loki-mode 6.18.0 → 6.19.0

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 v6.18.0
+# Loki Mode v6.19.0
 
 **You are an autonomous agent. You make decisions. You do not ask questions. You do not stop.**
 
@@ -267,4 +267,4 @@ 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 |
 
-**v6.18.0 | [Autonomi](https://www.autonomi.dev/) flagship product | ~260 lines core**
+**v6.19.0 | [Autonomi](https://www.autonomi.dev/) flagship product | ~260 lines core**

package/VERSION

@@ -1 +1 @@
-6.18.0
+6.19.0

package/autonomy/loki

@@ -435,6 +435,7 @@ show_help() {
     echo "  agent [cmd]      Agent type dispatch (list|info|run|start|review)"
     echo "  remote [PRD]     Start remote session (connect from phone/browser, Claude Pro/Max)"
   echo "  trigger          Event-driven autonomous execution (schedules, webhooks)"
+    echo "  failover [cmd]   Cross-provider auto-failover (status|--enable|--test|--chain)"
     echo "  plan <PRD>       Dry-run PRD analysis: complexity, cost, and execution plan"
     echo "  version          Show version"
     echo "  help             Show this help"
@@ -8299,6 +8300,225 @@ for line in sys.stdin:
     esac
 }
 
+# Cross-provider auto-failover management (v6.19.0)
+cmd_failover() {
+    local loki_dir="${TARGET_DIR:-.}/.loki"
+    local failover_file="$loki_dir/state/failover.json"
+    local script_dir="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+
+    if [ $# -eq 0 ]; then
+        # No args: show current failover config and health
+        if [ ! -f "$failover_file" ]; then
+            echo -e "${BOLD}Cross-Provider Auto-Failover${NC}"
+            echo ""
+            echo "Status: DISABLED"
+            echo ""
+            echo "Enable with: loki failover --enable"
+            echo "Or set:      LOKI_FAILOVER=true loki start"
+            return 0
+        fi
+
+        echo -e "${BOLD}Cross-Provider Auto-Failover${NC}"
+        echo ""
+        python3 << 'PYEOF'
+import json, os
+fpath = os.path.join(os.environ.get('TARGET_DIR', '.'), '.loki/state/failover.json')
+try:
+    with open(fpath) as f:
+        d = json.load(f)
+    status = "ENABLED" if d.get("enabled") else "DISABLED"
+    print(f"Status:           {status}")
+    print(f"Chain:            {' -> '.join(d.get('chain', []))}")
+    print(f"Current provider: {d.get('currentProvider', 'unknown')}")
+    print(f"Primary provider: {d.get('primaryProvider', 'unknown')}")
+    print(f"Failover count:   {d.get('failoverCount', 0)}")
+    last = d.get('lastFailover')
+    print(f"Last failover:    {last if last else 'never'}")
+    print()
+    print("Health Status:")
+    health = d.get('healthCheck', {})
+    for provider in d.get('chain', []):
+        h = health.get(provider, 'unknown')
+        indicator = '[OK]' if h == 'healthy' else '[--]' if h == 'unknown' else '[!!]'
+        print(f"  {indicator} {provider}: {h}")
+except FileNotFoundError:
+    print("Failover not initialized. Run: loki failover --enable")
+except Exception as e:
+    print(f"Error reading failover state: {e}")
+PYEOF
+        return 0
+    fi
+
+    while [[ $# -gt 0 ]]; do
+        case "$1" in
+            --enable)
+                mkdir -p "$loki_dir/state"
+                local current_provider
+                current_provider=$(cat "$loki_dir/state/provider" 2>/dev/null || echo "claude")
+                local chain="${LOKI_FAILOVER_CHAIN:-claude,codex,gemini}"
+
+                cat > "$failover_file" << FEOF
+{
+  "enabled": true,
+  "chain": $(printf '%s' "$chain" | python3 -c 'import sys,json; print(json.dumps(sys.stdin.read().strip().split(",")))' 2>/dev/null || echo '["claude","codex","gemini"]'),
+  "currentProvider": "$current_provider",
+  "primaryProvider": "$current_provider",
+  "lastFailover": null,
+  "failoverCount": 0,
+  "healthCheck": {}
+}
+FEOF
+                echo -e "${GREEN}Failover enabled${NC}"
+                echo "Chain: $chain"
+                echo "Primary: $current_provider"
+                shift
+                ;;
+            --disable)
+                if [ -f "$failover_file" ]; then
+                    python3 -c "
+import json
+with open('$failover_file') as f: d = json.load(f)
+d['enabled'] = False
+with open('$failover_file', 'w') as f: json.dump(d, f, indent=2)
+" 2>/dev/null
+                    echo -e "${YELLOW}Failover disabled${NC}"
+                else
+                    echo "Failover not initialized."
+                fi
+                shift
+                ;;
+            --chain)
+                shift
+                local new_chain="$1"
+                if [ -z "$new_chain" ]; then
+                    echo -e "${RED}Error: --chain requires a comma-separated list of providers${NC}"
+                    return 1
+                fi
+                # Validate each provider in chain
+                local IFS=','
+                for p in $new_chain; do
+                    case "$p" in
+                        claude|codex|gemini|cline|aider) ;;
+                        *) echo -e "${RED}Error: invalid provider '$p' in chain${NC}"; return 1 ;;
+                    esac
+                done
+                unset IFS
+
+                if [ ! -f "$failover_file" ]; then
+                    echo -e "${RED}Error: failover not initialized. Run: loki failover --enable${NC}"
