// rerouted through slog so old and new log sites share a format.
	logFormat := flag.String("log-format", envOr("DCHAIN_LOG_FORMAT", "text"), "log format: `text` or `json` (env: DCHAIN_LOG_FORMAT)")
	// Access control for the HTTP/WS API. Empty token = fully public node
	// (default). Non-empty token gates /api/tx (and WS submit_tx) behind
	// Authorization: Bearer <token>. Add --api-private to also gate every
	// read endpoint — useful for personal nodes where the operator
	// considers chat metadata private. See node/api_guards.go.
	apiToken      := flag.String("api-token", envOr("DCHAIN_API_TOKEN", ""), "Bearer token required to submit transactions (env: DCHAIN_API_TOKEN). Empty = public node")
	apiPrivate    := flag.Bool("api-private", envBoolOr("DCHAIN_API_PRIVATE", false), "also require the access token on READ endpoints (env: DCHAIN_API_PRIVATE)")
	updateSrcURL  := flag.String("update-source-url", envOr("DCHAIN_UPDATE_SOURCE_URL", ""), "Gitea /api/v1/repos/{owner}/{repo}/releases/latest URL for /api/update-check (env: DCHAIN_UPDATE_SOURCE_URL)")
	updateSrcToken := flag.String("update-source-token", envOr("DCHAIN_UPDATE_SOURCE_TOKEN", ""), "optional Gitea PAT used when polling a private repo (env: DCHAIN_UPDATE_SOURCE_TOKEN)")
	disableUI      := flag.Bool("disable-ui", envBoolOr("DCHAIN_DISABLE_UI", false), "do not register HTML block-explorer pages — JSON API and Swagger still work (env: DCHAIN_DISABLE_UI)")
	disableSwagger := flag.Bool("disable-swagger", envBoolOr("DCHAIN_DISABLE_SWAGGER", false), "do not register /swagger* endpoints (env: DCHAIN_DISABLE_SWAGGER)")
	// When --join is set, the seed's genesis_hash becomes the expected value
	// for our own block 0. After sync completes we compare; on mismatch we
	// refuse to start (fail loud) — otherwise a malicious or misconfigured
	// seed could silently trick us onto a parallel chain. Use --allow-genesis-mismatch
	// only for intentional migrations (e.g. importing data from another chain
	// into this network) — very dangerous.
	allowGenesisMismatch := flag.Bool("allow-genesis-mismatch", false, "skip the safety check that aborts when the local genesis hash differs from the seed's. Use only for explicit chain migration.")
	showVersion := flag.Bool("version", false, "print version info and exit")
	flag.Parse()

	if *showVersion {
		// Print one-line version summary then exit. Used by update.sh smoke
		// test and operators running `node --version`.
		fmt.Println(version.String())
		return
	}

	// Configure structured logging. Must run before any log.Printf call
	// so subsequent logs inherit the format.
	setupLogging(*logFormat)

	// Wire API access-control. A non-empty token gates writes; adding
	// --api-private also gates reads. Logged up-front so the operator
	// sees what mode they're in.
	node.SetAPIAccess(*apiToken, *apiPrivate)
	node.SetUpdateSource(*updateSrcURL, *updateSrcToken)
	switch {
	case *apiToken == "":
		log.Printf("[NODE] API access: public (no token set)")
	case *apiPrivate:
		log.Printf("[NODE] API access: fully private (token required on all endpoints)")
	default:
		log.Printf("[NODE] API access: public reads, token-gated writes")
	}

	// --- Zero-config onboarding: --join fetches a seed node's network info
	// and fills in --peers / --validators from it, so operators adding a new
	// node to an existing network don't have to hunt down multiaddrs or
	// validator pubkeys by hand.
	//
	// Multi-seed: --join accepts a comma-separated list. We try each URL in
	// order until one succeeds. This way a dead seed doesn't orphan a new
	// node — just list a couple backups.
	//
	// Persistence: on first successful --join we write <db>/seeds.json
	// with the tried URL list + live peer multiaddrs. On subsequent
	// restarts (no --join passed) we read this file and retry seeds
	// automatically, so the operator doesn't have to repeat the CLI flag
	// on every container restart.
	seedFile := filepath.Join(*dbPath, "seeds.json")
	seedURLs := parseSeedList(*joinSeedURL)
	if len(seedURLs) == 0 {
		// Fall back to persisted list from last run, if any.
		if persisted, err := loadSeedsFile(seedFile); err == nil && len(persisted.URLs) > 0 {
			seedURLs = persisted.URLs
			log.Printf("[NODE] loaded %d seed URL(s) from %s", len(seedURLs), seedFile)
		}
	}
	if len(seedURLs) > 0 {
		info, usedURL := tryJoinSeeds(seedURLs)
		if info == nil {
			log.Fatalf("[NODE] --join: all %d seeds failed. Check URLs and network.", len(seedURLs))
		}
		log.Printf("[NODE] --join: bootstrapped from %s", usedURL)

		if *peersFlag == "" && len(info.Peers) > 0 {
			var addrs []string
			for _, p := range info.Peers {
				for _, a := range p.Addrs {
					if a != "" {
						addrs = append(addrs, a)
						break
					}
				}
			}
			*peersFlag = strings.Join(addrs, ",")
			log.Printf("[NODE] --join: learned %d peer multiaddrs from seed", len(addrs))
		}
		if *validators == "" && len(info.Validators) > 0 {
			*validators = strings.Join(info.Validators, ",")
			log.Printf("[NODE] --join: learned %d validators from seed", len(info.Validators))
		}
		if *govContractID == "" {
			if gov, ok := info.Contracts["governance"]; ok && gov.ContractID != "" {
				*govContractID = gov.ContractID
				log.Printf("[NODE] --join: governance contract = %s", gov.ContractID)
			}
		}
		if info.ChainID != "" {
			log.Printf("[NODE] --join: connecting to chain %s (tip=%d)", info.ChainID, info.TipHeight)
		}
		if info.GenesisHash != "" {
			expectedGenesisHash = info.GenesisHash
			log.Printf("[NODE] --join: expected genesis hash = %s", info.GenesisHash)
		}

		// Persist the successful URL first, then others, so retries pick
		// the known-good one first. Include the current peer list so the
		// next cold boot can bypass the HTTP layer entirely if any peer
		// is still up.
		var orderedURLs []string
		orderedURLs = append(orderedURLs, usedURL)
		for _, u := range seedURLs {
			if u != usedURL {
				orderedURLs = append(orderedURLs, u)
			}
		}
		if err := saveSeedsFile(seedFile, &persistedSeeds{
			ChainID:     info.ChainID,
			GenesisHash: info.GenesisHash,
			URLs:        orderedURLs,
			Peers:       *peersFlag,
			SavedAt:     time.Now().UTC().Format(time.RFC3339),
		}); err != nil {
			log.Printf("[NODE] warn: could not persist seeds to %s: %v", seedFile, err)
		}
	}

	// --- Identity ---
	id := loadOrCreateIdentity(*keyFile)
	log.Printf("[NODE] pub_key: %s", id.PubKeyHex())
	log.Printf("[NODE] address: %s", wallet.PubKeyToAddress(id.PubKeyHex()))

