Set unique PLAYWRIGHT_CLI_SESSION environment variable for each spawned
agent subprocess to prevent concurrent agents from sharing a single
browser instance and interfering with each other's navigation.
- _spawn_coding_agent: session named "coding-{feature_id}"
- _spawn_coding_agent_batch: session named "coding-{primary_id}"
- _spawn_testing_agent: session named "testing-{counter}" using an
incrementing counter (since multiple testing agents can test
overlapping features, feature ID alone isn't sufficient)
Previously, after migrating from Playwright MCP to CLI, all parallel
agents shared the default browser session, causing them to navigate
away from each other's pages.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Address three issues reported after overnight AutoForge runs:
1. ~193GB of .node files in %TEMP% from V8 compile caching
2. Stale npm artifact folders on drive root when %TEMP% fills up
3. PNG screenshot files left in project root by Playwright
Changes:
- Widen .node cleanup glob from ".78912*.node" to ".[0-9a-f]*.node"
to match all V8 compile cache hex prefixes
- Add "node-compile-cache" directory to temp cleanup patterns
- Set NODE_COMPILE_CACHE="" in all subprocess environments (client.py,
parallel_orchestrator.py, process_manager.py) to disable V8 compile
caching at the source
- Add cleanup_project_screenshots() to remove stale .png files from
project directories (feature*-*.png, screenshot-*.png, step-*.png)
- Run cleanup_stale_temp() at server startup in lifespan()
- Add _run_inter_session_cleanup() to orchestrator, called after each
agent completes (both coding and testing paths)
- Update coding and testing prompt templates to instruct agents to use
inline (base64) screenshots only, never saving files to disk
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
File-based signal (.pause_drain) lets the orchestrator finish current
work before pausing instead of hard-freezing the process tree. New
status states pausing/paused_graceful flow through WebSocket to the UI
where a Pause button, draining indicator, and Resume button are shown.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Complete project rebrand from AutoCoder to AutoForge, touching 62 files
across Python backend, FastAPI server, React UI, documentation, config,
and CI/CD.
Key changes:
- Rename autocoder_paths.py -> autoforge_paths.py with backward-compat
migration from .autocoder/ -> .autoforge/ directories
- Update registry.py to migrate ~/.autocoder/ -> ~/.autoforge/ global
config directory with fallback support
- Update security.py with fallback reads from legacy .autocoder/ paths
- Rename .claude/commands and skills from gsd-to-autocoder-spec to
gsd-to-autoforge-spec
- Update all Python modules: client, prompts, progress, agent,
orchestrator, server routers and services
- Update React UI: package.json name, index.html title, localStorage
keys, all documentation sections, component references
- Update start scripts (bat/sh/py), examples, and .env.example
- Update CLAUDE.md and README.md with new branding and paths
- Update test files for new .autoforge/ directory structure
- Transfer git remote from leonvanzyl/autocoder to
AutoForgeAI/autoforge
Backward compatibility preserved: legacy .autocoder/ directories are
auto-detected and migrated on next agent start. Config fallback chain
checks both new and old paths.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
The parallel orchestrator was using AUTOCODER_ROOT as the working
directory when spawning coding, batch, and testing agent subprocesses.
This caused the Claude Code CLI to create .claude/ and .claude_worktrees/
directories in the autocoder installation folder instead of the project
directory, scattering output files across multiple locations.
Changed all 3 subprocess spawn sites (coding agent, batch agent, testing
agent) to use self.project_dir as cwd, matching the behavior of the
server's process_manager.py. The subprocess commands already use absolute
paths to autonomous_agent_demo.py, so Python imports are unaffected.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
The output reader was stamping every line with the primary feature ID
(e.g., [Feature #24]) even after the agent claimed a new feature in the
batch. Now parses feature_claim_and_get calls in the output stream and
switches the prefix to the newly claimed feature ID, so logs correctly
show [Feature #30] once the agent moves on.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Enable the orchestrator to assign 1-3 features per coding agent subprocess,
selected via dependency chain extension + same-category fill. This reduces
cold-start overhead and leverages shared context across related features.