+                    return 1
+                fi
+
+                python3 -c "
+import json
+with open('$failover_file') as f: d = json.load(f)
+d['chain'] = '$new_chain'.split(',')
+with open('$failover_file', 'w') as f: json.dump(d, f, indent=2)
+" 2>/dev/null
+                echo "Failover chain updated: $new_chain"
+                shift
+                ;;
+            --test)
+                echo -e "${BOLD}Testing all providers in failover chain...${NC}"
+                echo ""
+                # Source provider loader for health checks
+                if [ -f "$script_dir/../providers/loader.sh" ]; then
+                    source "$script_dir/../providers/loader.sh"
+                fi
+
+                local chain_providers="claude,codex,gemini"
+                if [ -f "$failover_file" ]; then
+                    chain_providers=$(python3 -c "import json; print(','.join(json.load(open('$failover_file')).get('chain', ['claude','codex','gemini'])))" 2>/dev/null || echo "claude,codex,gemini")
+                fi
+
+                local IFS=','
+                local all_pass=true
+                for p in $chain_providers; do
+                    printf "  %-10s " "$p:"
+
+                    # Check CLI installed
+                    local cli_name="$p"
+                    if command -v "$cli_name" &>/dev/null; then
+                        local cli_version
+                        cli_version=$("$cli_name" --version 2>/dev/null | head -1 || echo "unknown")
+                        printf "CLI %-20s " "[$cli_version]"
+                    else
+                        printf "CLI %-20s " "[NOT INSTALLED]"
+                        echo -e "${RED}FAIL${NC}"
+                        all_pass=false
+                        continue
+                    fi
+
+                    # Check API key
+                    local has_key=false
+                    case "$p" in
+                        claude) [ -n "${ANTHROPIC_API_KEY:-}" ] && has_key=true ;;
+                        codex) [ -n "${OPENAI_API_KEY:-}" ] && has_key=true ;;
+                        gemini) [ -n "${GOOGLE_API_KEY:-${GEMINI_API_KEY:-}}" ] && has_key=true ;;
+                        cline|aider) has_key=true ;;  # Key check varies
+                    esac
+
+                    if [ "$has_key" = "true" ]; then
+                        printf "Key [OK]  "
+                        echo -e "${GREEN}PASS${NC}"
+                    else
+                        printf "Key [MISSING]  "
+                        echo -e "${RED}FAIL${NC}"
+                        all_pass=false
+                    fi
+                done
+                unset IFS
+
+                echo ""
+                if [ "$all_pass" = "true" ]; then
+                    echo -e "${GREEN}All providers healthy${NC}"
+                else
+                    echo -e "${YELLOW}Some providers unavailable - failover chain may be limited${NC}"
+                fi
+                shift
+                ;;
+            --reset)
+                if [ -f "$failover_file" ]; then
+                    rm -f "$failover_file"
+                    echo "Failover state reset to defaults."
+                else
+                    echo "No failover state to reset."
+                fi
+                shift
+                ;;
+            --help|-h)
+                echo -e "${BOLD}loki failover${NC} - Cross-provider auto-failover management (v6.19.0)"
+                echo ""
+                echo "Usage: loki failover [options]"
+                echo ""
+                echo "Options:"
+                echo "  (no args)        Show failover status and health"
+                echo "  --enable         Enable auto-failover"
+                echo "  --disable        Disable auto-failover"
+                echo "  --chain X,Y,Z    Set failover chain (e.g., claude,codex,gemini)"
+                echo "  --test           Test all providers in chain"
+                echo "  --reset          Reset failover state to defaults"
+                echo "  --help, -h       Show this help"
+                echo ""
+                echo "Environment:"
+                echo "  LOKI_FAILOVER=true         Enable failover at startup"
+                echo "  LOKI_FAILOVER_CHAIN=X,Y,Z  Set default chain"
+                echo ""
+                echo "When a rate limit (429/529) is detected, Loki automatically switches"
+                echo "to the next healthy provider in the chain. After each successful"
+                echo "iteration on a fallback provider, the primary is health-checked and"
+                echo "execution switches back when it recovers."
+                return 0
+                ;;
+            *)
+                echo -e "${RED}Unknown option: $1${NC}"
+                echo "Run 'loki failover --help' for usage."
+                return 1
+                ;;
+        esac
+    done
+}
+
 # Dry-run PRD analysis and cost estimation (v6.18.0)
 cmd_plan() {
     local prd_file=""
@@ -8966,6 +9186,9 @@ main() {
         plan)
             cmd_plan "$@"
             ;;
+        failover)
+            cmd_failover "$@"
+            ;;
         version|--version|-v)
             cmd_version
             ;;

