loki-mode 6.67.3 → 6.68.0

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 v6.67.3
+# Loki Mode v6.68.0
 
 **You are an autonomous agent. You make decisions. You do not ask questions. You do not stop.**
 
@@ -268,4 +268,4 @@ 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 |
 
-**v6.67.3 | [Autonomi](https://www.autonomi.dev/) flagship product | ~260 lines core**
+**v6.68.0 | [Autonomi](https://www.autonomi.dev/) flagship product | ~260 lines core**

package/VERSION

@@ -1 +1 @@
-6.67.3
+6.68.0

package/autonomy/loki

@@ -7054,70 +7054,191 @@ QPRDEOF
 cmd_monitor() {
     local project_dir="${1:-.}"
 
+    # Resolve to absolute path
+    if [[ ! "$project_dir" = /* ]]; then
+        project_dir="$(cd "$project_dir" 2>/dev/null && pwd)" || {
+            echo -e "${RED}Error: directory not found: $1${NC}"
+            return 1
+        }
+    fi
+
     if [[ ! -f "$project_dir/docker-compose.yml" ]] && [[ ! -f "$project_dir/docker-compose.yaml" ]]; then
         echo -e "${RED}No docker-compose.yml found in $project_dir${NC}"
         return 1
     fi
 
-    echo -e "${CYAN}Monitoring Docker Compose services in $project_dir...${NC}"
+    echo -e "${CYAN}AI-Powered Docker Monitor${NC}"
+    echo -e "Project: ${BOLD}$project_dir${NC}"
+    echo -e "Provider: ${GREEN}${LOKI_PROVIDER:-claude}${NC}"
     echo "Press Ctrl+C to stop."
     echo ""
 
     local fix_count=0
-    local max_fixes=10
+    local max_fixes="${LOKI_MONITOR_MAX_FIXES:-10}"
+    local poll_interval="${LOKI_MONITOR_INTERVAL:-10}"
+    local consecutive_healthy=0
+
+    trap 'echo ""; echo -e "${CYAN}Monitor stopped.${NC}"; return 0' INT TERM
 
     while true; do
-        # Check service status
-        local status
-        status=$(cd "$project_dir" && docker compose ps -a --format json 2>/dev/null)
+        # 1. CAPTURE: Get service status (all services including stopped)
+        local ps_output
+        ps_output=$(cd "$project_dir" && docker compose ps -a --format json 2>/dev/null)
 
-        if [[ -z "$status" ]]; then
+        if [[ -z "$ps_output" ]]; then
             echo -e "${YELLOW}No services found. Waiting...${NC}"
-            sleep 10
+            sleep "$poll_interval"
             continue
         fi
 
-        # Parse each service
-        local has_failure=false
+        # Parse services and find failures
+        local failed_services=""
+        local all_healthy=true
+        local service_summary=""
+
         while IFS= read -r line; do
             [[ -z "$line" ]] && continue
-            local svc_name svc_state
-            svc_name=$(echo "$line" | python3 -c "import json,sys; d=json.loads(sys.stdin.read()); print(d.get('Service',d.get('Name','')))" 2>/dev/null)
-            svc_state=$(echo "$line" | python3 -c "import json,sys; d=json.loads(sys.stdin.read()); print(d.get('State',''))" 2>/dev/null)
+            local svc_name svc_state svc_exit
+            svc_name=$(echo "$line" | python3 -c "import json,sys; d=json.loads(sys.stdin.read()); print(d.get('Service',d.get('Name','')))" 2>/dev/null || echo "unknown")
+            svc_state=$(echo "$line" | python3 -c "import json,sys; d=json.loads(sys.stdin.read()); print(d.get('State','unknown'))" 2>/dev/null || echo "unknown")
+            svc_exit=$(echo "$line" | python3 -c "import json,sys; d=json.loads(sys.stdin.read()); print(d.get('ExitCode',0))" 2>/dev/null || echo "0")
+
+            service_summary="${service_summary}${svc_name}=${svc_state} "
 
             if [[ "$svc_state" == "exited" ]] || [[ "$svc_state" == "dead" ]]; then
-                has_failure=true
-                echo -e "${RED}Service '$svc_name' is $svc_state${NC}"
+                all_healthy=false
+                failed_services="${failed_services} ${svc_name}"
+            fi
+        done <<< "$ps_output"
 
