loki-mode 11.3.0 → 11.3.1

package/autonomy/run.sh

@@ -7578,12 +7578,13 @@ _write_pricing_json() {
   "updated": "${updated}",
   "source": "static",
   "models": {
-    "fable":           {"input": 10.00, "output": 50.00, "label": "Fable 5 (top, 2x Opus)", "provider": "claude"},
-    "claude-fable-5":  {"input": 10.00, "output": 50.00, "label": "Fable 5 (top, 2x Opus)", "provider": "claude"},
-    "opus":            {"input": 5.00,  "output": 25.00, "label": "Opus (latest)",   "provider": "claude"},
-    "sonnet":          {"input": 2.00,  "output": 10.00, "label": "Sonnet (latest)", "provider": "claude"},
-    "haiku":           {"input": 1.00,  "output": 5.00,  "label": "Haiku (latest)",  "provider": "claude"},
-    "claude-haiku-5-5": {"input": 0.10, "output": 0.50, "over_100k": {"input": 0.50, "output": 2.50}, "label": "Haiku 5.5", "provider": "claude"},
+    "fable":           {"input": 10.00, "output": 50.00, "cache_read": 0.25, "cache_write": 12.5, "cache_write_1h": 20, "label": "Fable 5 (top tier)", "provider": "claude"},
+    "claude-fable-5":  {"input": 10.00, "output": 50.00, "cache_read": 0.25, "cache_write": 12.5, "cache_write_1h": 20, "label": "Fable 5 (top tier)", "provider": "claude"},
+    "opus":            {"input": 4.00,  "output": 20.00, "cache_read": 0.2, "cache_write": 5, "cache_write_1h": 8, "label": "Opus (latest)",   "provider": "claude"},
+    "sonnet":          {"input": 2.00,  "output": 10.00, "cache_read": 0.1, "cache_write": 2.5, "cache_write_1h": 4, "label": "Sonnet (latest)", "provider": "claude"},
+    "haiku":           {"input": 0.10,  "output": 0.50, "cache_read": 0.01, "cache_write": 0.125, "cache_write_1h": 0.2, "over_100k": {"input": 0.50, "output": 2.50}, "label": "Haiku 5.5 (latest)",  "provider": "claude"},
+    "claude-haiku-4-5": {"input": 1.00, "output": 5.00, "cache_read": 0.1, "cache_write": 1.25, "cache_write_1h": 2, "label": "Haiku 4.5", "provider": "claude"},
+    "claude-haiku-5-5": {"input": 0.10, "output": 0.50, "cache_read": 0.01, "cache_write": 0.125, "cache_write_1h": 0.2, "over_100k": {"input": 0.50, "output": 2.50}, "label": "Haiku 5.5", "provider": "claude"},
     "gpt-5.3-codex":   {"input": 1.75,  "output": 14.00, "label": "GPT-5.3 Codex", "provider": "codex"}
   }
 }
@@ -16425,7 +16426,7 @@ _dispatch_reviewer() {
                         _crs_out="$(_loki_with_deadline "$_crs_cap" \
                             "${_cr_provider_env[@]}" "$_crs_loki" internal sdk-judge \
                             --prompt-file "$_crs_pf" --schema-file "$_crs_schema_file" \
-                            --model "${LOKI_SDK_REVIEW_MODEL:-claude-sonnet-5}" --effort high \
+                            --model "${LOKI_SDK_REVIEW_MODEL:-sonnet}" --effort high \
                             --timeout-ms "$(( _crs_to_s * 1000 ))")" || _crs_rc=$?
                         _crs_invoked=true
                         rm -f "$_crs_pf" 2>/dev/null || true
@@ -20248,15 +20249,20 @@ check_budget_limit() {
 import json, glob
 total = 0.0
 pricing = {
-    'fable': {'input': 10.00, 'output': 50.00},
-    'claude-fable-5': {'input': 10.00, 'output': 50.00},
-    'opus': {'input': 5.00, 'output': 25.00},
-    'sonnet': {'input': 2.00, 'output': 10.00},
-    'haiku': {'input': 1.00, 'output': 5.00},
+    'fable': {'input': 10.00, 'output': 50.00, 'cache_read': 0.25, 'cache_write': 12.5, 'cache_write_1h': 20},
