loki-mode 5.49.4 → 5.51.0

package/README.md

@@ -11,7 +11,7 @@
 [![Agent Types](https://img.shields.io/badge/Agent%20Types-41-blue)]()
 [![Benchmarks](https://img.shields.io/badge/Benchmarks-Infrastructure%20Ready-blue)](benchmarks/)
 
-**Current Version: v5.49.4**
+**Current Version: v5.51.0**
 
 **[Autonomi](https://www.autonomi.dev/)** | **[Documentation](https://www.autonomi.dev/docs)** | **[GitHub](https://github.com/asklokesh/loki-mode)**
 
@@ -85,7 +85,7 @@ gemini
 ### Verify Installation
 
 ```bash
-loki --version    # Should print 5.49.4
+loki --version    # Should print 5.51.0
 loki doctor       # Check skill symlinks and provider availability
 ```
 

package/SKILL.md

@@ -3,7 +3,7 @@ name: loki-mode
 description: Multi-agent autonomous startup system. Triggers on "Loki Mode". Takes PRD to deployed product with minimal human intervention. Requires --dangerously-skip-permissions flag.
 ---
 
-# Loki Mode v5.49.4
+# Loki Mode v5.51.0
 
 **You are an autonomous agent. You make decisions. You do not ask questions. You do not stop.**
 
@@ -263,4 +263,4 @@ The following features are documented in skill modules but not yet fully automat
 | Quality gates 3-reviewer system | Implemented (v5.35.0) | 5 specialist reviewers in `skills/quality-gates.md`; execution in run.sh |
 | Benchmarks (HumanEval, SWE-bench) | Infrastructure only | Runner scripts and datasets exist in `benchmarks/`; no published results |
 
-**v5.49.4 | [Autonomi](https://www.autonomi.dev/) flagship product | ~260 lines core**
+**v5.51.0 | [Autonomi](https://www.autonomi.dev/) flagship product | ~260 lines core**

package/VERSION

@@ -1 +1 @@
-5.49.4
+5.51.0

package/autonomy/completion-council.sh

@@ -216,6 +216,19 @@ council_vote() {
     local vote_dir="$COUNCIL_STATE_DIR/votes/iteration-$ITERATION_COUNT"
     mkdir -p "$vote_dir"
 
+    # Council v2: true blind review with sycophancy detection
+    if [ "${LOKI_COUNCIL_VERSION:-1}" = "2" ]; then
+        # Source council v2 if not already loaded
+        if ! type council_v2_vote &>/dev/null; then
+            source "${BASH_SOURCE[0]%/*}/council-v2.sh"
+        fi
+        # Gather evidence first (shared function)
+        local evidence_file="$vote_dir/evidence.md"
+        council_gather_evidence "$evidence_file" "$prd_path"
+        council_v2_vote "$prd_path" "$evidence_file" "$vote_dir" "${ITERATION_COUNT:-0}"
+        return $?
+    fi
+
     log_header "COMPLETION COUNCIL - Iteration $ITERATION_COUNT"
     log_info "Convening ${COUNCIL_SIZE}-member council..."
 