Orchestrator (parallel_orchestrator.py):
- Add batch tracking: _batch_features and _feature_to_primary data structures
- Add build_feature_batches() with dependency chain + category fill algorithm
- Add start_feature_batch() and _spawn_coding_agent_batch() methods
- Update _on_agent_complete() for batch cleanup across all features
- Update stop_feature() with _feature_to_primary lookup
- Update get_ready_features() to exclude all batch feature IDs
- Update main loop to build batches then spawn per available slot
CLI and agent layer:
- Add --feature-ids (comma-separated) and --batch-size CLI args
- Add feature_ids parameter to run_autonomous_agent() with batch prompt selection
- Add get_batch_feature_prompt() with sequential workflow instructions
WebSocket layer (server/websocket.py):
- Add BATCH_CODING_AGENT_START_PATTERN and BATCH_FEATURES_COMPLETE_PATTERN
- Add _handle_batch_agent_start() and _handle_batch_agent_complete() methods
- Add featureIds field to all agent_update messages
- Track current_feature_id updates as agent moves through batch
Frontend (React UI):
- Add featureIds to ActiveAgent and WSAgentUpdateMessage types
- Update KanbanColumn and DependencyGraph agent-feature maps for batch
- Update AgentCard to show "Batch: #X, #Y, #Z" with active feature highlight
- Add "Features per Agent" segmented control (1-3) in SettingsModal
Settings integration (full stack):
- Add batch_size to schemas, settings router, agent router, process manager
- Default batch_size=3, user-configurable 1-3 via settings UI
- batch_size=1 is functionally identical to pre-batching behavior
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Replace ineffective threading.Lock() with atomic SQL operations for
cross-process safety. Key changes:
- Add SQLAlchemy event hooks (do_connect/do_begin) for BEGIN IMMEDIATE
transactions in api/database.py
- Add atomic_transaction() context manager for multi-statement ops
- Convert all feature MCP write operations to atomic UPDATE...WHERE
with compare-and-swap patterns (feature_claim, mark_passing, etc.)
- Add WHERE passes=0 state guard to feature_mark_passing
- Add WAL checkpoint on shutdown and idempotent cleanup() in
parallel_orchestrator.py with async-safe signal handling
- Wrap SQLite connections with contextlib.closing() in progress.py
- Add thread-safe engine cache with double-checked locking in
assistant_database.py
- Migrate to SQLAlchemy 2.0 DeclarativeBase across all modules
Inspired by PR #108 (cabana8471-arch), with fixes for nested
BEGIN EXCLUSIVE bug and missing state guards.
Closes#106
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add comment on running_coding_agents explaining why feature_id keying
is safe (start_feature checks for duplicates before spawning), since
the sister dict running_testing_agents required PID keying to avoid
overwrites from concurrent same-feature testing
- Clear running_testing_agents dict in stop_all() after killing
processes so get_status() doesn't report stale agent counts while
_on_agent_complete callbacks are still in flight
Follow-up to PR #130 (runaway testing agent spawn fix).
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
running_testing_agents was keyed by feature_id, so when multiple agents
tested the same feature, each spawn overwrote the previous dict entry.
The count stayed at 1 regardless of how many processes were actually
running, causing the maintain loop to spawn agents indefinitely (~130+).
Re-key the dict by PID so each agent gets a unique entry and the
existing max-agent guards work correctly. Also check the return value
of _spawn_testing_agent() to break the loop on failure.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Follow-up to PR #89 - apply the same popen_kwargs pattern with
stdin=DEVNULL and CREATE_NO_WINDOW to _spawn_testing_agent() and
_run_initializer() for consistent Windows behavior.
Also fixes typo: _kill_process_tree -> kill_process_tree
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
When all features pass, the orchestrator continued spawning testing
agents for 10+ minutes, wasting tokens on unnecessary regression
tests. Added a check for get_all_complete() to prevent this.
Fixes: leonvanzyl/autocoder#66
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Added _kill_process_tree call in _read_output finally block to ensure
child processes (Claude CLI) are cleaned up when agents complete or fail.