+    'claude-fable-5': {'input': 10.00, 'output': 50.00, 'cache_read': 0.25, 'cache_write': 12.5, 'cache_write_1h': 20},
+    'opus': {'input': 4.00, 'output': 20.00, 'cache_read': 0.2, 'cache_write': 5, 'cache_write_1h': 8},
+    'sonnet': {'input': 2.00, 'output': 10.00, 'cache_read': 0.1, 'cache_write': 2.5, 'cache_write_1h': 4},
+    # haiku = Haiku 5.5 up to 100K prompt tokens (no over-100K tier in this gate)
+    'haiku': {'input': 0.10, 'output': 0.50, 'cache_read': 0.01, 'cache_write': 0.125, 'cache_write_1h': 0.2},
+    # Exact ids: this dict is keyed by the recorded model string with a sonnet
+    # fallback, so an id missing here is priced at sonnet (20x for Haiku 5.5).
+    'claude-haiku-4-5': {'input': 1.00, 'output': 5.00, 'cache_read': 0.1, 'cache_write': 1.25, 'cache_write_1h': 2},
+    'claude-haiku-5-5': {'input': 0.10, 'output': 0.50, 'cache_read': 0.01, 'cache_write': 0.125, 'cache_write_1h': 0.2},
     'gpt-5.3-codex': {'input': 1.75, 'output': 14.00},
-    'gpt-5.6-sol': {'input': 2.50, 'output': 20.00},
-    'gpt-5.6-terra': {'input': 1.50, 'output': 12.00},
-    'gpt-5.6-luna': {'input': 0.50, 'output': 4.00},
+    'gpt-5.6-sol': {'input': 4.00, 'output': 20.00},
+    'gpt-5.6-terra': {'input': 2.00, 'output': 12.00},
+    'gpt-5.6-luna': {'input': 0.20, 'output': 1.20},
 }
 for f in glob.glob('${efficiency_dir}/*.json'):
     try:
@@ -20273,8 +20279,9 @@ for f in glob.glob('${efficiency_dir}/*.json'):
             # DOMINATE real traffic: a measured iteration carried 797,496
             # cache-read tokens against 10,272 of plain input. Pricing them at
             # zero under-counted a real iteration 5.4x, so a breaker set to
-            # stop a runaway let it run far past the cap. Published multipliers:
-            # cache read 0.1x input, cache write 1.25x input.
+            # stop a runaway let it run far past the cap. Rates come from the
+            # table row (cache read is 0.05x on Opus 5.5 / Sonnet 5.5, 0.025x
+            # on Fable 5.1, 0.1x elsewhere); rows without them use 0.1x / 1.25x.
             #
             # This mirrors the TS route's calculateCostFromRecords
             # (loki-ts/src/runner/budget.ts). Both routes must agree or the
@@ -20282,7 +20289,7 @@ for f in glob.glob('${efficiency_dir}/*.json'):
             cr = d.get('cache_read_tokens', 0) or 0
             cw = d.get('cache_creation_tokens', 0) or 0
             total += (inp / 1_000_000) * p['input'] + (out / 1_000_000) * p['output']
-            total += (cr / 1_000_000) * (p['input'] * 0.1) + (cw / 1_000_000) * (p['input'] * 1.25)
+            total += (cr / 1_000_000) * p.get('cache_read', p['input'] * 0.1) + (cw / 1_000_000) * p.get('cache_write', p['input'] * 1.25)
     except: pass
 print(round(total, 4))
 " 2>/dev/null || echo "0")
@@ -28817,6 +28824,11 @@ main() {
     REMAINING_ARGS=()
     while [[ $# -gt 0 ]]; do
         case "$1" in
+            --attempts|--attempts=*)
+                # T5-D2: run.sh is the legacy loop; attempts run only on the engine10 path.
+                echo "run.sh: --attempts is only supported by 'loki start --attempts N --no-pr' (the Loki 10 engine), not the legacy loop." >&2
+                exit 2
+                ;;
             --parallel)
                 PARALLEL_MODE=true
                 shift