package/autonomy/council-v2.sh

@@ -0,0 +1,302 @@
+#!/usr/bin/env bash
+#===============================================================================
+# Council v2 - True blind review with sycophancy detection
+#
+# Upgraded council review system that provides:
+#   1. True blind review (isolated evidence packages, no cross-contamination)
+#   2. Sycophancy detection via statistical analysis
+#   3. Reviewer calibration tracking over time
+#   4. Anti-sycophancy devil's advocate on high sycophancy scores
+#
+# Activated via: LOKI_COUNCIL_VERSION=2
+#
+# Environment Variables:
+#   LOKI_COUNCIL_SYCOPHANCY_THRESHOLD - Score above which to trigger devil's advocate (default: 0.6)
+#   LOKI_COUNCIL_VERSION              - Set to "2" to activate this module
+#
+# Dependencies:
+#   - completion-council.sh (sourced by caller for shared functions)
+#   - swarm/sycophancy.py (sycophancy detection)
+#   - swarm/calibration.py (reviewer calibration)
+#
+#===============================================================================
+
+COUNCIL_V2_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+
+#===============================================================================
+# council_v2_vote() -- Main entry point for v2 voting
+#
+# 1. Create isolated evidence packages (no cross-contamination)
+# 2. Launch parallel reviewers with isolated evidence
+# 3. Collect votes independently
+# 4. Run sycophancy detection via Python
+# 5. Run calibration tracking
+# 6. Apply anti-sycophancy if needed (devil's advocate)
+# 7. Return final verdict
+#===============================================================================
+
+council_v2_vote() {
+    local prd_path="$1"
+    local evidence_file="$2"
+    local vote_dir="$3"
+    local iteration="${4:-0}"
+
+    local council_size="${COUNCIL_SIZE:-3}"
+
+    log_header "COMPLETION COUNCIL v2 - Iteration $iteration"
+    log_info "Convening ${council_size}-member blind review council..."
+
+    # Step 1: Create isolated evidence packages
+    local review_dirs=()
+    for i in $(seq 1 "$council_size"); do
+        local review_dir
+        review_dir=$(mktemp -d)
+        cp "$evidence_file" "$review_dir/evidence.md"
+        if [ -n "$prd_path" ] && [ -f "$prd_path" ]; then
+            cp "$prd_path" "$review_dir/prd.md"
+        fi
+        review_dirs+=("$review_dir")
+    done
+
+    # Step 2: Launch parallel reviewers
+    local vote_files=()
+    for i in $(seq 0 $((council_size - 1))); do
+        local role
+        case $i in
+            0) role="requirements_verifier" ;;
+            1) role="test_auditor" ;;
+            *) role="code_quality_reviewer" ;;
+        esac
+        local vote_file="${review_dirs[$i]}/vote.json"
+        vote_files+=("$vote_file")
+
+        # Each reviewer gets ONLY their evidence package (no other votes visible)
+        council_v2_run_reviewer "$role" "${review_dirs[$i]}" "$vote_file" &
+    done
+
+    # Wait for all reviewers to complete
+    wait
+
+    # Step 3: Collect votes
+    local votes_json="["
+    local first=true
+    local approve_count=0
+    local reject_count=0
+    for vote_file in "${vote_files[@]}"; do
+        if [ -f "$vote_file" ]; then
+            local vote_content
+            vote_content=$(cat "$vote_file")
+            if [ "$first" = "true" ]; then
+                first=false
+            else
+                votes_json="$votes_json,"
+            fi
+            votes_json="$votes_json$vote_content"
+
+            local verdict
+            verdict=$(echo "$vote_content" | python3 -c "import sys,json; print(json.load(sys.stdin).get('verdict','').upper())" 2>/dev/null || echo "UNKNOWN")
+            if [ "$verdict" = "APPROVE" ]; then
+                ((approve_count++))
+            else
+                ((reject_count++))
+            fi
+        fi
+    done
+    votes_json="$votes_json]"
+
+    log_info "Blind review results: $approve_count APPROVE / $reject_count REJECT"
+
+    # Step 4: Sycophancy detection
+    local sycophancy_score