This prevents accumulation of zombie processes that was causing 78+
Python processes when max concurrency was set to 5.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Resolved conflicts by combining:
- stdin=DEVNULL and CREATE_NO_WINDOW (blocking fix)
- PYTHONUNBUFFERED env var (output buffering fix)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add stdin=subprocess.DEVNULL to prevent blocking on stdin reads
- Add CREATE_NO_WINDOW flag on Windows to prevent console pop-ups
- Remove trailing pause from start_ui.bat
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Remove the testing_in_progress claim/release mechanism from the testing
agent architecture. Multiple testing agents can now test the same feature
concurrently, simplifying the system and eliminating potential stale lock
issues.
Changes:
- parallel_orchestrator.py:
- Remove claim_feature_for_testing() and release_testing_claim() methods
- Remove _cleanup_stale_testing_locks() periodic cleanup
- Replace with simple _get_random_passing_feature() selection
- Remove startup stale lock cleanup code
- Remove STALE_TESTING_LOCK_MINUTES constant
- Remove unused imports (timedelta, text)
- api/database.py:
- Remove testing_in_progress and last_tested_at columns from Feature model
- Update to_dict() to exclude these fields
- Convert _migrate_add_testing_columns() to no-op for backwards compat
- mcp_server/feature_mcp.py:
- Remove feature_release_testing tool entirely
- Remove unused datetime import
- prompts.py:
- Update testing prompt to remove feature_release_testing instruction
- Testing agents now just verify and exit (no cleanup needed)
- server/websocket.py:
- Update AgentTracker to use composite keys (feature_id, agent_type)
- Prevents ghost agent creation from ambiguous [Feature #X] messages
- Proper separation of coding vs testing agent tracking
Benefits:
- Eliminates artificial bottleneck from claim coordination
- No stale locks to clean up after crashes
- Simpler crash recovery (no testing state to restore)
- Reduced database writes (no claim/release transactions)
- Matches intended design: random, concurrent regression testing
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Replace agent-initiated feature selection with orchestrator pre-selection
for both coding and testing agents. This ensures Mission Control displays
correct feature numbers for testing agents (previously showed "Feature #0").
Key changes:
MCP Server (mcp_server/feature_mcp.py):
- Add feature_get_by_id tool for agents to fetch assigned feature details
- Remove obsolete tools: feature_get_next, feature_claim_next,
feature_claim_for_testing, feature_get_for_regression
- Remove helper functions and unused imports (text, OperationalError, func)
Orchestrator (parallel_orchestrator.py):
- Change running_testing_agents from list to dict[int, Popen]
- Add claim_feature_for_testing() with random selection
- Add release_testing_claim() method
- Pass --testing-feature-id to spawned testing agents
- Use unified [Feature #X] output format for both agent types
Agent Entry Points:
- autonomous_agent_demo.py: Add --testing-feature-id CLI argument
- agent.py: Pass testing_feature_id to get_testing_prompt()
Prompt Templates:
- coding_prompt.template.md: Update to use feature_get_by_id
- testing_prompt.template.md: Update workflow for pre-assigned features
- prompts.py: Update pre-claimed headers for both agent types
WebSocket (server/websocket.py):
- Simplify tracking with unified [Feature #X] pattern
- Remove testing-specific parsing code
Assistant (server/services/assistant_chat_session.py):
- Update help text with current available tools
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Major refactoring of the parallel orchestrator to run regression testing
agents independently from coding agents. This improves system reliability
and provides better control over testing behavior.