package/autonomy/run.sh

@@ -6525,6 +6525,278 @@ calculate_wait() {
     echo $wait_time
 }
 
+#===============================================================================
+# Cross-Provider Auto-Failover (v6.19.0)
+#===============================================================================
+
+# Initialize failover state file on startup
+init_failover_state() {
+    local failover_dir="${TARGET_DIR:-.}/.loki/state"
+    local failover_file="$failover_dir/failover.json"
+
+    # Only create if failover is enabled via env or config
+    if [ "${LOKI_FAILOVER:-false}" != "true" ]; then
+        return
+    fi
+
+    mkdir -p "$failover_dir"
+
+    if [ ! -f "$failover_file" ]; then
+        local chain="${LOKI_FAILOVER_CHAIN:-claude,codex,gemini}"
+        local primary="${PROVIDER_NAME:-claude}"
+        cat > "$failover_file" << FEOF
+{
+  "enabled": true,
+  "chain": $(printf '%s' "$chain" | python3 -c 'import sys,json; print(json.dumps(sys.stdin.read().strip().split(",")))' 2>/dev/null || echo '["claude","codex","gemini"]'),
+  "currentProvider": "$primary",
+  "primaryProvider": "$primary",
+  "lastFailover": null,
+  "failoverCount": 0,
+  "healthCheck": {
+    "$primary": "healthy"
+  }
+}
+FEOF
+        log_info "Failover initialized: chain=$chain, primary=$primary"
+    fi
+}
+
+# Read failover config from state file
+# Sets: FAILOVER_ENABLED, FAILOVER_CHAIN, FAILOVER_CURRENT, FAILOVER_PRIMARY
+read_failover_config() {
+    local failover_file="${TARGET_DIR:-.}/.loki/state/failover.json"
+
+    if [ ! -f "$failover_file" ]; then
+        FAILOVER_ENABLED="false"
+        return 1
+    fi
+
+    eval "$(python3 << 'PYEOF' 2>/dev/null || echo 'FAILOVER_ENABLED=false'
+import json, os
+try:
+    with open(os.path.join(os.environ.get('TARGET_DIR', '.'), '.loki/state/failover.json')) as f:
+        d = json.load(f)
+    chain = ','.join(d.get('chain', ['claude','codex','gemini']))
+    print(f'FAILOVER_ENABLED={str(d.get("enabled", False)).lower()}')
+    print(f'FAILOVER_CHAIN="{chain}"')
+    print(f'FAILOVER_CURRENT="{d.get("currentProvider", "claude")}"')
+    print(f'FAILOVER_PRIMARY="{d.get("primaryProvider", "claude")}"')
+    print(f'FAILOVER_COUNT={d.get("failoverCount", 0)}')
+except Exception:
+    print('FAILOVER_ENABLED=false')
+PYEOF
+    )"
+}
+
+# Update failover state file
+update_failover_state() {
+    local key="$1"
+    local value="$2"
+    local failover_file="${TARGET_DIR:-.}/.loki/state/failover.json"
+
+    [ ! -f "$failover_file" ] && return 1
+
+    python3 << PYEOF 2>/dev/null || true
+import json, os
+fpath = os.path.join(os.environ.get('TARGET_DIR', '.'), '.loki/state/failover.json')
+try:
+    with open(fpath) as f:
+        d = json.load(f)
+    key = "$key"
+    value = "$value"
+    # Handle type conversion
+    if value == "null":
+        d[key] = None
+    elif value == "true":
+        d[key] = True
+    elif value == "false":
+        d[key] = False
+    elif value.isdigit():
+        d[key] = int(value)
+    else:
+        d[key] = value
+    with open(fpath, 'w') as f:
+        json.dump(d, f, indent=2)
+except Exception:
+    pass
+PYEOF
+}
+
+# Update health status for a specific provider in failover.json
+update_failover_health() {
+    local provider="$1"
+    local status="$2"  # healthy, unhealthy, unknown
+    local failover_file="${TARGET_DIR:-.}/.loki/state/failover.json"
+
+    [ ! -f "$failover_file" ] && return 1
+
+    python3 << PYEOF 2>/dev/null || true
+import json, os
+fpath = os.path.join(os.environ.get('TARGET_DIR', '.'), '.loki/state/failover.json')
+try:
+    with open(fpath) as f:
+        d = json.load(f)
+    if 'healthCheck' not in d:
+        d['healthCheck'] = {}
+    d['healthCheck']["$provider"] = "$status"
+    with open(fpath, 'w') as f:
+        json.dump(d, f, indent=2)
+except Exception:
+    pass
+PYEOF
+}
+
+# Check provider health: API key exists + CLI installed
+# Returns: 0 if healthy, 1 if unhealthy
+check_provider_health() {
+    local provider="$1"
+
+    # Check CLI is installed
+    case "$provider" in
+        claude)
+            command -v claude &>/dev/null || return 1
+            [ -n "${ANTHROPIC_API_KEY:-}" ] || return 1
+            ;;
+        codex)
+            command -v codex &>/dev/null || return 1
+            [ -n "${OPENAI_API_KEY:-}" ] || return 1
+            ;;
+        gemini)
+            command -v gemini &>/dev/null || return 1
+            [ -n "${GOOGLE_API_KEY:-${GEMINI_API_KEY:-}}" ] || return 1
+            ;;
+        cline)
+            command -v cline &>/dev/null || return 1