+    sycophancy_score=$(python3 -c "
+import sys
+sys.path.insert(0, '${COUNCIL_V2_DIR}/../swarm')
+from sycophancy import detect_sycophancy
+import json
+votes = json.loads(sys.argv[1])
+print('{:.3f}'.format(detect_sycophancy(votes)))
+" "$votes_json" 2>/dev/null || echo "0.000")
+
+    log_info "Sycophancy score: $sycophancy_score"
+
+    # Step 5: Anti-sycophancy check
+    if [ "$approve_count" -eq "$council_size" ]; then
+        local threshold="${LOKI_COUNCIL_SYCOPHANCY_THRESHOLD:-0.6}"
+        local should_challenge
+        should_challenge=$(python3 -c "print('yes' if float('$sycophancy_score') >= float('$threshold') else 'no')" 2>/dev/null || echo "no")
+
+        if [ "$should_challenge" = "yes" ]; then
+            log_warn "Sycophancy score $sycophancy_score >= $threshold -- adding devil's advocate"
+            # Run devil's advocate with fresh evidence (no visibility of other votes)
+            local da_dir
+            da_dir=$(mktemp -d)
+            cp "$evidence_file" "$da_dir/evidence.md"
+            [ -n "$prd_path" ] && [ -f "$prd_path" ] && cp "$prd_path" "$da_dir/prd.md"
+            local da_vote="$da_dir/vote.json"
+            council_v2_run_reviewer "devils_advocate" "$da_dir" "$da_vote"
+
+            if [ -f "$da_vote" ]; then
+                local da_verdict
+                da_verdict=$(cat "$da_vote" | python3 -c "import sys,json; print(json.load(sys.stdin).get('verdict','').upper())" 2>/dev/null || echo "REJECT")
+                if [ "$da_verdict" = "REJECT" ]; then
+                    log_warn "Devil's advocate REJECTED unanimous approval"
+                    approve_count=$((approve_count - 1))
+                    reject_count=$((reject_count + 1))
+                else
+                    log_info "Devil's advocate confirmed unanimous approval"
+                fi
+            fi
+            rm -rf "$da_dir"
+        fi
+    fi
+
+    # Step 6: Calibration tracking
+    local final_decision
+    if [ "$approve_count" -ge "${COUNCIL_THRESHOLD:-2}" ]; then
+        final_decision="approve"
+    else
+        final_decision="reject"
+    fi
+
+    python3 -c "
+import sys
+sys.path.insert(0, '${COUNCIL_V2_DIR}/../swarm')
+from calibration import CalibrationTracker
+import json
+tracker = CalibrationTracker('.loki/council/calibration.json')
+votes = json.loads(sys.argv[1])
+for i, v in enumerate(votes):
+    v['reviewer_id'] = 'reviewer-' + str(i + 1)
+tracker.record_round(int(sys.argv[2]), votes, sys.argv[3])
+tracker.save()
+" "$votes_json" "$iteration" "$final_decision" 2>/dev/null || true
+
+    # Step 7: Save results
+    mkdir -p "$vote_dir"
+    echo "$votes_json" > "$vote_dir/all-votes.json"
+    cat > "$vote_dir/summary.json" << SUMMARY_EOF
+{
+    "version": 2,
+    "approve": $approve_count,
+    "reject": $reject_count,
+    "sycophancy_score": $sycophancy_score,
+    "decision": "$final_decision",
+    "council_size": $council_size,
+    "iteration": $iteration
+}
+SUMMARY_EOF
+
+    log_info "Council v2 verdict: $approve_count APPROVE / $reject_count REJECT -> $final_decision (sycophancy: $sycophancy_score)"
+
+    # Emit event for dashboard
+    emit_event_json "council_vote" \
+        "version=2" \
+        "iteration=$iteration" \
+        "approve=$approve_count" \
+        "reject=$reject_count" \
+        "threshold=${COUNCIL_THRESHOLD:-2}" \
+        "sycophancy_score=$sycophancy_score" \
+        "result=$(echo "$final_decision" | tr '[:lower:]' '[:upper:]')" 2>/dev/null || true
+
+    # Cleanup isolated review directories
+    for dir in "${review_dirs[@]}"; do
+        rm -rf "$dir"
+    done
+
+    # Return result
+    if [ "$final_decision" = "approve" ]; then
+        return 0
+    else
+        return 1
+    fi
+}
+
+#===============================================================================
+# council_v2_run_reviewer() -- Run a single isolated reviewer
+#
+# Each reviewer receives only their evidence package and produces a JSON vote.
+# No reviewer can see another reviewer's output (true blind review).