	// --- Optional payout wallet ---
	var payoutWallet *wallet.Wallet
	if *walletFile != "" {
		w, err := wallet.Load(*walletFile, *walletPass)
		if err != nil {
			log.Fatalf("[NODE] load wallet: %v", err)
		}
		payoutWallet = w
		log.Printf("[NODE] payout wallet: %s", w.Short())
	}

	// --- Validator set ---
	var valSet []string
	if *validators != "" {
		for _, v := range strings.Split(*validators, ",") {
			if v = strings.TrimSpace(v); v != "" {
				valSet = append(valSet, v)
			}
		}
	}
	if len(valSet) == 0 {
		valSet = []string{id.PubKeyHex()}
	}
	log.Printf("[NODE] validator set (%d): %v", len(valSet), shortKeys(valSet))

	// --- Chain ---
	chain, err := blockchain.NewChain(*dbPath)
	if err != nil {
		log.Fatalf("[NODE] open chain: %v", err)
	}
	defer chain.Close()
	log.Printf("[NODE] chain height: %d  tip: %s", tipIndex(chain), tipHash(chain))

	ctx, cancel := context.WithCancel(context.Background())
	defer cancel()

	// One-time catch-up GC: reclaims any garbage that accumulated before
	// this version (which introduced the background GC loop). Nodes with
	// clean DBs finish in milliseconds; nodes upgraded from pre-GC builds
	// may take a minute or two to shrink their value log from multi-GB
	// back to actual live size. Safe to skip — the background loop will
	// eventually reclaim the same space — but doing it once up front
	// makes the upgrade visibly free disk.
	go func() {
		log.Printf("[NODE] running startup value-log compaction…")
		chain.CompactNow()
		log.Printf("[NODE] startup compaction done")
	}()

	// Background value-log garbage collector for the chain DB.
	// Without this, every overwrite of a hot key (`height`, `netstats`) leaves
	// the previous bytes pinned in BadgerDB's value log forever — the DB ends
	// up multiple gigabytes on disk even when live state is megabytes. Runs
	// every 5 minutes; stops automatically when ctx is cancelled.
	chain.StartValueLogGC(ctx)

	// Genesis-hash sanity check for --join'd nodes. Run in background so
	// startup isn't blocked while waiting for sync to complete on first
	// launch, but fatal-error the process if the local genesis eventually
	// diverges from what the seed reported.
	if expectedGenesisHash != "" {
		go startGenesisVerifier(ctx, chain, expectedGenesisHash, *allowGenesisMismatch)
	}

	// --- WASM VM ---
	contractVM := vm.NewVM(ctx)
	defer contractVM.Close(ctx)
	chain.SetVM(contractVM)

	// --- Native system contracts ---
	// Register built-in Go contracts so they're callable via CALL_CONTRACT
	// with the same on-chain semantics as WASM contracts, but without the
	// WASM VM in the hot path. The username registry lives here because
	// it's the most latency-sensitive service (every chat/contact-add
	// does a resolve/reverseResolve).
	chain.RegisterNative(blockchain.NewUsernameRegistry())
	log.Printf("[NODE] registered native contract: %s", blockchain.UsernameRegistryID)

	// --- Governance contract (optional) ---
	if *govContractID != "" {
		chain.SetGovernanceContract(*govContractID)
	}

	// --- SSE event hub ---
	sseHub := node.NewSSEHub()

	// --- WebSocket gateway ---
	// Same event sources as SSE but pushed over a persistent ws:// connection
	// so mobile clients don't have to poll. Hello frame carries chain_id +
	// tip_height so the client can verify it's on the right chain.
	wsHub := node.NewWSHub(
		func() string {
			if g, err := chain.GetBlock(0); err == nil && g != nil {
				return "dchain-" + g.HashHex()[:12]
			}
			return "dchain-unknown"
		},
		chain.TipIndex,
	)
	// Event bus: one emit site per event; consumers (SSE, WS, future
	// indexers) are registered once below. Eliminates the duplicated
	// `go sseHub.EmitX / go wsHub.EmitX` pattern at every commit callsite.
	eventBus := node.NewEventBus()
	eventBus.Register(node.WrapSSE(sseHub))
	eventBus.Register(node.WrapWS(wsHub))

	// submit_tx handler is wired after `engine` and `h` are constructed —
	// see further down where `wsHub.SetSubmitTxHandler(...)` is called.

	// Auth binding: Ed25519 pubkey → X25519 pubkey from the identity registry.
	// Lets the hub enforce `inbox:<x25519>` subscriptions so a client can
	// only listen to their OWN inbox, not anyone else's.
	wsHub.SetX25519ForPub(func(edHex string) (string, error) {
		info, err := chain.IdentityInfo(edHex)
		if err != nil || info == nil {
			return "", err
		}
		return info.X25519Pub, nil
	})

	// Live peer-count gauge is registered below (after `h` is constructed).

	// --- Stats tracker ---
	stats := node.NewTracker()

	// --- Announce addresses (internet deployment) ---
	// Parse --announce flag into []multiaddr.Multiaddr.
	// When set, only these addresses are advertised to peers — libp2p's
	// auto-detected addresses (internal interfaces, loopback) are suppressed.
	// This is essential when running on a VPS or inside Docker with a fixed IP:
	//   --announce /ip4/203.0.113.10/tcp/4001
	var announceAddrs []multiaddr.Multiaddr
	if *announceFlag != "" {
		for _, s := range strings.Split(*announceFlag, ",") {
			s = strings.TrimSpace(s)
			if s == "" {
				continue
			}
			ma, err := multiaddr.NewMultiaddr(s)
			if err != nil {
				log.Fatalf("[NODE] bad --announce addr %q: %v", s, err)
			}
			announceAddrs = append(announceAddrs, ma)
			log.Printf("[NODE] announce addr: %s", s)
		}
	}

	// --- P2P ---
	h, err := p2p.NewHost(ctx, id, *listenAddr, announceAddrs)
	if err != nil {
		log.Fatalf("[NODE] p2p: %v", err)
	}
	defer h.Close()

	// Start peer-version gossip. Best-effort: errors here are logged but do
	// not block node startup — the version map is purely advisory.
	if err := h.StartVersionGossip(ctx, node.ProtocolVersion); err != nil {
		log.Printf("[NODE] peer-version gossip unavailable: %v", err)
	}

	// Wire peer connect/disconnect into stats
	// Live peer-count gauge for Prometheus — queried at scrape time.
	// Registered here (not earlier) because we need `h` to exist.
	node.NewGaugeFunc("dchain_peer_count_live",
		"Live libp2p peer count (queried on scrape)",
		func() int64 { return int64(h.PeerCount()) })

	// engine captured below after creation; use a pointer-to-pointer pattern
	var engineRef *consensus.Engine
	h.OnPeerConnected(func(pid libp2ppeer.ID) {
		stats.PeerConnected(pid)
		eng := engineRef // capture current value
		go syncOnConnect(ctx, h, chain, stats, pid, func(next uint64) {
			if eng != nil {
				eng.SyncSeqNum(next)
			}
		})
	})

	// --- Genesis ---
	if *genesisFlag && chain.Tip() == nil {
		genesis := blockchain.GenesisBlock(id.PubKeyHex(), id.PrivKey)
		if err := chain.AddBlock(genesis); err != nil {
			log.Fatalf("[NODE] add genesis: %v", err)
		}
		log.Printf("[NODE] genesis block: %s", genesis.HashHex())
	}