-                if [[ $fix_count -lt $max_fixes ]]; then
-                    # Get logs for the failed service
-                    local logs
-                    logs=$(cd "$project_dir" && docker compose logs --tail 30 "$svc_name" 2>/dev/null)
+        if [[ "$all_healthy" == "true" ]]; then
+            consecutive_healthy=$((consecutive_healthy + 1))
+            printf "\r${GREEN}[healthy]${NC} %s (%s)   " "$service_summary" "$(date +%H:%M:%S)"
+            sleep "$poll_interval"
+            continue
+        fi
 
-                    echo -e "${CYAN}Auto-fixing $svc_name...${NC}"
+        consecutive_healthy=0
 
-                    # Run loki quick with targeted fix
-                    LOKI_MAX_ITERATIONS=5 LOKI_AUTO_FIX=true \
-                    loki quick "Fix the '$svc_name' Docker service. It exited with an error. Logs: $logs" \
-                        2>&1 | while IFS= read -r fline; do echo "  $fline"; done
+        # 2. CAPTURE LOGS and CONTEXT for each failed service, then feed to AI
+        for svc_name in $failed_services; do
+            if [[ $fix_count -ge $max_fixes ]]; then
+                echo -e "\n${RED}Max fix attempts ($max_fixes) reached for this session.${NC}"
+                echo "Manual intervention needed. Check: cd $project_dir && docker compose logs $svc_name"
+                sleep "$poll_interval"
+                continue 2
+            fi
 
-                    # Rebuild and restart
-                    echo -e "${CYAN}Rebuilding $svc_name...${NC}"
-                    (cd "$project_dir" && docker compose up -d --build --no-deps "$svc_name" 2>&1) | while IFS= read -r fline; do echo "  $fline"; done
+            echo -e "\n${RED}[FAILURE] $svc_name is down${NC}"
 
-                    fix_count=$((fix_count + 1))
-                    echo -e "${GREEN}Fix attempt $fix_count/$max_fixes for $svc_name${NC}"
-                else
-                    echo -e "${YELLOW}Max fix attempts ($max_fixes) reached. Manual intervention needed.${NC}"
+            local logs
+            logs=$(cd "$project_dir" && docker compose logs --tail 50 "$svc_name" 2>/dev/null | head -c 3000)
+
+            # Gather docker-compose.yml for AI context (truncated to avoid ARG_MAX)
+            local compose_content=""
+            if [[ -f "$project_dir/docker-compose.yml" ]]; then
+                compose_content=$(head -c 5000 "$project_dir/docker-compose.yml")
+            elif [[ -f "$project_dir/docker-compose.yaml" ]]; then
+                compose_content=$(head -c 5000 "$project_dir/docker-compose.yaml")
+            fi
+
+            # Gather Dockerfile for the failing service (truncated)
+            local dockerfile_content=""
+            for df_path in "$project_dir/$svc_name/Dockerfile" "$project_dir/Dockerfile" "$project_dir/Dockerfile.$svc_name"; do
+                if [[ -f "$df_path" ]]; then
+                    dockerfile_content=$(head -c 3000 "$df_path")
+                    break
                 fi
+            done
+
+            # 3. FEED TO AI: Construct a rich prompt with all context
+            local ai_prompt="You are debugging a Docker Compose service that has failed.
+
+SERVICE: $svc_name
+STATUS: exited/dead
+
+DOCKER COMPOSE LOGS (last 50 lines):
+$logs
+
+DOCKER-COMPOSE.YML:
+$compose_content"
+
+            if [[ -n "$dockerfile_content" ]]; then
+                ai_prompt="${ai_prompt}
+
+DOCKERFILE ($svc_name/Dockerfile):
+$dockerfile_content"
             fi
-        done <<< "$status"
 