+#===============================================================================
+
+council_v2_run_reviewer() {
+    local role="$1"
+    local review_dir="$2"
+    local output_file="$3"
+
+    # Build review prompt based on role
+    local prompt
+    case "$role" in
+        requirements_verifier)
+            prompt="You are a requirements verification reviewer. Review the evidence and PRD below. Check if all requirements are implemented. Output a JSON object with keys: verdict (APPROVE or REJECT), reasoning (string), issues (array of {severity, description})."
+            ;;
+        test_auditor)
+            prompt="You are a test auditor reviewer. Review the evidence below. Check test coverage, test quality, and assertion density. Output a JSON object with keys: verdict (APPROVE or REJECT), reasoning (string), issues (array of {severity, description})."
+            ;;
+        code_quality_reviewer)
+            prompt="You are a code quality reviewer. Review the evidence below. Check for SOLID violations, security issues, performance problems. Output a JSON object with keys: verdict (APPROVE or REJECT), reasoning (string), issues (array of {severity, description})."
+            ;;
+        devils_advocate)
+            prompt="You are a devil's advocate reviewer. Your job is to find reasons this code should NOT be approved. Be skeptical and contrarian. Output a JSON object with keys: verdict (APPROVE or REJECT), reasoning (string), issues (array of {severity, description})."
+            ;;
+        *)
+            prompt="You are a general reviewer. Evaluate project completion. Output a JSON object with keys: verdict (APPROVE or REJECT), reasoning (string), issues (array of {severity, description})."
+            ;;
+    esac
+
+    local evidence=""
+    [ -f "$review_dir/evidence.md" ] && evidence=$(cat "$review_dir/evidence.md")
+    local prd=""
+    [ -f "$review_dir/prd.md" ] && prd=$(head -100 "$review_dir/prd.md")
+
+    # Use the current provider to run the review
+    local full_prompt="$prompt
+
+## Evidence
+$evidence
+
+## PRD (first 100 lines)
+$prd
+
+Respond ONLY with a valid JSON object. No markdown fencing."
+
+    local result
+    case "${PROVIDER_NAME:-claude}" in
+        claude)
+            if command -v claude &>/dev/null; then
+                result=$(echo "$full_prompt" | claude --model haiku -p 2>/dev/null || echo '{"verdict":"REJECT","reasoning":"review execution failed","issues":[]}')
+            else
+                result='{"verdict":"REJECT","reasoning":"reviewer CLI unavailable","issues":[]}'
+            fi
+            ;;
+        codex)
+            if command -v codex &>/dev/null; then
+                result=$(codex exec -q "$full_prompt" 2>/dev/null || echo '{"verdict":"REJECT","reasoning":"review execution failed","issues":[]}')
+            else
+                result='{"verdict":"REJECT","reasoning":"reviewer CLI unavailable","issues":[]}'
+            fi
+            ;;
+        gemini)
+            if command -v gemini &>/dev/null; then
+                result=$(echo "$full_prompt" | gemini 2>/dev/null || echo '{"verdict":"REJECT","reasoning":"review execution failed","issues":[]}')
+            else
+                result='{"verdict":"REJECT","reasoning":"reviewer CLI unavailable","issues":[]}'
+            fi
+            ;;
+        *)
+            result='{"verdict":"REJECT","reasoning":"review not supported for this provider","issues":[]}'
+            ;;
+    esac
+
+    # Extract JSON from result (handle markdown fencing)
+    local extracted
+    extracted=$(echo "$result" | sed -n '/^{/,/^}/p' | head -50)
+    if [ -z "$extracted" ]; then
+        # Try removing markdown fencing
+        extracted=$(echo "$result" | sed 's/^```json//;s/^```//' | sed -n '/^{/,/^}/p' | head -50)
+    fi
+    if [ -z "$extracted" ]; then
+        extracted='{"verdict":"REJECT","reasoning":"failed to parse review output","issues":[]}'
+    fi
+
+    echo "$extracted" > "$output_file"
+}