Key changes:
Database & MCP Layer:
- Add testing_in_progress and last_tested_at columns to Feature model
- Add feature_claim_for_testing() for atomic test claim with retry
- Add feature_release_testing() to release claims after testing
- Refactor claim functions to iterative loops (no recursion)
- Add OperationalError retry handling for transient DB errors
- Reduce MAX_CLAIM_RETRIES from 10 to 5
Orchestrator:
- Decouple testing agent lifecycle from coding agents
- Add _maintain_testing_agents() for continuous testing maintenance
- Fix TOCTOU race in _spawn_testing_agent() - hold lock during spawn
- Add _cleanup_stale_testing_locks() with 30-min timeout
- Fix log ordering - start_session() before stale flag cleanup
- Add stale testing_in_progress cleanup on startup
Dead Code Removal:
- Remove count_testing_in_concurrency from entire stack (12+ files)
- Remove ineffective with_for_update() from features router
API & UI:
- Pass testing_agent_ratio via CLI to orchestrator
- Update testing prompt template to use new claim/release tools
- Rename UI label to "Regression Agents" with clearer description
- Add process_utils.py for cross-platform process tree management
Testing agents now:
- Run continuously as long as passing features exist
- Can re-test features multiple times to catch regressions
- Are controlled by fixed count (0-3) via testing_agent_ratio setting
- Have atomic claiming to prevent concurrent testing of same feature
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
This fix addresses two root causes that caused testing agents to
accumulate (10-12 agents) instead of maintaining a 1:1 ratio with
coding agents:
1. Testing agents now exit after one session (agent.py)
- Added `or agent_type == "testing"` to the exit condition
- Previously, testing agents never hit the exit condition since
they're spawned with feature_id=None
2. Testing agents now spawn when coding agents START, not complete
- Moved spawn logic from _on_agent_complete() to start_feature()
- Removed the old spawn logic from _on_agent_complete()
- This ensures proper 1:1 ratio and prevents accumulation
Expected behavior after fix:
- First coding agent: no testing agent (no passing features yet)
- Subsequent coding agents: one testing agent spawns per start
- Each testing agent tests ONE feature then terminates immediately
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Introduce a new testing agent architecture that runs regression tests
independently from coding agents, improving quality assurance in
parallel mode.
Key changes:
Testing Agent System:
- Add testing_prompt.template.md for dedicated testing agent role
- Add feature_mark_failing MCP tool for regression detection
- Add --agent-type flag to select initializer/coding/testing mode
- Remove regression testing from coding prompt (now handled by testing agents)
Parallel Orchestrator Enhancements:
- Add testing agent spawning with configurable ratio (--testing-agent-ratio)
- Add comprehensive debug logging system (DebugLog class)
- Improve database session management to prevent stale reads
- Add engine.dispose() calls to refresh connections after subprocess commits
- Fix f-string linting issues (remove unnecessary f-prefixes)
UI Improvements:
- Add testing agent mascot (Chip) to AgentAvatar
- Enhance AgentCard to display testing agent status
- Add testing agent ratio slider in SettingsModal
- Update WebSocket handling for testing agent updates
- Improve ActivityFeed to show testing agent activity
API & Server Updates:
- Add testing_agent_ratio to settings schema and endpoints
- Update process manager to support testing agent type
- Enhance WebSocket messages for agent_update events
Template Changes:
- Delete coding_prompt_yolo.template.md (consolidated into main prompt)
- Update initializer_prompt.template.md with improved structure
- Streamline coding_prompt.template.md workflow
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
The parallel orchestrator was exiting prematurely with "All features
complete!" while pending features remained. This was caused by SQLAlchemy
session caching not seeing commits made by agent subprocesses.