-        if [[ "$has_failure" == "false" ]]; then
-            echo -e "\r${GREEN}All services healthy${NC} $(date +%H:%M:%S)  \c"
-        fi
+            ai_prompt="${ai_prompt}
+
+INSTRUCTIONS:
+1. Analyze the error in the logs above
+2. Identify the root cause
+3. Fix the issue by editing the necessary files (docker-compose.yml, Dockerfile, source code, package.json, requirements.txt, etc.)
+4. Make sure the fix works on any platform (Docker Desktop, Linux, Docker-in-Docker, Kubernetes)
+5. Do NOT just restart -- actually fix the underlying code/config problem
+6. After fixing, the system will rebuild with 'docker compose up --build'
+7. Common issues: named volumes for node_modules (use anonymous), missing dependencies, port conflicts, wrong commands"
+
+            echo -e "${CYAN}[AI] Analyzing $svc_name failure with ${LOKI_PROVIDER:-claude}...${NC}"
+
+            # 4. LET AI FIX: Run loki quick with the AI prompt
+            # The AI provider (claude/codex/gemini/ollama) decides what to fix
+            fix_count=$((fix_count + 1))
+            echo -e "${CYAN}[FIX] Attempt $fix_count/$max_fixes${NC}"
+
+            (
+                cd "$project_dir"
+                LOKI_MAX_ITERATIONS=5 LOKI_AUTO_FIX=true \
+                "$0" quick "$ai_prompt" 2>&1 | while IFS= read -r fline; do
+                    echo "  $fline"
+                done
+            )
+
+            # 5. REBUILD: Docker compose up --build for the fixed service
+            echo -e "${CYAN}[REBUILD] Rebuilding $svc_name...${NC}"
+            (cd "$project_dir" && docker compose up -d --build --no-deps "$svc_name" 2>&1) | while IFS= read -r fline; do
+                echo "  $fline"
+            done
+
+            # 6. VERIFY: Wait and check if the fix worked
+            echo -e "${CYAN}[VERIFY] Waiting 15s for $svc_name to stabilize...${NC}"
+            sleep 15
+
+            local verify_state
+            verify_state=$(cd "$project_dir" && docker compose ps -a --format json 2>/dev/null | VERIFY_SVC="$svc_name" python3 -c "
+import json, sys, os
+svc = os.environ['VERIFY_SVC']
+raw = sys.stdin.read().strip()
+if not raw:
+    print('unknown')
+    sys.exit(0)
+# Handle both JSON array (v2.21+) and NDJSON formats
+try:
+    parsed = json.loads(raw)
+    services = parsed if isinstance(parsed, list) else [parsed]
+except json.JSONDecodeError:
+    services = []
+    for line in raw.split(chr(10)):
+        try: services.append(json.loads(line))
+        except: pass
+for s in services:
+    if not isinstance(s, dict): continue
+    if s.get('Service', s.get('Name','')) == svc:
+        print(s.get('State','unknown'))
+        sys.exit(0)
+print('unknown')
+" 2>/dev/null || echo "unknown")
+
+            if [[ "$verify_state" == "running" ]]; then
+                echo -e "${GREEN}[SUCCESS] $svc_name is now running!${NC}"
+            else
+                echo -e "${YELLOW}[RETRY] $svc_name still not healthy (state: $verify_state). Will retry on next poll.${NC}"
+            fi
+        done
 
-        sleep 10
+        sleep "$poll_interval"
     done
 }
 

package/autonomy/run.sh

@@ -630,6 +630,8 @@ if [ -f "$TELEMETRY_SCRIPT" ]; then
     source "$TELEMETRY_SCRIPT"
 fi
 
+
+
 # 2026 Research Enhancements (minimal additions)
 PROMPT_REPETITION=${LOKI_PROMPT_REPETITION:-true}
 CONFIDENCE_ROUTING=${LOKI_CONFIDENCE_ROUTING:-true}

package/dashboard/__init__.py

@@ -7,7 +7,7 @@ Modules:
     control: Session control API (start/stop/pause/resume)
 """
 
-__version__ = "6.67.3"
+__version__ = "6.68.0"
 
 # 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:** v6.67.3
+**Version:** v6.68.0
 
 ---
 

package/mcp/__init__.py

@@ -57,4 +57,4 @@ try:
 except ImportError:
     __all__ = ['mcp']
 
-__version__ = '6.67.3'
+__version__ = '6.68.0'

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "loki-mode",
-  "version": "6.67.3",
+  "version": "6.68.0",
   "description": "Loki Mode by Autonomi - Multi-agent autonomous startup system for Claude Code, Codex CLI, and Gemini CLI",
   "keywords": [
     "agent",

package/web-app/server.py

@@ -1481,7 +1481,7 @@ class DevServerManager:
                 if framework == "docker":
                     await asyncio.to_thread(
                         subprocess.run,
-                        ["docker", "compose", "up", "-d", "--build"],
+                        ["docker", "compose", "up", "-d", "--build", "--no-deps"],
                         capture_output=True, cwd=str(target), timeout=120
                     )
                 # Restart the dev server
@@ -1732,17 +1732,17 @@ class DevServerManager:
                 was_running = prev.get("status") in ("running", None)
                 now_failed = svc["state"] in ("exited", "dead")
 