package/autonomy/loki

@@ -130,6 +130,81 @@ get_version() {
     fi
 }
 
+# Ensure dashboard Python venv with all deps installed.
+# Uses ~/.loki/dashboard-venv (persistent, writable, survives npm/brew upgrades).
+# Sets DASHBOARD_PYTHON to the venv python3 path on success.
+# Returns 0 on success, 1 on failure.
+ensure_dashboard_venv() {
+    local dashboard_venv="$HOME/.loki/dashboard-venv"
+    DASHBOARD_PYTHON="python3"
+
+    # Use existing venv python if available
+    if [ -x "${dashboard_venv}/bin/python3" ]; then
+        DASHBOARD_PYTHON="${dashboard_venv}/bin/python3"
+    fi
+
+    # Check all required imports
+    if "$DASHBOARD_PYTHON" -c "import fastapi; import sqlalchemy; import aiosqlite" 2>/dev/null; then
+        return 0
+    fi
+
+    echo -e "${YELLOW}Setting up dashboard virtualenv...${NC}"
+
+    # Create venv if missing or broken
+    if ! [ -x "${dashboard_venv}/bin/python3" ]; then
+        # Remove broken venv if exists
+        if [ -d "$dashboard_venv" ]; then
+            echo -e "${YELLOW}Removing broken dashboard venv...${NC}"
+            rm -rf "$dashboard_venv"
+        fi
+        mkdir -p "$HOME/.loki"
+        python3 -m venv "$dashboard_venv" 2>/dev/null || python3.13 -m venv "$dashboard_venv" 2>/dev/null || {
+            echo -e "${RED}Failed to create dashboard virtualenv${NC}"
+            echo "You may need to install python3-venv:"
+            echo "  sudo apt install python3-venv    (Debian/Ubuntu)"
+            echo "  brew install python3              (macOS)"
+            return 1
+        }
+    fi
+
+    DASHBOARD_PYTHON="${dashboard_venv}/bin/python3"
+    echo -e "${YELLOW}Installing dashboard dependencies...${NC}"
+
+    # Try pinned requirements first, then unpinned fallback
+    local req_file="${SKILL_DIR}/dashboard/requirements.txt"
+    local installed=false
+    if [ -f "$req_file" ]; then
+        if "${dashboard_venv}/bin/pip" install -r "$req_file" 2>&1 | tail -1; then
+            installed=true
+        else
+            echo -e "${YELLOW}Pinned deps failed, trying unpinned...${NC}"
+        fi
+    fi
+
+    if [ "$installed" = false ]; then
+        # Install without greenlet first (it may need a C compiler).
+        # SQLAlchemy only requires greenlet for sync-in-async; aiosqlite is pure async.
+        if ! "${dashboard_venv}/bin/pip" install fastapi uvicorn pydantic websockets sqlalchemy aiosqlite 2>&1 | tail -1; then
+            echo -e "${RED}Failed to install dashboard dependencies${NC}"
+            echo "Try manually: ${dashboard_venv}/bin/pip install fastapi uvicorn sqlalchemy aiosqlite"
+            return 1
+        fi
+        # Try greenlet separately (optional, needs C compiler on some platforms)
+        "${dashboard_venv}/bin/pip" install greenlet 2>/dev/null || true
+    fi
+
+    # Verify imports work after install
+    if ! "$DASHBOARD_PYTHON" -c "import fastapi; import sqlalchemy; import aiosqlite" 2>/dev/null; then
+        echo -e "${RED}Dashboard dependencies installed but imports still fail${NC}"
+        echo "Try removing the venv and retrying:"
+        echo "  rm -rf ${dashboard_venv}"
+        echo "  loki dashboard start"
+        return 1
+    fi
+
+    return 0
+}
+
 # Check if jq is available (called by functions that need it)
 require_jq() {
     if ! command -v jq &> /dev/null; then
@@ -1565,14 +1640,12 @@ cmd_dashboard_start() {
         exit 1
     fi
 
-    # Determine python command -- prefer dashboard venv if available
-    local python_cmd="python3"
-    local dashboard_venv="${SKILL_DIR}/dashboard/.venv"
-    if [ -x "${dashboard_venv}/bin/python3" ]; then
-        python_cmd="${dashboard_venv}/bin/python3"
-    elif ! command -v python3 &> /dev/null; then
-        python_cmd="python"
+    # Set up dashboard venv and python command
+    if ! ensure_dashboard_venv; then
+        echo -e "${RED}Cannot start dashboard without Python dependencies${NC}"
+        return 1
     fi
+    local python_cmd="$DASHBOARD_PYTHON"
 
     # Check if already running
     if [ -f "$DASHBOARD_PID_FILE" ]; then
@@ -1618,34 +1691,6 @@ cmd_dashboard_start() {
         tls_info=" (TLS enabled)"
     fi
 