package/autonomy/sandbox.sh

@@ -279,7 +279,7 @@ check_docker() {
 }
 
 build_sandbox_image() {
-    local dockerfile="$SKILL_DIR/Dockerfile.sandbox"
+    local dockerfile="$SKILL_DIR/docker/Dockerfile.sandbox"
 
     if [[ ! -f "$dockerfile" ]]; then
         log_error "Sandbox Dockerfile not found at $dockerfile"

package/autonomy/tui.sh

@@ -390,16 +390,17 @@ display_token_cost() {
     local input_tokens="$2"
     local output_tokens="$3"
 
-    # Pricing per 1M tokens (March 2026 rates)
+    # Pricing per 1M tokens (2026-10-08 rates, loki-ts/data/model-pricing.json)
     local input_rate=0
     local output_rate=0
     case "$model" in
-        *opus*)   input_rate=15.0;  output_rate=75.0 ;;
-        *sonnet*) input_rate=3.0;   output_rate=15.0 ;;
-        *haiku*)  input_rate=0.25;  output_rate=1.25 ;;
+        *opus*)   input_rate=4.0;   output_rate=20.0 ;;
+        *sonnet*) input_rate=2.0;   output_rate=10.0 ;;
+        # Haiku 5.5 up to 100K prompt tokens (no over-100K tier here)
+        *haiku*)  input_rate=0.10;  output_rate=0.50 ;;
         *gpt-4o-mini*) input_rate=0.15; output_rate=0.6 ;;
         *gpt-4o*) input_rate=2.5;   output_rate=10.0 ;;
-        *)        input_rate=3.0;   output_rate=15.0 ;;  # default sonnet-like
+        *)        input_rate=2.0;   output_rate=10.0 ;;  # default sonnet-like
     esac
 
     local input_cost

package/autonomy/verify.sh

@@ -207,7 +207,7 @@ _verify_llm_review() {
 }
 SCHEMA
 
-    model="${LOKI_VERIFY_LLM_MODEL:-claude-sonnet-5}"
+    model="${LOKI_VERIFY_LLM_MODEL:-sonnet}"
     local to_s="${LOKI_VERIFY_LLM_TIMEOUT_S:-120}"
     local wrap=""
     if command -v timeout >/dev/null 2>&1; then wrap="timeout $(( to_s + 15 ))"

package/bin/loki

@@ -32,6 +32,39 @@ SCRIPT_DIR=$(dirname "$SCRIPT_PATH")
 REPO_ROOT=$(cd "$SCRIPT_DIR/.." 2>/dev/null && pwd)
 BASH_CLI="$REPO_ROOT/autonomy/loki"
 
+# FC-38 (scoped): `loki start --attempts` never reaches the legacy bash loop. Every other
+# start, and every other command, keeps the plain bash fallback.
+_LOKI_START_ATTEMPTS=0
+if [ "${1:-}" = "start" ]; then
+    for _loki_a0 in "${@:2}"; do
+        case "$_loki_a0" in --attempts|--attempts=*) _LOKI_START_ATTEMPTS=1 ;; esac
+    done
+    unset _loki_a0
+fi
+_loki_bash_or_refuse() {
+    if [ "${1:-}" = "start" ] && [ "$_LOKI_START_ATTEMPTS" = "1" ]; then
+        echo "loki: 'loki start' runs only the Loki 10 engine, which needs the Bun entry (loki-ts). It cannot fall back to the removed legacy loop; reinstall loki-mode or fix LOKI_TS_ENTRY / BUN_FROM_SOURCE." >&2
+        exit 1
+    fi
+    exec "$BASH_CLI" "$@"
+}
+
+# T5-D2 (FC-38 sweep): quick, run, demo and tour have no attempts engine; --attempts there would
+# reach the legacy loop, so it is refused before any route split.
+case "${1:-}" in
+    quick|run|demo|tour)
+        for _loki_a1 in "${@:2}"; do
+            case "$_loki_a1" in
+                --attempts|--attempts=*)
+                    echo "loki: --attempts is only supported by 'loki start --attempts N --no-pr' (the Loki 10 engine). 'loki $1' would run the removed legacy loop; drop --attempts or use 'loki start --attempts N --no-pr'." >&2
+                    exit 2
+                    ;;
+            esac
+        done
+        unset _loki_a1
+        ;;
+esac
+
 # v8.1: resolve the one-switch SDK mode (LOKI_SDK_MODE=off|judges|full) into the
 # per-site LOKI_SDK_* flags HERE, in the outermost shim, BEFORE the `start`
 # routing fork below reads LOKI_SDK_LOOP. Without this, `LOKI_SDK_MODE=full loki
@@ -78,7 +111,7 @@ if [ -n "${LOKI_TS_ENTRY:-}" ]; then
     # invoke commands while they fix the path.
     if [ ! -f "$LOKI_TS_ENTRY" ]; then
         echo "ERROR: LOKI_TS_ENTRY=$LOKI_TS_ENTRY does not exist; falling through to bash CLI. Unset the variable or fix the path." >&2
-        exec "$BASH_CLI" "$@"
+        _loki_bash_or_refuse "$@"
     fi
     BUN_CLI="$LOKI_TS_ENTRY"
 elif [ "${BUN_FROM_SOURCE:-0}" = "1" ] || [ "${BUN_FROM_SOURCE:-}" = "true" ]; then
@@ -89,7 +122,7 @@ elif [ "${BUN_FROM_SOURCE:-0}" = "1" ] || [ "${BUN_FROM_SOURCE:-}" = "true" ]; t
         BUN_CLI="$REPO_ROOT/loki-ts/dist/loki.js"
     else
         echo "ERROR: BUN_FROM_SOURCE set but neither src/cli.ts nor dist/loki.js found; falling through to bash" >&2
-        exec "$BASH_CLI" "$@"
+        _loki_bash_or_refuse "$@"
     fi
 elif [ -f "$REPO_ROOT/loki-ts/dist/loki.js" ]; then
     BUN_CLI="$REPO_ROOT/loki-ts/dist/loki.js"
@@ -112,7 +145,7 @@ elif [ -f "$REPO_ROOT/loki-ts/src/cli.ts" ]; then
     BUN_CLI="$REPO_ROOT/loki-ts/src/cli.ts"
 else
     # Neither dist nor src available; fall through to bash silently.
-    exec "$BASH_CLI" "$@"
+    _loki_bash_or_refuse "$@"
 fi
 