	// Seed the on-chain validator set from CLI flags (idempotent).
	if err := chain.InitValidators(valSet); err != nil {
		log.Fatalf("[NODE] init validators: %v", err)
	}
	// Prefer the on-chain set if it has been extended by ADD_VALIDATOR txs.
	if onChain, err := chain.ValidatorSet(); err == nil && len(onChain) > 0 {
		valSet = onChain
		log.Printf("[NODE] loaded %d validators from chain state", len(valSet))
	}

	// --- Consensus engine ---
	var seqNum uint64
	if tip := chain.Tip(); tip != nil {
		seqNum = tip.Index + 1
	}

	engine := consensus.NewEngine(
		id, valSet, seqNum,
		func(b *blockchain.Block) {
			// Time the AddBlock call end-to-end so we can see slow commits
			// (typically contract calls) in the Prometheus histogram.
			commitStart := time.Now()
			if err := chain.AddBlock(b); err != nil {
				log.Printf("[NODE] add block #%d: %v", b.Index, err)
				return
			}
			node.MetricBlockCommitSeconds.Observe(time.Since(commitStart).Seconds())
			node.MetricBlocksTotal.Inc()
			if len(b.Transactions) > 0 {
				node.MetricTxsTotal.Add(uint64(len(b.Transactions)))
			}
			// Remove committed transactions from our mempool so we don't
			// re-propose them in the next round (non-proposing validators
			// keep txs in their pending list until explicitly pruned).
			if engineRef != nil {
				engineRef.PruneTxs(b.Transactions)
			}
			stats.BlocksCommitted.Add(1)
			// Push live events to every registered bus consumer (SSE,
			// WS, future indexers). Single emit per event type.
			go eventBus.EmitBlockWithTxs(b)
			go emitContractLogsViaBus(eventBus, chain, b)

			r := economy.ComputeBlockReward(b)
			log.Printf("[NODE] %s", r.Summary())

			// Show balance of whoever receives the reward
			rewardKey := b.Validator
			if binding, _ := chain.WalletBinding(b.Validator); binding != "" {
				rewardKey = binding
			}
			bal, _ := chain.Balance(rewardKey)
			log.Printf("[NODE] reward target %s balance: %s",
				wallet.PubKeyToAddress(rewardKey), economy.FormatTokens(bal))

			// Hot-reload validator set if this block changed it.
			for _, tx := range b.Transactions {
				if tx.Type == blockchain.EventAddValidator || tx.Type == blockchain.EventRemoveValidator {
					if newSet, err := chain.ValidatorSet(); err == nil {
						if eng := engineRef; eng != nil {
							eng.UpdateValidators(newSet)
						}
					}
					break
				}
			}

			// Gossip committed block to peers
			if err := h.PublishBlock(b); err != nil {
				log.Printf("[NODE] publish block: %v", err)
			}
			stats.BlocksGossipSent.Add(1)
		},
		func(msg *blockchain.ConsensusMsg) {
			if err := h.BroadcastConsensus(msg); err != nil {
				log.Printf("[NODE] broadcast consensus: %v", err)
			}
			stats.ConsensusMsgsSent.Add(1)
		},
	)

	// Assign engine to the ref captured by peer-connect handler
	engineRef = engine

	// Wire the WS submit_tx handler now that `engine` and `h` exist.
	// Mirrors the HTTP POST /api/tx path so clients get the same
	// validation regardless of transport.
	wsHub.SetSubmitTxHandler(func(txJSON []byte) (string, error) {
		var tx blockchain.Transaction
		if err := json.Unmarshal(txJSON, &tx); err != nil {
			return "", fmt.Errorf("invalid tx JSON: %w", err)
		}
		if err := node.ValidateTxTimestamp(&tx); err != nil {
			return "", fmt.Errorf("bad timestamp: %w", err)
		}
		if err := identity.VerifyTx(&tx); err != nil {
			return "", fmt.Errorf("invalid signature: %w", err)
		}
		if err := engine.AddTransaction(&tx); err != nil {
			return "", err
		}
		if err := h.PublishTx(&tx); err != nil {
			// Mempool accepted — gossip failure is non-fatal for the caller.
			log.Printf("[NODE] ws submit_tx: gossip publish failed (mempool has it): %v", err)
		}
		return tx.ID, nil
	})

	// Wrap consensus stats
	engine.OnPropose(func() { stats.BlocksProposed.Add(1) })
	engine.OnVote(func() { stats.VotesCast.Add(1) })
	engine.OnViewChange(func() { stats.ViewChanges.Add(1) })

	// Register direct-stream consensus handler
	h.SetConsensusMsgHandler(func(msg *blockchain.ConsensusMsg) {
		stats.ConsensusMsgsRecv.Add(1)
		engine.HandleMessage(msg)
	})

	// --- Sync protocol handler ---
	h.SetSyncHandler(
		func(index uint64) (*blockchain.Block, error) { return chain.GetBlock(index) },
		func() uint64 {
			if tip := chain.Tip(); tip != nil {
				return tip.Index + 1
			}
			return 0
		},
	)

	// --- Bootstrap peers ---
	if *peersFlag != "" {
		for _, addr := range strings.Split(*peersFlag, ",") {
			if addr = strings.TrimSpace(addr); addr != "" {
				go connectWithRetry(ctx, h, addr)
			}
		}
	}

	h.Advertise(ctx)
	h.DiscoverPeers(ctx)

	// --- Incoming gossip loops ---
	txMsgs := h.TxMsgs(ctx)
	blockMsgs := h.BlockMsgs(ctx)

	go func() {
		for tx := range txMsgs {
			stats.TxsGossipRecv.Add(1)
			if err := engine.AddTransaction(tx); err != nil {
				log.Printf("[NODE] rejected gossip tx %s: %v", tx.ID, err)
				continue
			}
			// Trigger an immediate block proposal so new transactions are
			// committed quickly instead of waiting for the next tick.
			if engine.IsLeader() {
				go engine.Propose(chain.Tip())
			}
		}
	}()

	// gapFillRecent tracks the last time we asked a given peer to fill in
	// missing blocks, so a burst of out-of-order gossips can't cause a
	// sync storm. One sync request per peer per minute is plenty —
	// SyncFromPeerFull drains whatever is needed in a single call.
	gapFillRecent := make(map[libp2ppeer.ID]time.Time)
	var gapFillMu sync.Mutex

	go func() {
		for m := range blockMsgs {
			b := m.Block
			stats.BlocksGossipRecv.Add(1)
			tip := chain.Tip()
			if tip != nil && b.Index <= tip.Index {
				continue
			}
			// Gap detected — we received block N but our tip is N-k for k>1.
			// Gossipsub won't replay the missing blocks, so we have to ask
			// the gossiper (or any peer) to stream them to us explicitly.
			// Without this the node silently falls behind forever and never
			// catches up, even though the rest of the network is live.
			if tip != nil && b.Index > tip.Index+1 {
				gapFillMu.Lock()
				last := gapFillRecent[m.From]
				if time.Since(last) > 60*time.Second {
					gapFillRecent[m.From] = time.Now()
					gapFillMu.Unlock()
					log.Printf("[NODE] gap detected: gossiped #%d, tip=#%d — requesting sync from %s",
						b.Index, tip.Index, m.From)
					eng := engineRef
					go syncOnConnect(ctx, h, chain, stats, m.From, func(next uint64) {
						if eng != nil {
							eng.SyncSeqNum(next)
						}
					})
				} else {
					gapFillMu.Unlock()
					log.Printf("[NODE] gap detected but sync from %s throttled (cooldown); dropping #%d",
						m.From, b.Index)
				}
				continue
			}
			if err := chain.AddBlock(b); err != nil {
				log.Printf("[NODE] add gossip block #%d: %v", b.Index, err)
				continue
			}
			engine.SyncSeqNum(b.Index + 1)
			log.Printf("[NODE] applied gossip block #%d hash=%s", b.Index, b.HashHex()[:8])
			go eventBus.EmitBlockWithTxs(b)
		}
	}()