-                # Also detect services that never started (failed on initial docker compose up).
-                # On first poll, prev is empty ({}) so prev.get("status") is None, which counts
-                # as "was_running" above. But if a service was already "exited" before we ever
-                # saw it as "running", the next poll would see prev.status="exited" and skip it.
-                # Detect this case: first time we see a service in a failed state with no prior fix.
-                first_seen_failed = (
-                    not prev  # No previous health record for this service
-                    and now_failed
+                # Detect services that need fixing:
+                # 1. was_running and now_failed: service transitioned from running to exited
+                # 2. Persistently failed: service is exited AND has never been successfully fixed
+                #    (fix_attempts == 0 or fix_status != "fixed")
+                persistently_failed = (
+                    now_failed
+                    and prev.get("fix_status") != "fixed"
+                    and prev.get("status") in ("exited", "dead", None)
                 )
 
-                if (was_running and now_failed) or first_seen_failed:
+                if (was_running and now_failed) or persistently_failed:
                     svc_health["restarts"] = prev.get("restarts", 0) + 1
                     logger.warning("Docker service '%s' failed (exit %s)", name, svc.get("exit_code"))
 
@@ -1766,14 +1766,51 @@ class DevServerManager:
 
                         info["_auto_fixing"] = True
                         svc_logs = docker_ctx.get("service_logs", {}).get(name, "")
-                        diagnoses = _diagnose_errors(svc_logs)
-                        diag_text = "\n".join(f"- {d['diagnosis']}: {d['suggestion']}" for d in diagnoses)
-
-                        fix_prompt = f"The '{name}' Docker service crashed (exit code {svc.get('exit_code', 1)}).\n"
-                        if diag_text:
-                            fix_prompt += f"\nDiagnosis:\n{diag_text}\n"
-                        fix_prompt += f"\nService logs:\n{svc_logs[:2000]}\n"
-                        fix_prompt += "\nFix the issue and ensure the service starts correctly."
+
+                        # Read compose file for AI context
+                        compose_content = ""
+                        compose_file = project_dir / "docker-compose.yml"
+                        if not compose_file.exists():
+                            compose_file = project_dir / "docker-compose.yaml"
+                        if compose_file.exists():
+                            try:
+                                compose_content = compose_file.read_text(errors="replace")[:5000]
+                            except OSError:
+                                pass
+
+                        # Read service Dockerfile for AI context
+                        dockerfile_content = ""
+                        for df_path in [project_dir / name / "Dockerfile", project_dir / "Dockerfile"]:
+                            if df_path.exists():
+                                try:
+                                    dockerfile_content = df_path.read_text(errors="replace")[:3000]
+                                    break
+                                except OSError:
+                                    pass
+
+                        fix_prompt = f"""You are debugging a Docker Compose service that has failed.
+
+SERVICE: {name}
+STATUS: exited/dead (exit code {svc.get('exit_code', 1)})
+
+DOCKER COMPOSE LOGS (last 50 lines):
+{svc_logs[:3000]}
+
+DOCKER-COMPOSE.YML:
+{compose_content}
+"""
+                        if dockerfile_content:
+                            fix_prompt += f"\nDOCKERFILE ({name}/Dockerfile):\n{dockerfile_content}\n"
+
+                        fix_prompt += """
+INSTRUCTIONS:
+1. Analyze the error in the logs above
+2. Identify the root cause
+3. Fix the issue by editing the necessary files (docker-compose.yml, Dockerfile, source code, package.json, requirements.txt, etc.)
+4. Make sure the fix works on any platform (Docker Desktop, Linux, Docker-in-Docker, Kubernetes)
+5. Do NOT just restart -- actually fix the underlying code/config problem
+6. After fixing, the system will rebuild with 'docker compose up --build'
+7. Common issues: named volumes for node_modules (use anonymous), missing dependencies, port conflicts, wrong commands"""
 
                         svc_health["fix_attempts"] += 1
                         svc_health["fix_timestamps"] = recent_fixes + [now]