 # FC-30: an install artifact must not outlive the install that made it. If a provider
@@ -259,7 +292,7 @@ esac
 
 # Force-fall-through to bash when the rollback flag is set.
 if [ "${LOKI_LEGACY_BASH:-0}" = "1" ] || [ "${LOKI_LEGACY_BASH:-}" = "true" ]; then
-    exec "$BASH_CLI" "$@"
+    _loki_bash_or_refuse "$@"
 fi
 
 # Loki 10 bun resolver: bun on PATH, then an npm-installed @oven/bun-* optional
@@ -335,6 +368,105 @@ case "${1:-}" in
         ;;
 esac
 
+_loki_start_needs_bash() {
+    # $@ = the start args (already know $1 = start). Return 0 (divert to bash) if
+    # any bash-only orchestration flag or an issue-ref spec is present.
+    local a
+    local explicit_provider=""
+    local provider_value_next=0
+    for a in "$@"; do
+        if [ "$provider_value_next" = "1" ]; then
+            explicit_provider="$a"
+            provider_value_next=0
+            continue
+        fi
+        case "$a" in
+            --provider)
+                provider_value_next=1 ;;
+            --provider=*)
+                explicit_provider="${a#--provider=}" ;;
+            --parallel|--bg|--background|--github|--api|--sandbox \
+            |--fast|--repo-fast|--full|--repo-full \
+            |--bmad-project|--openspec \
+            |--mirofish|--mirofish-docker|--mirofish-rounds|--mirofish-timeout|--mirofish-bg \
+            |--issue|--dry-run|--no-start|--output|--worktree|-w|--pr|--ship|--detach|-d)
+                return 0 ;;
+            --isolation)
+                # value follows; only worktree/docker/sandbox divert (none stays Bun).
+                _loki_iso_next=1 ;;
+            worktree|wt|docker|sandbox)
+                [ "${_loki_iso_next:-0}" = "1" ] && { unset _loki_iso_next; return 0; }
+                _loki_iso_next=0 ;;
+            *)
+                # issue-ref positional: owner/repo#N, a tracker URL, or *#<digits>.
+                case "$a" in
+                    */*\#[0-9]*|http*://*/issues/*|http*://*/browse/*) return 0 ;;
+                esac
+                _loki_iso_next=0 ;;
+        esac
+    done
+    unset _loki_iso_next
+
+    # The Bun RARV loop has not ported the opencode invoker yet. Keep the
+    # released, autonomous shell adapter reachable when SDK mode is enabled
+    # instead of accepting the selection and then failing after intake. CLI
+    # selection wins over env, which wins over the project-level saved choice.
+    local effective_provider="$explicit_provider"
+    if [ -z "$effective_provider" ] && [ -n "${LOKI_PROVIDER:-}" ]; then
+        effective_provider="$LOKI_PROVIDER"
+    fi
+    if [ -z "$effective_provider" ] && [ -f "${PWD}/.loki/state/provider" ]; then
+        effective_provider="$(sed -n '1p' "${PWD}/.loki/state/provider" 2>/dev/null || true)"
+    fi
+    [ "$effective_provider" = "opencode" ] && return 0
+    return 1
+}
+_loki_attempts_refuse() {
+    if _loki_start_needs_bash "$@"; then
+        echo "loki: 'loki start --attempts' runs only the Loki 10 engine; an option or provider in this command (for example --parallel, --github, --sandbox, --issue, --dry-run, --detach, or opencode) was only honored by the removed legacy loop. Remove it, or use a task string, an issue ref or a PRD path." >&2
+        exit 2
+    fi
+    case "${LOKI_PROVIDER:-claude}" in
+        claude|codex|cline|aider) ;;
+        *) echo "loki: the Loki 10 engine does not support LOKI_PROVIDER=${LOKI_PROVIDER:-}; use claude, codex, cline or aider." >&2; exit 2 ;;
+    esac
+}
+
+# FC-37: --attempts is handled HERE, once, before any route split (engine10 positional task, Bun
+# loop, bash cmd_start). A flag accepted on one route must be honored or refused loudly on every
+# other. N=1 is stripped so every route sees today's argv byte-for-byte; N=2..5 runs on the Bun