	// --- Relay mailbox ---
	mailbox, err := relay.OpenMailbox(*mailboxDB)
	if err != nil {
		log.Fatalf("[NODE] relay mailbox: %v", err)
	}
	defer mailbox.Close()
	go mailbox.RunGC()
	log.Printf("[NODE] relay mailbox: %s", *mailboxDB)

	// Push-notify bus consumers whenever a fresh envelope lands in the
	// mailbox. Clients subscribed to `inbox:<my_x25519>` (via WS) get the
	// event immediately so they no longer need to poll /relay/inbox.
	//
	// We send a minimal summary (no ciphertext) — the client refetches from
	// /relay/inbox if it needs the full envelope. Keeps WS frames small and
	// avoids a fat push for every message.
	mailbox.SetOnStore(func(env *relay.Envelope) {
		sum, _ := json.Marshal(map[string]any{
			"id":            env.ID,
			"recipient_pub": env.RecipientPub,
			"sender_pub":    env.SenderPub,
			"sent_at":       env.SentAt,
		})
		eventBus.EmitInbox(env.RecipientPub, sum)
	})

	// --- Relay router ---
	relayKP, err := relay.LoadOrCreateKeyPair(*relayKeyFile)
	if err != nil {
		log.Fatalf("[NODE] relay keypair: %v", err)
	}
	relayRouter, err := relay.NewRouter(
		h.LibP2PHost(),
		h.GossipSub(),
		relayKP,
		id,
		mailbox,
		func(envID, senderPub string, msg []byte) {
			senderShort := senderPub
			if len(senderShort) > 8 {
				senderShort = senderShort[:8]
			}
			log.Printf("[RELAY] delivered envelope %s from %s (%d bytes)", envID, senderShort, len(msg))
		},
		func(tx *blockchain.Transaction) error {
			if err := engine.AddTransaction(tx); err != nil {
				return err
			}
			return h.PublishTx(tx)
		},
	)
	if err != nil {
		log.Fatalf("[NODE] relay router: %v", err)
	}
	go relayRouter.Run(ctx)
	log.Printf("[NODE] relay pub key: %s", relayRouter.RelayPubHex())

	// --- Register relay service on-chain (optional) ---
	// Observers can still act as relays (they forward encrypted envelopes
	// and serve /relay/inbox the same way validators do) — but only if
	// the operator explicitly asks. Default-off for observers avoids
	// surprise on-chain registration txs.
	if *registerRelay && !*observerMode {
		go autoRegisterRelay(ctx, id, relayKP, *relayFee, h.AddrStrings(), chain, engine, h)
	} else if *registerRelay && *observerMode {
		log.Printf("[NODE] --register-relay ignored in observer mode; run with both flags off an observer to register as relay")
	}

	// --- Validator liveness metric updater ---
	// Engine tracks per-validator last-seen seqNum; we scan periodically
	// and publish the worst offender into Prometheus so alerts can fire.
	go func() {
		t := time.NewTicker(15 * time.Second)
		defer t.Stop()
		for {
			select {
			case <-ctx.Done():
				return
			case <-t.C:
				report := engine.LivenessReport()
				var worst uint64
				for _, missed := range report {
					if missed > worst {
						worst = missed
					}
				}
				node.MetricMaxMissedBlocks.Set(int64(worst))
			}
		}
	}()

	// --- Periodic heartbeat transaction ---
	// Observers don't heartbeat — they're not validators, their uptime
	// doesn't affect consensus, and heartbeat is an on-chain fee-paying
	// tx that would waste tokens.
	if *heartbeatEnabled && !*observerMode {
		go heartbeatLoop(ctx, id, chain, h, stats)
	} else if *observerMode {
		log.Printf("[NODE] heartbeat disabled (observer mode)")
	} else {
		log.Printf("[NODE] heartbeat disabled by config")
	}

	// --- Bind wallet at startup if provided and not yet bound ---
	if payoutWallet != nil {
		go autoBindWallet(ctx, id, payoutWallet, chain, engine)
	}

	// --- Block production loop (leader only) ---
	//
	// Two tiers:
	//   fastTicker  (500 ms)  — proposes only when the mempool is non-empty.
	//                           Keeps tx latency low without flooding empty blocks.
	//   idleTicker  (5 s)     — heartbeat: proposes even if the mempool is empty
	//                           so the genesis block is produced quickly on startup
	//                           and the chain height advances so peers can sync.
	// Observer mode skips the producer entirely — they never propose or
	// vote. They still receive blocks via gossipsub and forward txs, so
	// this keeps them light-weight + certain not to participate in
	// consensus even if erroneously listed as a validator.
	if *observerMode {
		log.Printf("[NODE] observer mode: block producer disabled")
	}
	go func() {
		if *observerMode {
			return
		}
		fastTicker := time.NewTicker(500 * time.Millisecond)
		idleTicker := time.NewTicker(5 * time.Second)
		// heartbeatTicker: independent of tip-reads and mempool. Its sole job
		// is to prove this producer goroutine is actually alive so we can
		// tell "chain frozen because consensus halted" apart from "chain
		// frozen because producer deadlocked on c.mu". Logs at most once per
		// minute; should appear continuously on any live node.
		lifeTicker := time.NewTicker(60 * time.Second)
		defer fastTicker.Stop()
		defer idleTicker.Stop()
		defer lifeTicker.Stop()
		lastProposedHeight := uint64(0)
		for {
			select {
			case <-ctx.Done():
				return
			case <-fastTicker.C:
				if engine.IsLeader() && engine.HasPendingTxs() {
					tip := chain.Tip()
					engine.Propose(tip)
					if tip != nil {
						lastProposedHeight = tip.Index
					}
				}
			case <-idleTicker.C:
				if engine.IsLeader() {
					tip := chain.Tip()
					engine.Propose(tip) // heartbeat block, may be empty
					if tip != nil {
						lastProposedHeight = tip.Index
					}
				}
			case <-lifeTicker.C:
				// Proof of life + last proposal height so we can see from
				// logs whether the loop is running AND reaching the chain.
				log.Printf("[NODE] producer alive (leader=%v lastProposed=%d tip=%d)",
					engine.IsLeader(), lastProposedHeight, chain.TipIndex())
			}
		}
	}()

