feat(solvers): liveness heartbeats with automatic offline detection

.env.example

@@ -429,6 +429,16 @@ WS_DRAIN_TIMEOUT_MS=25000
 # Empty disables JWT auth on the WS gateway.
 AUTH_JWT_SECRET=
 
+# ─── Solver liveness heartbeats (issue #445) ─────────────────────────────────
+# Cadence clients heartbeat at (ms); a solver is auto-marked offline after
+# interval × misses without a beat (default 10 000 × 3 = 30 s).
+SOLVER_HEARTBEAT_INTERVAL_MS=10000
+SOLVER_HEARTBEAT_MISSES=3
+# Shared liveness store URL (defaults to REDIS_URL). Leave both empty for a
+# process-local store — single replica only; multi-replica deployments must
+# share liveness or each replica only sees its own heartbeats.
+SOLVER_HEARTBEAT_REDIS_URL=
+
 # ─── API keys & distributed rate limiting (issue #441) ─────────────────────────
 # How often (ms) the local rate-limiter fallback prunes expired window entries.
 # Only relevant during a Redis outage; keeps the fallback map bounded.
@@ -549,6 +559,26 @@ DATASETS_STORAGE_KIND=memory
 DATASETS_LOCAL_DIR=./data/datasets
 
 
+# ─── Solver reputation (issue #444) ──────────────────────────────────────────
+# Weights for each reputation sub-component. Sum must be exactly 1.0 in production;
+# dev/test renormalises any valid positive set so local experimentation doesn't
+# prevent boot.
+REP_WEIGHT_FILL_RATE=0.35
+REP_WEIGHT_LATENCY=0.15
+REP_WEIGHT_SLASHES=0.25
+REP_WEIGHT_QUOTE_HONOUR=0.15
+REP_WEIGHT_VOLUME=0.10
+# Shared exponential-decay half-life (seconds) for all events.
+# Default: 30 days — scores dominated by the last ~1 month of activity.
+REP_DECAY_HALFLIFE_SECONDS=2592000
+# Beta-distribution priors for the Bayesian cold-start lower bound on fill rate.
+# (α=4, β=1) ≈ prior mean 80%, 2.5% lower bound ~37%.
+REP_BAYES_ALPHA=4
+REP_BAYES_BETA=1
+# Volume-component knee in USD-equivalent notional; $100k default.
+REP_VOLUME_LAMBDA_USD=100000
+# How many trailing days of daily snapshots the /reputation endpoint exposes (1..365).
+REP_HISTORY_WINDOW_DAYS=30
 # ─── Read replicas (#411) ─────────────────────────────────────────────────────
 # Comma-separated read-replica Postgres connection strings.
 # Leave blank to route all reads to the primary.

.env.mainnet.example

@@ -305,6 +305,16 @@ WS_DRAIN_TIMEOUT_MS=25000
 # Empty disables JWT auth on the WS gateway.
 AUTH_JWT_SECRET=
 
+# ─── Solver liveness heartbeats (issue #445) ─────────────────────────────────
+# Cadence clients heartbeat at (ms); a solver is auto-marked offline after
+# interval × misses without a beat (default 10 000 × 3 = 30 s).
+SOLVER_HEARTBEAT_INTERVAL_MS=10000
+SOLVER_HEARTBEAT_MISSES=3
+# Shared liveness store URL (defaults to REDIS_URL). Leave both empty for a
+# process-local store — single replica only; multi-replica deployments must
+# share liveness or each replica only sees its own heartbeats.
+SOLVER_HEARTBEAT_REDIS_URL=
+
 # ─── API keys & distributed rate limiting (issue #441) ─────────────────────────
 RATE_LIMIT_LOCAL_PRUNE_MS=60000
 # Redis URL for the shared rate-limit window. Empty = bounded local limiter.

.env.staging.example

@@ -192,6 +192,16 @@ WS_DRAIN_TIMEOUT_MS=25000
 # Empty disables JWT auth on the WS gateway.
 AUTH_JWT_SECRET=
 
+# ─── Solver liveness heartbeats (issue #445) ─────────────────────────────────
+# Cadence clients heartbeat at (ms); a solver is auto-marked offline after
+# interval × misses without a beat (default 10 000 × 3 = 30 s).
+SOLVER_HEARTBEAT_INTERVAL_MS=10000
+SOLVER_HEARTBEAT_MISSES=3
+# Shared liveness store URL (defaults to REDIS_URL). Leave both empty for a
+# process-local store — single replica only; multi-replica deployments must
+# share liveness or each replica only sees its own heartbeats.
+SOLVER_HEARTBEAT_REDIS_URL=
+
 # ─── API keys & distributed rate limiting (issue #441) ─────────────────────────
 RATE_LIMIT_LOCAL_PRUNE_MS=60000
 # Redis URL for the shared rate-limit window. Empty = bounded local limiter.

.env.testnet.example

@@ -281,6 +281,16 @@ DATASETS_PUBLIC_BUCKET=
 DATASETS_STORAGE_KIND=memory
 DATASETS_LOCAL_DIR=
 
+# ─── Solver liveness heartbeats (issue #445) ─────────────────────────────────
+# Cadence clients heartbeat at (ms); a solver is auto-marked offline after
+# interval × misses without a beat (default 10 000 × 3 = 30 s).
+SOLVER_HEARTBEAT_INTERVAL_MS=10000
+SOLVER_HEARTBEAT_MISSES=3
+# Shared liveness store URL (defaults to REDIS_URL). Leave both empty for a
+# process-local store — single replica only; multi-replica deployments must
+# share liveness or each replica only sees its own heartbeats.
+SOLVER_HEARTBEAT_REDIS_URL=
+
 # ─── Health probes (issue #492) ──────────────────────────────────────────────
 # Roles served by this process (api, ws, worker); readiness checks follow them.
 SERVICE_ROLES=api,ws,worker

.eslintrc.json

@@ -52,7 +52,7 @@
   "ignorePatterns": ["dist", "node_modules"],
   "overrides": [
     {
-      "files": ["scripts/**/*.ts", "tools/**/*.ts"],
+      "files": ["scripts/**/*.ts", "tools/**/*.ts", "test/chaos/**/*.ts"],
       "rules": {
         "no-restricted-syntax": "off"
       }