+# start command (loki-ts/src/commands/start.ts), which spawns one engine10 run per worktree.
+if [ "${1:-}" = "start" ]; then
+    _loki_att_args=()
+    _loki_att_n=""
+    _loki_att_skip=0
+    for _loki_att_a in "${@:2}"; do
+        if [ "$_loki_att_skip" = "1" ]; then _loki_att_n="$_loki_att_a"; _loki_att_skip=0; continue; fi
+        case "$_loki_att_a" in
+            --attempts) _loki_att_skip=1 ;;
+            --attempts=*) _loki_att_n="${_loki_att_a#--attempts=}" ;;
+            *) _loki_att_args+=("$_loki_att_a") ;;
+        esac
+    done
+    if [ "$_loki_att_skip" = "1" ]; then
+        echo "loki: --attempts requires a value from 1 to 5." >&2
+        exit 2
+    fi
+    if [ -n "$_loki_att_n" ]; then
+        case "$_loki_att_n" in
+            1) set -- start ${_loki_att_args[@]+"${_loki_att_args[@]}"} ;;
+            [2-5])
+                _lb=$(_loki_bun)
+                [ -n "$_lb" ] || { echo "loki: --attempts $_loki_att_n needs bun (install from https://bun.sh); refusing to run a single attempt silently." >&2; exit 1; }
+                _loki_attempts_refuse "${@:2}"
+                exec "$_lb" "$BUN_CLI" "$@"
+                ;;
+            *) echo "loki: --attempts must be an integer from 1 to 5, got '$_loki_att_n'." >&2; exit 2 ;;
+        esac
+    fi
+    unset _loki_att_args _loki_att_n _loki_att_skip _loki_att_a
+fi
+
 # Loki 10 engine (D29, the only engine from 11.0.0). Routes `loki "<task>"`,
 # `loki owner/repo#N`, `loki quick "<task>"`, `keys` and `verify --pubkey` to engine10.
 # shellcheck disable=SC2194
@@ -376,7 +508,23 @@ case "v10" in
             # `loki <ref>`; any other start call (PRD file, flags first, one word) stays legacy.
             start)
                 case "${2:-}" in
-                    */*\#[0-9]*|http*://*/issues/*|http*://*/browse/*|jira:*|linear:*|https://linear.app/*/issue/*|*" "*) _e10=1; _e10_args=("${@:2}") ;;
+                    */*\#[0-9]*|http*://*/issues/*|http*://*/browse/*|jira:*|linear:*|https://linear.app/*/issue/*|*" "*)
+                        _e10=1; _e10_args=("${@:2}")
+                        if [ "$_LOKI_START_ATTEMPTS" = "1" ]; then
+                            _e10_hard=1  # FC-38: --attempts never falls back to the legacy engine
+                            # FC-37: engine10 has no --budget; unmapped it would be glued onto the task text. Its per-run cap is --max-cost.
+                            _e10_args=(); _e10_bnext=0
+                            for _e10_a in "${@:2}"; do
+                                if [ "$_e10_bnext" = "1" ]; then _e10_args+=("$_e10_a"); _e10_bnext=0; continue; fi
+                                case "$_e10_a" in
+                                    --budget|--budget-limit) _e10_args+=(--max-cost); _e10_bnext=1 ;;
+                                    --budget=*|--budget-limit=*) _e10_args+=("--max-cost=${_e10_a#*=}") ;;
+                                    *) _e10_args+=("$_e10_a") ;;
+                                esac
+                            done
+                            unset _e10_a _e10_bnext
+                        fi
+                        ;;
                     *) _e10=0 ;;
                 esac
                 ;;
@@ -425,6 +573,10 @@ esac
 # Bun must be installed for the new path; if it isn't, fall through to bash
 # silently rather than fail. This keeps users on systems without Bun working.
 if ! command -v bun &>/dev/null; then
+    if [ "${1:-}" = "start" ] && [ "$_LOKI_START_ATTEMPTS" = "1" ]; then
+        echo "loki: 'loki start' runs the Loki 10 engine, which needs bun. Install bun from https://bun.sh. The legacy loop is not reachable from start." >&2
+        exit 1
+    fi
     exec "$BASH_CLI" "$@"
 fi
 