+            ;;
+        aider)
+            command -v aider &>/dev/null || return 1
+            ;;
+        *)
+            return 1
+            ;;
+    esac
+
+    return 0
+}
+
+# Attempt failover to next healthy provider in chain
+# Called when rate limit is detected on current provider
+# Returns: 0 if failover succeeded, 1 if all providers exhausted
+attempt_provider_failover() {
+    read_failover_config || return 1
+
+    if [ "$FAILOVER_ENABLED" != "true" ]; then
+        return 1
+    fi
+
+    local current="${FAILOVER_CURRENT:-${PROVIDER_NAME:-claude}}"
+    log_warn "Failover: rate limit on $current, checking chain: $FAILOVER_CHAIN"
+
+    # Mark current as unhealthy
+    update_failover_health "$current" "unhealthy"
+
+    # Walk the chain looking for the next healthy provider
+    local IFS=','
+    local found_current=false
+    local tried_wrap=false
+
+    # Two passes: first from current position to end, then from start to current
+    for provider in $FAILOVER_CHAIN $FAILOVER_CHAIN; do
+        if [ "$provider" = "$current" ]; then
+            if [ "$found_current" = "true" ]; then
+                # We've wrapped around, all exhausted
+                break
+            fi
+            found_current=true
+            continue
+        fi
+
+        [ "$found_current" != "true" ] && continue
+
+        # Check if this provider is healthy
+        if check_provider_health "$provider"; then
+            log_info "Failover: switching from $current to $provider"
+
+            # Load the new provider config
+            local provider_dir
+            provider_dir="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)/providers"
+            if [ -f "$provider_dir/$provider.sh" ]; then
+                source "$provider_dir/$provider.sh"
+            fi
+
+            # Update state
+            update_failover_state "currentProvider" "$provider"
+            update_failover_state "lastFailover" "$(date -u +%Y-%m-%dT%H:%M:%SZ)"
+            update_failover_state "failoverCount" "$((FAILOVER_COUNT + 1))"
+            update_failover_health "$provider" "healthy"
+
+            # Update runtime provider vars
+            PROVIDER_NAME="$provider"
+
+            emit_event_json "provider_failover" \
+                "from=$current" \
+                "to=$provider" \
+                "reason=rate_limit" \
+                "iteration=$ITERATION_COUNT" 2>/dev/null || true
+
+            log_info "Failover: now using $provider (failover #$((FAILOVER_COUNT + 1)))"
+            return 0
+        else
+            log_debug "Failover: $provider is unhealthy, skipping"
+            update_failover_health "$provider" "unhealthy"
+        fi
+    done
+
+    log_warn "Failover: all providers in chain exhausted, falling back to retry"
+    return 1
+}
+
+# Check if primary provider has recovered after running on a fallback
+# Called after each successful iteration when on a non-primary provider
+# Returns: 0 if switched back to primary, 1 if still on fallback
+check_primary_recovery() {
+    read_failover_config || return 1
+
+    if [ "$FAILOVER_ENABLED" != "true" ]; then
+        return 1
+    fi
+
+    local current="${FAILOVER_CURRENT:-${PROVIDER_NAME:-claude}}"
+    local primary="${FAILOVER_PRIMARY:-claude}"
+
+    # Already on primary
+    if [ "$current" = "$primary" ]; then
+        return 1
+    fi
+
+    # Check if primary is healthy again
+    if check_provider_health "$primary"; then
+        log_info "Failover: primary provider $primary appears healthy, switching back"
+
+        # Load primary provider config
+        local provider_dir
+        provider_dir="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)/providers"
+        if [ -f "$provider_dir/$primary.sh" ]; then
+            source "$provider_dir/$primary.sh"
+        fi
+
+        update_failover_state "currentProvider" "$primary"
+        update_failover_health "$primary" "healthy"
+
+        PROVIDER_NAME="$primary"
+
+        emit_event_json "provider_recovery" \
+            "from=$current" \
+            "to=$primary" \
+            "iteration=$ITERATION_COUNT" 2>/dev/null || true
+
+        log_info "Failover: recovered to primary provider $primary"
+        return 0
+    fi
+
+    return 1
+}
+
 #===============================================================================
 # Rate Limit Detection
 #===============================================================================
@@ -8145,6 +8417,9 @@ run_autonomous() {
     load_state
     local retry=$RETRY_COUNT
 