-    # Ensure dashboard Python dependencies via virtualenv (PEP 668 safe)
-    if ! "$python_cmd" -c "import fastapi" 2>/dev/null; then
-        echo -e "${YELLOW}Setting up dashboard virtualenv...${NC}"
-        local req_file="${SKILL_DIR}/dashboard/requirements.txt"
-        if ! [ -d "$dashboard_venv" ]; then
-            python3 -m venv "$dashboard_venv" 2>/dev/null || python3.13 -m venv "$dashboard_venv" 2>/dev/null || true
-        fi
-        if [ -x "${dashboard_venv}/bin/python3" ]; then
-            python_cmd="${dashboard_venv}/bin/python3"
-            echo -e "${YELLOW}Installing dashboard dependencies into venv...${NC}"
-            if [ -f "$req_file" ]; then
-                "${dashboard_venv}/bin/pip" install -q -r "$req_file" 2>/dev/null || {
-                    echo -e "${YELLOW}Pinned deps failed, installing core deps...${NC}"
-                    "${dashboard_venv}/bin/pip" install -q fastapi uvicorn pydantic websockets 2>/dev/null || true
-                }
-            else
-                "${dashboard_venv}/bin/pip" install -q fastapi uvicorn pydantic websockets 2>/dev/null || true
-            fi
-        else
-            # Fallback: try direct pip (may fail on PEP 668 systems)
-            pip3 install -q fastapi uvicorn pydantic websockets 2>/dev/null || pip install -q fastapi uvicorn pydantic websockets 2>/dev/null || {
-                echo -e "${RED}Failed to install dashboard dependencies${NC}"
-                echo "Run manually: python3 -m venv ${dashboard_venv} && ${dashboard_venv}/bin/pip install fastapi uvicorn pydantic websockets"
-                exit 1
-            }
-        fi
-    fi
-
     echo -e "${GREEN}Starting dashboard server...${NC}"
     echo -e "${CYAN}Host:${NC} $host"
     echo -e "${CYAN}Port:${NC} $port"
@@ -3347,35 +3392,12 @@ cmd_api() {
                 fi
             fi
 
-            # Ensure dashboard Python dependencies via virtualenv (PEP 668 safe)
-            local dashboard_venv="${SKILL_DIR}/dashboard/.venv"
-            local api_python="python3"
-            if [ -x "${dashboard_venv}/bin/python3" ]; then
-                api_python="${dashboard_venv}/bin/python3"
-            fi
-            if ! "$api_python" -c "import fastapi" 2>/dev/null; then
-                echo -e "${YELLOW}Setting up dashboard virtualenv...${NC}"
-                if ! [ -d "$dashboard_venv" ]; then
-                    python3 -m venv "$dashboard_venv" 2>/dev/null || python3.13 -m venv "$dashboard_venv" 2>/dev/null || true
-                fi
-                if [ -x "${dashboard_venv}/bin/python3" ]; then
-                    api_python="${dashboard_venv}/bin/python3"
-                    local req_file="${SKILL_DIR}/dashboard/requirements.txt"
-                    if [ -f "$req_file" ]; then
-                        "${dashboard_venv}/bin/pip" install -q -r "$req_file" 2>/dev/null || {
-                            "${dashboard_venv}/bin/pip" install -q fastapi uvicorn pydantic websockets 2>/dev/null || true
-                        }
-                    else
-                        "${dashboard_venv}/bin/pip" install -q fastapi uvicorn pydantic websockets 2>/dev/null || true
-                    fi
-                else
-                    pip3 install -q fastapi uvicorn pydantic websockets 2>/dev/null || {
-                        echo -e "${RED}Failed to install dashboard dependencies${NC}"
-                        echo "Run manually: python3 -m venv ${dashboard_venv} && ${dashboard_venv}/bin/pip install fastapi uvicorn pydantic websockets"
-                        exit 1
-                    }
-                fi
+            # Ensure dashboard Python dependencies
+            if ! ensure_dashboard_venv; then
+                echo -e "${RED}Cannot start dashboard without Python dependencies${NC}"
+                exit 1
             fi
+            local api_python="$DASHBOARD_PYTHON"
 
             # Start server
             mkdir -p "$LOKI_DIR/logs" "$LOKI_DIR/dashboard"
@@ -3941,7 +3963,13 @@ SESSEOF
             sleep 1
         fi
 
-        LOKI_DIR="$LOKI_DIR" PYTHONPATH="$SKILL_DIR" python3 -m uvicorn dashboard.server:app \
+        # Set up dashboard venv
+        if ! ensure_dashboard_venv; then
+            echo -e "${YELLOW}Warning: Dashboard dependencies not available, continuing without dashboard${NC}"
+        fi
+        local demo_python="$DASHBOARD_PYTHON"
+
+        LOKI_DIR="$LOKI_DIR" PYTHONPATH="$SKILL_DIR" "$demo_python" -m uvicorn dashboard.server:app \
             --host 127.0.0.1 --port 57374 --log-level warning &>/dev/null &
         local dash_pid=$!
         echo "$dash_pid" > "$LOKI_DIR/dashboard/dashboard.pid"