@@ -448,59 +600,14 @@ fi
 # guarantee that makes the eventual loop default-flip a no-regression. Runner
 # flags (--simple, --allow-haiku, --regen-prd, --skip-memory, provider knobs) are
 # handled by start.ts and do NOT divert. See artifacts/ITER1-FLAG-RECONCILE.md.
-_loki_start_needs_bash() {
-    # $@ = the start args (already know $1 = start). Return 0 (divert to bash) if
-    # any bash-only orchestration flag or an issue-ref spec is present.
-    local a
-    local explicit_provider=""
-    local provider_value_next=0
-    for a in "$@"; do
-        if [ "$provider_value_next" = "1" ]; then
-            explicit_provider="$a"
-            provider_value_next=0
-            continue
-        fi
-        case "$a" in
-            --provider)
-                provider_value_next=1 ;;
-            --provider=*)
-                explicit_provider="${a#--provider=}" ;;
-            --parallel|--bg|--background|--github|--api|--sandbox \
-            |--fast|--repo-fast|--full|--repo-full \
-            |--bmad-project|--openspec \
-            |--mirofish|--mirofish-docker|--mirofish-rounds|--mirofish-timeout|--mirofish-bg \
-            |--issue|--dry-run|--no-start|--output|--worktree|-w|--pr|--ship|--detach|-d)
-                return 0 ;;
-            --isolation)
-                # value follows; only worktree/docker/sandbox divert (none stays Bun).
-                _loki_iso_next=1 ;;
-            worktree|wt|docker|sandbox)
-                [ "${_loki_iso_next:-0}" = "1" ] && { unset _loki_iso_next; return 0; }
-                _loki_iso_next=0 ;;
-            *)
-                # issue-ref positional: owner/repo#N, a tracker URL, or *#<digits>.
-                case "$a" in
-                    */*\#[0-9]*|http*://*/issues/*|http*://*/browse/*) return 0 ;;
-                esac
-                _loki_iso_next=0 ;;
-        esac
-    done
-    unset _loki_iso_next
-
-    # The Bun RARV loop has not ported the opencode invoker yet. Keep the
-    # released, autonomous shell adapter reachable when SDK mode is enabled
-    # instead of accepting the selection and then failing after intake. CLI
-    # selection wins over env, which wins over the project-level saved choice.
-    local effective_provider="$explicit_provider"
-    if [ -z "$effective_provider" ] && [ -n "${LOKI_PROVIDER:-}" ]; then
-        effective_provider="$LOKI_PROVIDER"
-    fi
-    if [ -z "$effective_provider" ] && [ -f "${PWD}/.loki/state/provider" ]; then
-        effective_provider="$(sed -n '1p' "${PWD}/.loki/state/provider" 2>/dev/null || true)"
-    fi
-    [ "$effective_provider" = "opencode" ] && return 0
-    return 1
-}
+# FC-38 (scoped to --attempts; plain start keeps the 11.3.0 routing below): no `start --attempts` reaches the legacy loop (bash cmd_start, run.sh, runAutonomous).
+# Issue refs and multi-word tasks were sent to engine10 above; every other start (PRD path, brief,
+# flags) goes to the Bun start command, which runs engine10 and seals a receipt. A flag only the
+# legacy route honored, or a provider engine10 cannot invoke, is refused loudly instead.
+if [ "${1:-}" = "start" ] && [ "$_LOKI_START_ATTEMPTS" = "1" ]; then
+    _loki_attempts_refuse "${@:2}"
+    exec bun "$BUN_CLI" "$@"
+fi
 if [ "${1:-}" = "start" ] && { [ "${LOKI_SDK_LOOP:-}" = "1" ] || [ "${LOKI_SDK_LOOP:-}" = "true" ]; }; then
     if _loki_start_needs_bash "${@:2}"; then
         exec "$BASH_CLI" "$@"
@@ -603,7 +710,7 @@ fi
 # Two-token routes (provider show/list, memory list/index) match on the first
 # token only; the Bun dispatcher handles subcommand routing internally.
 case "${1:-}" in
-    version|--version|-v|status|stats|doctor|provider|memory|rollback|internal|kpis|trust|proof|receipt|wiki|crash|control|contract|slack|answer|completion|__complete)
+    version|--version|-v|status|stats|doctor|provider|memory|rollback|internal|kpis|trust|proof|receipt|wiki|crash|control|contract|slack|queue|answer|completion|__complete)
         # v7.5.2: rollback added (wires loki-ts/src/commands/rollback.ts).
         # v7.5.3: internal added for autonomy/run.sh phase1-hooks calls.
         # v7.5.28: kpis added (Phase K MVP: read-only KPI snapshot).