+    # Initialize Cross-Provider Failover (v6.19.0)
+    init_failover_state
+
     # Initialize Completion Council (v5.25.0)
     if type council_init &>/dev/null; then
         council_init "$prd_path"
@@ -8784,6 +9059,9 @@ if __name__ == "__main__":
                 log_warn "Council will evaluate at next check interval (every ${COUNCIL_CHECK_INTERVAL:-5} iterations)"
             fi
 
+            # Cross-provider failover: check if primary has recovered (v6.19.0)
+            check_primary_recovery 2>/dev/null || true
+
             # SUCCESS exit - continue IMMEDIATELY to next iteration (no wait!)
             log_step "Starting next iteration..."
             ((retry++))
@@ -8801,6 +9079,13 @@ if __name__ == "__main__":
         local wait_time
 
         if [ $rate_limit_wait -gt 0 ]; then
+            # Cross-provider failover (v6.19.0): try switching provider before waiting
+            if attempt_provider_failover 2>/dev/null; then
+                log_info "Failover succeeded - retrying immediately with ${PROVIDER_NAME}"
+                ((retry++))
+                continue
+            fi
+
             wait_time=$rate_limit_wait
             local human_time=$(format_duration $wait_time)
             log_warn "Rate limit detected! Waiting until reset (~$human_time)..."

package/dashboard/__init__.py

@@ -7,7 +7,7 @@ Modules:
     control: Session control API (start/stop/pause/resume)
 """
 
-__version__ = "6.18.0"
+__version__ = "6.19.0"
 
 # Expose the control app for easy import
 try:

package/docs/INSTALLATION.md

@@ -2,7 +2,7 @@
 
 The flagship product of [Autonomi](https://www.autonomi.dev/). Complete installation instructions for all platforms and use cases.
 
-**Version:** v6.18.0
+**Version:** v6.19.0
 
 ---
 

package/mcp/__init__.py

@@ -57,4 +57,4 @@ try:
 except ImportError:
     __all__ = ['mcp']
 
-__version__ = '6.18.0'
+__version__ = '6.19.0'

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "loki-mode",
-  "version": "6.18.0",
+  "version": "6.19.0",
   "description": "Loki Mode by Autonomi - Multi-agent autonomous startup system for Claude Code, Codex CLI, and Gemini CLI",
   "keywords": [
     "agent",