	// --- Stats HTTP server ---
	statsQuery := node.QueryFunc{
		PubKey:     func() string { return id.PubKeyHex() },
		PeerID:     func() string { return h.PeerID() },
		PeersCount: func() int { return h.PeerCount() },
		ChainTip:   func() *blockchain.Block { return chain.Tip() },
		Balance: func(pk string) uint64 {
			b, _ := chain.Balance(pk)
			return b
		},
		WalletBinding: func(pk string) string {
			binding, _ := chain.WalletBinding(pk)
			if binding == "" {
				return ""
			}
			return wallet.PubKeyToAddress(binding)
		},
		Reputation: func(pk string) blockchain.RepStats {
			r, _ := chain.Reputation(pk)
			return r
		},
	}
	explorerQuery := node.ExplorerQuery{
		GetBlock:        chain.GetBlock,
		GetTx:           chain.TxByID,
		AddressToPubKey: chain.AddressToPubKey,
		Balance:         func(pk string) (uint64, error) { return chain.Balance(pk) },
		Reputation: func(pk string) (blockchain.RepStats, error) {
			return chain.Reputation(pk)
		},
		WalletBinding: func(pk string) (string, error) {
			return chain.WalletBinding(pk)
		},
		TxsByAddress: func(pk string, limit, offset int) ([]*blockchain.TxRecord, error) {
			return chain.TxsByAddress(pk, limit, offset)
		},
		RecentBlocks:     chain.RecentBlocks,
		RecentTxs:        chain.RecentTxs,
		NetStats:         chain.NetworkStats,
		RegisteredRelays: chain.RegisteredRelays,
		IdentityInfo: func(pubKeyOrAddr string) (*blockchain.IdentityInfo, error) {
			return chain.IdentityInfo(pubKeyOrAddr)
		},
		ValidatorSet: chain.ValidatorSet,
		SubmitTx: func(tx *blockchain.Transaction) error {
			if err := engine.AddTransaction(tx); err != nil {
				return err
			}
			return h.PublishTx(tx)
		},
		GetContract:      chain.GetContract,
		GetContracts:     chain.Contracts,
		GetContractState: chain.GetContractState,
		GetContractLogs:  chain.ContractLogs,
		Stake:            chain.Stake,
		GetToken:         chain.Token,
		GetTokens:        chain.Tokens,
		TokenBalance:     chain.TokenBalance,
		GetNFT:            chain.NFT,
		GetNFTs:           chain.NFTs,
		NFTsByOwner:       chain.NFTsByOwner,
		GetChannel:        chain.Channel,
		GetChannelMembers: chain.ChannelMembers,
		Events:           sseHub,
		WS:               wsHub,
		// Onboarding: expose libp2p peers + chain_id so new nodes/clients can
		// fetch /api/network-info to bootstrap without static configuration.
		ConnectedPeers: func() []node.ConnectedPeerRef {
			if h == nil {
				return nil
			}
			src := h.ConnectedPeers()
			// Pull peer-version map once per call so the N peers share a
			// consistent snapshot rather than a racing one.
			versions := h.PeerVersions()
			out := make([]node.ConnectedPeerRef, len(src))
			for i, p := range src {
				ref := node.ConnectedPeerRef{ID: p.ID, Addrs: p.Addrs}
				if pv, ok := versions[p.ID]; ok {
					ref.Version = &node.PeerVersionRef{
						Tag:             pv.Tag,
						Commit:          pv.Commit,
						ProtocolVersion: pv.ProtocolVersion,
						Timestamp:       pv.Timestamp,
						ReceivedAt:      pv.ReceivedAt.Format(time.RFC3339),
					}
				}
				out[i] = ref
			}
			return out
		},
		ChainID: func() string {
			// Derive from genesis block so every node with the same genesis
			// reports the same chain_id. Fall back to a static string if the
			// chain has no genesis yet (fresh DB, pre-sync).
			if g, err := chain.GetBlock(0); err == nil && g != nil {
				return "dchain-" + g.HashHex()[:12]
			}
			return "dchain-unknown"
		},
		// Native contracts registered with the chain are exposed through the
		// well-known endpoint so clients don't have to know which ones are
		// native vs WASM. They use the contract_id transparently.
		NativeContracts: func() []node.NativeContractInfo {
			all := chain.NativeContracts()
			out := make([]node.NativeContractInfo, len(all))
			for i, nc := range all {
				out[i] = node.NativeContractInfo{
					ContractID: nc.ID(),
					ABIJson:    nc.ABI(),
				}
			}
			return out
		},
	}

	relayConfig := node.RelayConfig{
		Mailbox: mailbox,
		Send: func(recipientPubHex string, msg []byte) (string, error) {
			return relayRouter.Send(recipientPubHex, msg, 0)
		},
		Broadcast: func(env *relay.Envelope) error {
			return relayRouter.Broadcast(env)
		},
		ContactRequests: func(pubKey string) ([]blockchain.ContactInfo, error) {
			return chain.ContactRequests(pubKey)
		},
	}

	go func() {
		log.Printf("[NODE] stats API:   http://0.0.0.0%s/stats", *statsAddr)
		if *disableUI {
			log.Printf("[NODE] explorer UI: disabled (--disable-ui)")
		} else {
			log.Printf("[NODE] explorer UI: http://0.0.0.0%s/", *statsAddr)
		}
		if *disableSwagger {
			log.Printf("[NODE] swagger:     disabled (--disable-swagger)")
		} else {
			log.Printf("[NODE] swagger:     http://0.0.0.0%s/swagger", *statsAddr)
		}
		log.Printf("[NODE] relay inbox: http://0.0.0.0%s/relay/inbox?pub=<x25519hex>", *statsAddr)
		routeFlags := node.ExplorerRouteFlags{
			DisableUI:      *disableUI,
			DisableSwagger: *disableSwagger,
		}
		if err := stats.ListenAndServe(*statsAddr, statsQuery, func(mux *http.ServeMux) {
			node.RegisterExplorerRoutes(mux, explorerQuery, routeFlags)
			node.RegisterRelayRoutes(mux, relayConfig)

			// POST /api/governance/link — link deployed contracts at runtime.
			// Body: {"governance": "<id>"}
			// Allows deploy script to wire governance without node restart.
			mux.HandleFunc("/api/governance/link", func(w http.ResponseWriter, r *http.Request) {
				if r.Method != http.MethodPost {
					http.Error(w, `{"error":"method not allowed"}`, http.StatusMethodNotAllowed)
					return
				}
				var body struct {
					Governance string `json:"governance"`
				}
				if err := json.NewDecoder(r.Body).Decode(&body); err != nil {
					http.Error(w, `{"error":"bad JSON"}`, http.StatusBadRequest)
					return
				}
				if body.Governance != "" {
					chain.SetGovernanceContract(body.Governance)
				}
				w.Header().Set("Content-Type", "application/json")
				fmt.Fprintf(w, `{"status":"ok","governance":%q}`, body.Governance)
			})
		}); err != nil {
			log.Printf("[NODE] stats server error: %v", err)
		}
	}()

	log.Printf("[NODE] running — peer ID: %s", h.PeerID())
	log.Printf("[NODE] addrs: %v", h.AddrStrings())

	sig := make(chan os.Signal, 1)
	signal.Notify(sig, syscall.SIGINT, syscall.SIGTERM)
	<-sig
	log.Println("[NODE] shutting down...")
}