Changes:
- Add session.expire_all() to get_resumable_features() to force fresh reads
- Add session.expire_all() to get_ready_features() to force fresh reads
- Add session.expire_all() to get_all_complete() to force fresh reads
- Add defensive retry logic in run_loop() when no features are ready
but nothing is running - now forces a fresh check before declaring blocked
- Add debug logging to get_all_complete() and get_ready_features() to
track passing/pending/in_progress counts for easier diagnosis
The root cause was cross-process database visibility: when an agent
subprocess committed feature completion, the orchestrator's session
had cached the old state and didn't see the update.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Remove unused imports and organize import statements to pass ruff
linting checks:
- mcp_server/feature_mcp.py: Remove unused imports (are_dependencies_satisfied,
get_blocking_dependencies) and alphabetize import block
- parallel_orchestrator.py: Remove unused imports (time, Awaitable) and
add blank lines between import groups per PEP 8
- server/routers/features.py: Alphabetize imports in dependency resolver
These changes were identified by running `ruff check .` and auto-fixed
with `--fix` flag.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Critical fixes:
- Lock file TOCTOU race condition: Use atomic O_CREAT|O_EXCL for lock creation
- PID reuse vulnerability on Windows: Store PID:CREATE_TIME in lock file to
detect when a different process has reused the same PID
- WAL mode on network drives: Detect network paths (UNC, mapped drives, NFS,
CIFS) and fall back to DELETE journal mode to prevent corruption
High priority fixes:
- JSON migration now preserves dependencies field during legacy migration
- Process tree termination on Windows: Use psutil to kill child processes
recursively to prevent orphaned browser instances
- Retry backoff jitter: Add random 30% jitter to prevent synchronized retries
under high contention with 5 concurrent agents
Files changed:
- server/services/process_manager.py: Atomic lock creation, PID+create_time
- api/database.py: Network filesystem detection for WAL mode fallback
- api/migration.py: Add dependencies field to JSON migration
- parallel_orchestrator.py: _kill_process_tree helper function
- mcp_server/feature_mcp.py: Add jitter to exponential backoff
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Changes:
- Add per-agent log viewer with copy-to-clipboard functionality
- New AgentLogEntry type for structured log entries
- Logs stored per-agent in WebSocket state (up to 500 entries)
- Log modal rendered via React Portal to avoid overflow issues
- Click log icon on agent card to view full activity history
- Fix agents getting stuck in "failed" state
- Wrap client context manager in try/except (agent.py)
- Remove failed agents from UI on error state (useWebSocket.ts)
- Handle permanently failed features in get_all_complete()
- Add friendlier agent state labels
- "Hit an issue" → "Trying plan B..."
- "Retrying..." → "Being persistent..."
- Softer colors (yellow/orange instead of red)
- Add scheduling scores for smarter feature ordering
- compute_scheduling_scores() in dependency_resolver.py
- Features that unblock others get higher priority
- Update CLAUDE.md with parallel mode documentation
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Major feature implementation for parallel agent execution with dependency-aware
scheduling and an engaging multi-agent UI experience.
Backend Changes:
- Add parallel_orchestrator.py for concurrent feature processing
- Add api/dependency_resolver.py with cycle detection (Kahn's algorithm + DFS)
- Add atomic feature_claim_next() with retry limit and exponential backoff
- Fix circular dependency check arguments in 4 locations
- Add AgentTracker class for parsing agent output and emitting updates
- Add browser isolation with --isolated flag for Playwright MCP
- Extend WebSocket protocol with agent_update messages and log attribution
- Add WSAgentUpdateMessage schema with agent states and mascot names
- Fix WSProgressMessage to include in_progress field
New UI Components:
- AgentMissionControl: Dashboard showing active agents with collapsible activity
- AgentCard: Individual agent status with avatar and thought bubble
- AgentAvatar: SVG mascots (Spark, Fizz, Octo, Hoot, Buzz) with animations
- ActivityFeed: Recent activity stream with stable keys (no flickering)
- CelebrationOverlay: Confetti animation with click/Escape dismiss
- DependencyGraph: Interactive node graph visualization with dagre layout
- DependencyBadge: Visual indicator for feature dependencies
- ViewToggle: Switch between Kanban and Graph views
- KeyboardShortcutsHelp: Help overlay accessible via ? key
UI/UX Improvements:
- Celebration queue system to handle rapid success messages
- Accessibility attributes on AgentAvatar (role, aria-label, aria-live)
- Collapsible Recent Activity section with persisted preference
- Agent count display in header
- Keyboard shortcut G to toggle Kanban/Graph view
- Real-time thought bubbles and state animations
Bug Fixes:
- Fix circular dependency validation (swapped source/target arguments)
- Add MAX_CLAIM_RETRIES=10 to prevent stack overflow under contention
- Fix THOUGHT_PATTERNS to match actual [Tool: name] format
- Fix ActivityFeed key prop to prevent re-renders on new items
- Add featureId/agentIndex to log messages for proper attribution
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