package/completions/_loki

@@ -209,6 +209,7 @@ function _loki_commands {
         'otel:OpenTelemetry management'
         'wiki:Wiki commands'
         'answer:Resume a BLOCKED run with an answer'
+        'queue:Overnight issue queue with a morning digest'
         'control:Control plane (preview, LOKI_CONTROL=1)'
         'workspace:Multi-project workspaces (preview, LOKI_WORKSPACES=1)'
         'backlog:Run every matching open issue, N in parallel'
@@ -264,6 +265,7 @@ function _loki_start {
         '-y[Skip confirmation prompts]' \
         '--budget[Cost budget limit in USD]:amount:' \
         '--max-iterations[Max iterations]:number:' \
+        '--attempts[Independent attempts in worktrees, 1-5]:number:(1 2 3 4 5)' \
         '*:spec file:_loki_spec_files'
 }
 

package/completions/loki.bash

@@ -5,7 +5,7 @@ _loki_completion() {
     _init_completion || return
 
     # Main subcommands (must match autonomy/loki main case statement)
-    local main_commands="start backlog quick monitor demo tour welcome init stop pause resume steer status next ship dashboard web serve api sandbox notify import github issue intent config provider reset memory compound checkpoint council dogfood projects enterprise secrets cockpit secure own handoff doctor watchdog audit metrics syslog onboard share proof receipt outcomes explain plan report cost estimate kpis stats test ci watch telemetry agent context ctx code run export review optimize heal modernize migrate cluster worktree wt trigger failover remote deploy docker mcp magic assets analyze compliance crash control workspace open otel preview quickstart rc rollback merge self-update sentrux setup-skill spec state template trust trust-metrics ultracode update verify voice why wiki bench cleanup logs grill docs cp version completions help keys answer acp"
+    local main_commands="start backlog quick monitor demo tour welcome init stop pause resume steer status next ship dashboard web serve api sandbox notify import github issue intent config provider reset memory compound checkpoint council dogfood projects enterprise secrets cockpit secure own handoff doctor watchdog audit metrics syslog onboard share proof receipt outcomes explain plan report cost estimate kpis stats test ci watch telemetry agent context ctx code run export review optimize heal modernize migrate cluster worktree wt trigger failover remote deploy docker mcp magic assets analyze compliance crash control workspace open otel preview quickstart rc rollback merge self-update sentrux setup-skill spec state template trust trust-metrics ultracode update verify voice why wiki bench cleanup logs grill docs cp version completions help keys answer queue acp"
 
     # 1. If we are on the first argument (subcommand)
     if [[ $cword -eq 1 ]]; then
@@ -27,7 +27,7 @@ _loki_completion() {
 
             # If the word starts with a dash, show flags
             if [[ "$cur" == -* ]]; then
-                local flags="--provider --max-iterations --parallel --background --bg --simple --complex --github --no-dashboard --sandbox --skip-memory --yes --budget --help"
+                local flags="--provider --max-iterations --parallel --background --bg --simple --complex --github --no-dashboard --sandbox --skip-memory --yes --budget --attempts --help"
                 COMPREPLY=( $(compgen -W "${flags}" -- "$cur") )
                 return 0
             fi

package/dashboard/__init__.py

@@ -7,7 +7,7 @@ Modules:
     control: Session control API (start/stop/pause/resume)
 """
 
-__version__ = "11.3.0"
+__version__ = "11.3.1"
 