// emitContractLogsViaBus scans b's CALL_CONTRACT transactions and emits a
// contract_log event for each log entry through the event bus. Bus fans
// it out to every registered consumer (SSE, WS, future indexers) — no
// hub-by-hub emit calls here.
func emitContractLogsViaBus(bus *node.EventBus, chain *blockchain.Chain, b *blockchain.Block) {
	for _, tx := range b.Transactions {
		if tx.Type != blockchain.EventCallContract {
			continue
		}
		var p blockchain.CallContractPayload
		if err := json.Unmarshal(tx.Payload, &p); err != nil || p.ContractID == "" {
			continue
		}
		// Fetch only logs from this block (limit=50 to avoid loading all history).
		logs, err := chain.ContractLogs(p.ContractID, 50)
		if err != nil {
			continue
		}
		for _, entry := range logs {
			if entry.BlockHeight == b.Index && entry.TxID == tx.ID {
				bus.EmitContractLog(entry)
			}
		}
	}
}

// heartbeatLoop publishes a HEARTBEAT transaction once per hour.
func heartbeatLoop(ctx context.Context, id *identity.Identity, chain *blockchain.Chain, h *p2p.Host, stats *node.Tracker) {
	ticker := time.NewTicker(60 * time.Minute)
	defer ticker.Stop()
	for {
		select {
		case <-ctx.Done():
			return
		case <-ticker.C:
			bal, err := chain.Balance(id.PubKeyHex())
			if err != nil {
				log.Printf("[NODE] heartbeat balance check failed: %v", err)
				continue
			}
			if bal < heartbeatFeeUT {
				log.Printf("[NODE] heartbeat skipped: balance %s below required fee %s",
					economy.FormatTokens(bal), economy.FormatTokens(heartbeatFeeUT))
				continue
			}

			var height uint64
			if tip := chain.Tip(); tip != nil {
				height = tip.Index
			}
			payload := blockchain.HeartbeatPayload{
				PubKey:      id.PubKeyHex(),
				ChainHeight: height,
				PeerCount:   h.PeerCount(),
				Version:     "0.1.0",
			}
			pb, _ := json.Marshal(payload)
			tx := &blockchain.Transaction{
				ID:        fmt.Sprintf("hb-%s-%d", id.PubKeyHex()[:8], time.Now().Unix()),
				Type:      blockchain.EventHeartbeat,
				From:      id.PubKeyHex(),
				Amount:    0,
				Payload:   pb,
				Fee:       heartbeatFeeUT,
				Memo:      "Heartbeat: liveness proof",
				Timestamp: time.Now().UTC(),
			}
			tx.Signature = id.Sign(identity.TxSignBytes(tx))

			if err := h.PublishTx(tx); err != nil {
				log.Printf("[NODE] heartbeat publish: %v", err)
			} else {
				stats.TxsGossipSent.Add(1)
				log.Printf("[NODE] heartbeat sent (height=%d peers=%d)", height, h.PeerCount())
			}
		}
	}
}

// autoBindWallet submits a BIND_WALLET tx if the node doesn't have one yet.
func autoBindWallet(ctx context.Context, id *identity.Identity, w *wallet.Wallet, chain *blockchain.Chain, engine *consensus.Engine) {
	// Wait a bit for chain to sync
	time.Sleep(5 * time.Second)

	binding, err := chain.WalletBinding(id.PubKeyHex())
	if err != nil || binding == w.ID.PubKeyHex() {
		return // already bound or error
	}

	payload := blockchain.BindWalletPayload{
		WalletPubKey: w.ID.PubKeyHex(),
		WalletAddr:   w.Address,
	}
	pb, _ := json.Marshal(payload)
	tx := &blockchain.Transaction{
		ID:        fmt.Sprintf("bind-%d", time.Now().UnixNano()),
		Type:      blockchain.EventBindWallet,
		From:      id.PubKeyHex(),
		To:        w.ID.PubKeyHex(),
		Payload:   pb,
		Fee:       blockchain.MinFee,
		Memo:      "Bind payout wallet",
		Timestamp: time.Now().UTC(),
	}
	tx.Signature = id.Sign(identity.TxSignBytes(tx))

	if err := engine.AddTransaction(tx); err != nil {
		log.Printf("[NODE] BIND_WALLET tx rejected: %v", err)
		return
	}
	log.Printf("[NODE] queued BIND_WALLET → %s", w.Address)
}

// startGenesisVerifier watches the local chain and aborts the node if, once
// block 0 is present, its hash differs from what the --join seed advertised.
//
// Lifecycle:
//   - Polls every 500 ms for up to 2 minutes waiting for sync to produce
//     block 0. On a fresh-db joiner this may take tens of seconds while
//     SyncFromPeerFull streams blocks.
//   - Once the block exists, compares HashHex() with `expected`:
//       match   → log success, exit goroutine
//       mismatch → log.Fatal, killing the process (unless `allowMismatch`
//                  is set, in which case we just warn and return)
//   - If the window expires without a genesis, we assume something went
//     wrong with peer connectivity and abort — otherwise PBFT would
//     helpfully produce its OWN genesis, permanently forking us from the
//     target network.
func startGenesisVerifier(ctx context.Context, chain *blockchain.Chain, expected string, allowMismatch bool) {
	deadline := time.NewTimer(2 * time.Minute)
	defer deadline.Stop()
	tick := time.NewTicker(500 * time.Millisecond)
	defer tick.Stop()

	for {
		select {
		case <-ctx.Done():
			return
		case <-deadline.C:
			if allowMismatch {
				log.Printf("[NODE] genesis verifier: no genesis after 2m — giving up (allow-genesis-mismatch set)")
				return
			}
			log.Fatalf("[NODE] genesis verifier: could not load block 0 within 2 minutes; seed expected hash %s. Check network connectivity to peers.", expected)
		case <-tick.C:
			g, err := chain.GetBlock(0)
			if err != nil || g == nil {
				continue // still syncing
			}
			actual := g.HashHex()
			if actual == expected {
				log.Printf("[NODE] genesis verified: hash %s matches seed", actual)
				return
			}
			if allowMismatch {
				log.Printf("[NODE] WARNING: genesis hash mismatch (local=%s seed=%s) — continuing because --allow-genesis-mismatch is set", actual, expected)
				return
			}
			log.Fatalf("[NODE] FATAL: genesis hash mismatch — local=%s seed=%s. Either the seed at --join is on a different chain, or your local DB was created against a different genesis. Wipe --db and retry, or pass --allow-genesis-mismatch if you know what you're doing.",
				actual, expected)
		}
	}
}

// syncOnConnect syncs chain from a newly connected peer.
// engineSyncFn is called with next seqNum after a successful sync so the
// consensus engine knows which block to propose/commit next.
func syncOnConnect(ctx context.Context, h *p2p.Host, chain *blockchain.Chain, stats *node.Tracker, pid libp2ppeer.ID, engineSyncFn func(uint64)) {
	time.Sleep(500 * time.Millisecond)
	syncCtx, cancel := context.WithTimeout(ctx, 30*time.Second)
	defer cancel()

	var localCount uint64
	if tip := chain.Tip(); tip != nil {
		localCount = tip.Index + 1
	}

	n, err := h.SyncFromPeerFull(syncCtx, pid, localCount, func(b *blockchain.Block) error {
		if tip := chain.Tip(); tip != nil && b.Index <= tip.Index {
			return nil
		}
		if err := chain.AddBlock(b); err != nil {
			return err
		}
		log.Printf("[SYNC] applied block #%d hash=%s", b.Index, b.HashHex()[:8])
		return nil
	})
	if err != nil {
		log.Printf("[SYNC] sync from %s: %v", pid, err)
		return
	}
	if n > 0 {
		stats.RecordSyncFrom(pid, n)
		if tip := chain.Tip(); tip != nil {
			engineSyncFn(tip.Index + 1)
			log.Printf("[SYNC] synced %d blocks from %s, chain now at #%d", n, pid, tip.Index)
		}
	} else if localCount == 0 {
		// No new blocks but we confirmed our genesis is current — still notify engine
		if tip := chain.Tip(); tip != nil {
			engineSyncFn(tip.Index + 1)
		}
	}
}

// connectWithRetry dials a bootstrap peer and keeps the connection alive.
//
// Behaviour:
//   - Retries indefinitely with exponential backoff (1 s → 2 s → … → 30 s max).
//   - After a successful connection it re-dials every 30 s.
//     libp2p.Connect is idempotent — if the peer is still connected the call
//     returns immediately. If it dropped, this transparently reconnects it.
//   - Stops only when ctx is cancelled (node shutdown).
//
// This ensures that even if a bootstrap peer restarts or is temporarily
// unreachable the node will reconnect without manual intervention.
func connectWithRetry(ctx context.Context, h *p2p.Host, addr string) {
	const (
		minBackoff   = 1 * time.Second
		maxBackoff   = 30 * time.Second
		keepAlive    = 30 * time.Second
	)
	backoff := minBackoff
	for {
		select {
		case <-ctx.Done():
			return
		default:
		}
		if err := h.Connect(ctx, addr); err != nil {
			log.Printf("[P2P] bootstrap %s unavailable: %v (retry in %s)", addr, err, backoff)
			select {
			case <-ctx.Done():
				return
			case <-time.After(backoff):
			}
			if backoff < maxBackoff {
				backoff *= 2
				if backoff > maxBackoff {
					backoff = maxBackoff
				}
			}
			continue
		}
		// Connected (or already connected). Log only on first connect or reconnect.
		if backoff > minBackoff {
			log.Printf("[P2P] reconnected to bootstrap %s", addr)
		} else {
			log.Printf("[P2P] connected to bootstrap %s", addr)
		}
		backoff = minBackoff // reset for next reconnection cycle
		// Keep-alive: re-check every 30 s so we detect and repair drops.
		select {
		case <-ctx.Done():
			return
		case <-time.After(keepAlive):
		}
	}
}

// --- helpers ---

type keyJSON struct {
	PubKey      string `json:"pub_key"`
	PrivKey     string `json:"priv_key"`
	X25519Pub   string `json:"x25519_pub,omitempty"`
	X25519Priv  string `json:"x25519_priv,omitempty"`
}

func loadOrCreateIdentity(keyFile string) *identity.Identity {
	if data, err := os.ReadFile(keyFile); err == nil {
		var kj keyJSON
		if err := json.Unmarshal(data, &kj); err == nil {
			if id, err := identity.FromHexFull(kj.PubKey, kj.PrivKey, kj.X25519Pub, kj.X25519Priv); err == nil {
				// If the file is missing X25519 keys, backfill and re-save.
				if kj.X25519Pub == "" {
					kj.X25519Pub = id.X25519PubHex()
					kj.X25519Priv = id.X25519PrivHex()
					if out, err2 := json.MarshalIndent(kj, "", "  "); err2 == nil {
						_ = os.WriteFile(keyFile, out, 0600)
					}
				}
				log.Printf("[NODE] loaded identity from %s", keyFile)
				return id
			}
		}
	}
	id, err := identity.Generate()
	if err != nil {
		log.Fatalf("generate identity: %v", err)
	}
	kj := keyJSON{
		PubKey:     id.PubKeyHex(),
		PrivKey:    id.PrivKeyHex(),
		X25519Pub:  id.X25519PubHex(),
		X25519Priv: id.X25519PrivHex(),
	}
	data, _ := json.MarshalIndent(kj, "", "  ")
	if err := os.WriteFile(keyFile, data, 0600); err != nil {
		log.Printf("[NODE] warning: could not save key: %v", err)
	} else {
		log.Printf("[NODE] new identity saved to %s", keyFile)
	}
	return id
}

func tipIndex(c *blockchain.Chain) uint64 {
	if tip := c.Tip(); tip != nil {
		return tip.Index
	}
	return 0
}

func tipHash(c *blockchain.Chain) string {
	if tip := c.Tip(); tip != nil {
		return tip.HashHex()[:12]
	}
	return "(empty)"
}

// autoRegisterRelay submits a REGISTER_RELAY tx and retries until the registration
// is confirmed on-chain. This handles the common case where the initial gossip has
// no peers (node2 hasn't started yet) or the leader hasn't included the tx yet.
func autoRegisterRelay(
	ctx context.Context,
	id *identity.Identity,
	relayKP *relay.KeyPair,
	feePerMsgUT uint64,
	addrs []string,
	chain *blockchain.Chain,
	engine *consensus.Engine,
	h *p2p.Host,
) {
	var multiaddr string
	if len(addrs) > 0 {
		multiaddr = addrs[0]
	}
	payload := blockchain.RegisterRelayPayload{
		X25519PubKey: relayKP.PubHex(),
		FeePerMsgUT:  feePerMsgUT,
		Multiaddr:    multiaddr,
	}
	pb, _ := json.Marshal(payload)

	ticker := time.NewTicker(10 * time.Second)
	defer ticker.Stop()

	// First attempt after 5 seconds; subsequent retries every 10 seconds.
	select {
	case <-ctx.Done():
		return
	case <-time.After(5 * time.Second):
	}

	for {
		// Stop if already confirmed on-chain.
		if relays, err := chain.RegisteredRelays(); err == nil {
			for _, r := range relays {
				if r.PubKey == id.PubKeyHex() && r.Relay.X25519PubKey == relayKP.PubHex() {
					log.Printf("[NODE] relay registration confirmed on-chain (x25519=%s)", relayKP.PubHex()[:16])
					return
				}
			}
		}

		// Build and sign a fresh tx each attempt (unique ID, fresh timestamp).
		now := time.Now().UTC()
		tx := &blockchain.Transaction{
			ID:        fmt.Sprintf("relay-reg-%d", now.UnixNano()),
			Type:      blockchain.EventRegisterRelay,
			From:      id.PubKeyHex(),
			// Pay the standard MinFee. The original code set this to 0 with the
			// note "REGISTER_RELAY is free"; that's incompatible with validateTx
			// in consensus/pbft.go which rejects any tx below MinFee, causing
			// the auto-register retry loop to spam the logs forever without
			// ever getting the tx into the mempool.
			Fee:       blockchain.MinFee,
			Payload:   pb,
			Memo:      "Register relay service",
			Timestamp: now,
		}
		tx.Signature = id.Sign(identity.TxSignBytes(tx))

		if err := engine.AddTransaction(tx); err != nil {
			log.Printf("[NODE] REGISTER_RELAY tx rejected: %v", err)
		} else if err := h.PublishTx(tx); err != nil {
			log.Printf("[NODE] REGISTER_RELAY publish failed: %v", err)
		} else {
			log.Printf("[NODE] submitted REGISTER_RELAY (x25519=%s fee=%dµT), waiting for commit...",
				relayKP.PubHex()[:16], feePerMsgUT)
		}

		select {
		case <-ctx.Done():
			return
		case <-ticker.C:
		}
	}
}

func shortKeys(keys []string) []string {
	out := make([]string, len(keys))
	for i, k := range keys {
		if len(k) > 8 {
			out[i] = k[:8] + "…"
		} else {
			out[i] = k
		}
	}
	return out
}