 # Expose the control app for easy import
 try:

package/dashboard/server.py

@@ -1520,9 +1520,9 @@ async def get_provider_models() -> dict:
         "schema_version": 1,
         "providers": {
             "claude": {
-                "latest_planning": "claude-opus-4-7",
-                "latest_development": "claude-opus-4-7",
-                "latest_fast": "claude-sonnet-4-6",
+                "latest_planning": "opus",
+                "latest_development": "opus",
+                "latest_fast": "sonnet",
                 "models": [],
             }
         },
@@ -3360,7 +3360,7 @@ def _provider_model_offers(provider: str) -> list[dict]:
     Every other provider is offered the generic tiers (small/medium/high), which
     are provider-independent, each annotated with the concrete model id the
     catalog says that provider dispatches. That is what makes the picker read
-    "medium -> gpt-5.6-terra" on codex and "medium -> claude-sonnet-5" on claude
+    "medium -> gpt-5.6-terra" on codex and "medium -> <catalog sonnet id>" on claude
     without the frontend knowing a single model id.
     """
     if provider == "claude":
@@ -7738,23 +7738,24 @@ async def stop_session(request: Request):
 # At runtime, overridden by .loki/pricing.json if available
 _DEFAULT_PRICING = {
     # Claude (Anthropic)
-    # Fable 5 is the top-tier advisory model at exactly 2x Opus per token.
-    "fable":  {"input": 10.00, "output": 50.00},
-    "claude-fable-5": {"input": 10.00, "output": 50.00},
-    "opus":   {"input": 5.00, "output": 25.00},
-    "sonnet": {"input": 2.00, "output": 10.00},
-    "haiku":  {"input": 1.00, "output": 5.00},
+    # Fable 5 is the top-tier advisory model at 2.5x Opus 5.5 per token.
+    "fable":  {"input": 10.00, "output": 50.00, "cache_read": 0.25, "cache_write": 12.5, "cache_write_1h": 20},
+    "claude-fable-5": {"input": 10.00, "output": 50.00, "cache_read": 0.25, "cache_write": 12.5, "cache_write_1h": 20},
+    "opus":   {"input": 4.00, "output": 20.00, "cache_read": 0.2, "cache_write": 5, "cache_write_1h": 8},
+    "sonnet": {"input": 2.00, "output": 10.00, "cache_read": 0.1, "cache_write": 2.5, "cache_write_1h": 4},
+    # haiku = Haiku 5.5 up to 100K prompt tokens (no over-100K tier here)
+    "haiku":  {"input": 0.10, "output": 0.50, "cache_read": 0.01, "cache_write": 0.125, "cache_write_1h": 0.2},
+    # Exact ids (a lookup by recorded model string must not fall back to sonnet)
+    "claude-haiku-4-5": {"input": 1.00, "output": 5.00, "cache_read": 0.1, "cache_write": 1.25, "cache_write_1h": 2},
+    "claude-haiku-5-5": {"input": 0.10, "output": 0.50, "cache_read": 0.01, "cache_write": 0.125, "cache_write_1h": 0.2},
     # OpenAI Codex
-    "gpt-5.3-codex": {"input": 1.50, "output": 12.00},
+    "gpt-5.3-codex": {"input": 1.75, "output": 14.00},
     # gpt-5.6 line: sol (high) / terra (medium, default) / luna (small).
-    # UNVERIFIED RATES. The model IDs are confirmed against
-    # developers.openai.com/api/docs/models, but OpenAI's published per-token
-    # prices for this line were not, so these are placeholders scaled from the
-    # gpt-5.3 rate. They drive a display estimate only, never a gate. Replace
-    # from the pricing page; tools/probe-model-catalog.py is the refresh path.
-    "gpt-5.6-sol":   {"input": 2.50, "output": 20.00},
-    "gpt-5.6-terra": {"input": 1.50, "output": 12.00},
-    "gpt-5.6-luna":  {"input": 0.50, "output": 4.00},
+    # Rates from developers.openai.com/api/docs/pricing (read 2026-10-08),
+    # Standard tier, short context (up to 272K). tests/test-pricing-parity.sh pins them.
+    "gpt-5.6-sol":   {"input": 4.00, "output": 20.00},
+    "gpt-5.6-terra": {"input": 2.00, "output": 12.00},
+    "gpt-5.6-luna":  {"input": 0.20, "output": 1.20},
 }
 
 # Active pricing - starts with defaults, updated from .loki/pricing.json
@@ -8565,11 +8566,11 @@ async def get_gate_policy():
 # =============================================================================
 
 _PROVIDER_LABELS = {
-    # v7.104.0: current model IDs (model_catalog.json): opus=claude-opus-4-8,
-    # sonnet=claude-sonnet-5 (the default execution model), haiku=claude-haiku-4-5.
-    "opus": "Opus 4.8",
-    "sonnet": "Sonnet 5",
-    "haiku": "Haiku 4.5",
+    # v7.104.0: current model IDs (model_catalog.json): opus, sonnet (the default execution model), haiku: ids live in
+    # the catalog cli_aliases, not here.
+    "opus": "Opus 5.5",
+    "sonnet": "Sonnet 5.5",
+    "haiku": "Haiku 5.5",
     "gpt-5.3-codex": "GPT-5.3 Codex",
     "gpt-5.6-sol": "GPT-5.6 Sol",
     "gpt-5.6-terra": "GPT-5.6 Terra",
@@ -8580,18 +8581,11 @@ _PROVIDER_LABELS = {
 # the UI WITHOUT changing any computed cost number. The Sonnet 5 intro price
 # ($2/$10 per MTok) became the standard price, so the 2026-09-01 increase to
 # $3/$15 will not occur (platform.claude.com pricing page, read 2026-10-03).
-_UNVERIFIED_RATE_NOTE = (
-    "Unverified placeholder rate, scaled from gpt-5.3; not from OpenAI's pricing page."
-)
 _MODEL_PRICING_NOTES = {
     "sonnet": (
         "Standard pricing: $2 / $10 per MTok. "
         "The intro price became the standard price; no increase to $3 / $15."
     ),
-    # See the UNVERIFIED RATES comment on the pricing table: shown, not hidden.
-    "gpt-5.6-sol": _UNVERIFIED_RATE_NOTE,
-    "gpt-5.6-terra": _UNVERIFIED_RATE_NOTE,
-    "gpt-5.6-luna": _UNVERIFIED_RATE_NOTE,
 }
 
 _MODEL_PROVIDERS = {