// setupLogging initialises the slog handler and routes Go's std log through
// it so existing log.Printf calls get the chosen format without a
// file-by-file migration.
//
// "text" (default) is handler-default human-readable format, same as bare
// log.Printf. "json" emits one JSON object per line with `time/level/msg`
// + any key=value attrs — what Loki/ELK ingest natively.
func setupLogging(format string) {
	var handler slog.Handler
	switch strings.ToLower(format) {
	case "json":
		handler = slog.NewJSONHandler(os.Stderr, &slog.HandlerOptions{
			Level: slog.LevelInfo,
		})
	default:
		handler = slog.NewTextHandler(os.Stderr, &slog.HandlerOptions{
			Level: slog.LevelInfo,
		})
	}
	slog.SetDefault(slog.New(handler))

	// Reroute Go's standard log package through slog so `log.Printf("…")`
	// calls elsewhere in the codebase land in the same stream/format.
	log.SetFlags(0)
	log.SetOutput(slogWriter{slog.Default()})
}

// slogWriter adapts an io.Writer-style byte stream onto slog. Each line
// becomes one Info-level record. The whole line is stored in `msg` — we
// don't try to parse out attrs; that can happen at callsite by switching
// from log.Printf to slog.Info.
type slogWriter struct{ l *slog.Logger }

func (w slogWriter) Write(p []byte) (int, error) {
	msg := strings.TrimRight(string(p), "\r\n")
	w.l.Info(msg)
	return len(p), nil
}

// ─── Env-var helpers for flag defaults ──────────────────────────────────────
// Pattern: `flag.String(name, envOr("DCHAIN_X", default), help)` lets the
// same binary be driven by either CLI flags (dev) or env-vars (Docker) with
// CLI winning. Flag.Parse() is called once; by that time os.Getenv has
// already populated the fallback via these helpers.

func envOr(key, def string) string {
	if v := os.Getenv(key); v != "" {
		return v
	}
	return def
}

func envBoolOr(key string, def bool) bool {
	v := strings.ToLower(strings.TrimSpace(os.Getenv(key)))
	switch v {
	case "":
		return def
	case "1", "true", "yes", "on":
		return true
	case "0", "false", "no", "off":
		return false
	default:
		log.Printf("[NODE] warn: env %s=%q is not a valid bool, using default %v", key, v, def)
		return def
	}
}

func envUint64Or(key string, def uint64) uint64 {
	v := os.Getenv(key)
	if v == "" {
		return def
	}
	var out uint64
	if _, err := fmt.Sscanf(v, "%d", &out); err != nil {
		log.Printf("[NODE] warn: env %s=%q is not a valid uint64, using default %d", key, v, def)
		return def
	}
	return out
}

// ─── Multi-seed --join support ──────────────────────────────────────────────

// persistedSeeds is the on-disk shape of <db>/seeds.json. Kept minimal so
// format churn doesn't break old files; unknown fields are ignored on read.
type persistedSeeds struct {
	ChainID     string   `json:"chain_id,omitempty"`
	GenesisHash string   `json:"genesis_hash,omitempty"`
	URLs        []string `json:"urls"`
	Peers       string   `json:"peers,omitempty"`     // comma-separated multiaddrs
	SavedAt     string   `json:"saved_at,omitempty"`
}

// parseSeedList splits a comma-separated --join value into trimmed, non-empty
// URL entries. Preserves order so the first is tried first.
func parseSeedList(raw string) []string {
	if raw == "" {
		return nil
	}
	var out []string
	for _, s := range strings.Split(raw, ",") {
		if t := strings.TrimSpace(s); t != "" {
			out = append(out, t)
		}
	}
	return out
}

// tryJoinSeeds walks the list and returns the first seedNetworkInfo that
// responded. The matching URL is returned alongside so the caller can
// persist it as the known-good at the head of the list.
func tryJoinSeeds(urls []string) (*seedNetworkInfo, string) {
	for _, u := range urls {
		info, err := fetchNetworkInfo(u)
		if err != nil {
			log.Printf("[NODE] --join: %s failed: %v (trying next)", u, err)
			continue
		}
		return info, u
	}
	return nil, ""
}

func loadSeedsFile(path string) (*persistedSeeds, error) {
	b, err := os.ReadFile(path)
	if err != nil {
		return nil, err
	}
	var p persistedSeeds
	if err := json.Unmarshal(b, &p); err != nil {
		return nil, err
	}
	return &p, nil
}

func saveSeedsFile(path string, p *persistedSeeds) error {
	if err := os.MkdirAll(filepath.Dir(path), 0o755); err != nil {
		return err
	}
	b, err := json.MarshalIndent(p, "", "  ")
	if err != nil {
		return err
	}
	return os.WriteFile(path, b, 0o600) // 0600: may contain private IPs of peers
}

// seedNetworkInfo is a minimal subset of /api/network-info needed for
// bootstrap. We intentionally DO NOT import the node package's response
// types here to keep the struct stable across schema evolution; missing
// fields simply unmarshal to zero values.
type seedNetworkInfo struct {
	ChainID     string `json:"chain_id"`
	GenesisHash string `json:"genesis_hash"`
	TipHeight   uint64 `json:"tip_height"`
	Validators  []string `json:"validators"`
	Peers       []struct {
		ID    string   `json:"id"`
		Addrs []string `json:"addrs"`
	} `json:"peers"`
	Contracts map[string]struct {
		ContractID string `json:"contract_id"`
		Name       string `json:"name"`
		Version    string `json:"version"`
	} `json:"contracts"`
}

// fetchNetworkInfo performs a GET /api/network-info on the given HTTP base
// URL (e.g. "http://seed.example:8080") and returns a parsed seedNetworkInfo.
// Times out after 10s so a misconfigured --join URL fails fast instead of
// hanging node startup.
func fetchNetworkInfo(baseURL string) (*seedNetworkInfo, error) {
	baseURL = strings.TrimRight(baseURL, "/")
	client := &http.Client{Timeout: 10 * time.Second}
	resp, err := client.Get(baseURL + "/api/network-info")
	if err != nil {
		return nil, fmt.Errorf("fetch: %w", err)
	}
	defer resp.Body.Close()
	if resp.StatusCode != 200 {
		return nil, fmt.Errorf("status %d", resp.StatusCode)
	}
	var info seedNetworkInfo
	if err := json.NewDecoder(resp.Body).Decode(&info); err != nil {
		return nil, fmt.Errorf("decode: %w", err)
	}
	return &info, nil
}