* Allow bridge sessions to run concurrently
* Stabilize bridge concurrency test
* Set bridge approval timeout to 120 seconds
* harden bridge approval concurrency
---------
Co-authored-by: Codex <codex@openai.com>
* fix: recursive skill scan for nested sub-category directories
The Web UI skill scanner (scanSkillsDir) only checked one level deep:
skills/<category>/<subdir>/SKILL.md. Sub-category containers like
mlops/evaluation/ (which has DESCRIPTION.md + subdirs but no SKILL.md)
were skipped entirely, hiding all 12 nested skills under mlops.
Changes:
- scanSkillsDir: extract collectSkills() recursive helper that depth-first
searches for SKILL.md at every level under a category directory.
Directories without SKILL.md but with subdirectories are recursively
descended into.
- listFiles handler: replace hardcoded join(category, skill) path with
recursive findSkillDir() search, so nested skill file browsing works
(e.g., mlops/evaluation/lm-evaluation-harness).
Fixes mlops category showing 1 skill instead of 13. All 20 other
categories verified with zero regression.
* fix: pA handler also needs recursive search for nested skill file content
The readFile_ (pA) handler was constructing direct paths like
skills/category/skill/... which fails for nested sub-category
skills (mlops/evaluation/lm-evaluation-harness). Added fallback
recursive search when direct path returns 404.
Also fixed listFiles (sA) handler to use recursive search for
the same reason - previous fix to dist was not in source TS.
Verified:
- lm-evaluation-harness SKILL.md content: 200 ✅
- vllm SKILL.md: 200 ✅
- huggingface-hub (non-nested): 200 ✅
- reference file in nested skill: 200 ✅
* fix: pA handler also needs recursive search for nested skill file content
The readFile_ (pA) handler was constructing direct paths like
skills/category/skill/... which fails for nested sub-category
skills (mlops/evaluation/lm-evaluation-harness). Added fallback
recursive search when direct path returns 404.
Also fixed listFiles (sA) handler to use recursive search for
the same reason - previous fix to dist was not in source TS.
Verified:
- lm-evaluation-harness SKILL.md content: 200 ✅
- vllm SKILL.md: 200 ✅
- huggingface-hub (non-nested): 200 ✅
- reference file in nested skill: 200 ✅
* harden recursive skill lookup
---------
Co-authored-by: gs <gs@localhost>
Co-authored-by: gutanulaif <gutanulaifa@gmail.com>
* feat: support profile-aware group chat bridge flows
* feat: route cron jobs through hermes cli
* Fix group chat routing and isolate bridge tests
* Add Grok image-to-video media skill
* Default Grok videos to media directory
* Fix bridge profile fallback and cron repeat clearing
* Refine bridge chat and gateway platform handling
* Filter bridge tool-call text deltas
* Preserve structured bridge chat history
* Prepare beta release build artifacts
* Fix Windows run profile resolution
* Fix Windows path compatibility checks
* Fix profile-scoped model page display
* Hide Windows subprocess windows for jobs and updates
* Hide Windows file backend subprocess windows
* Avoid Windows gateway restart lock conflicts
* Treat Windows gateway lock as running on startup
* Force release Windows gateway lock on restart
* Tighten Windows gateway lock cleanup
* Update chat e2e source expectation
* Bump package version to 0.5.30
---------
Co-authored-by: Codex <codex@openai.com>
Revert the dynamic import() for i18n locales and highlight.js core+
registration from #696. Dynamic imports create separate chunk files
that cause 404 errors for users after updating when the browser still
references old chunk hashes.
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
When starting gateways, automatically switch to 'default' profile if
current profile is not 'default'. This ensures consistent gateway startup
behavior and prevents port conflicts from different profile configurations.
Changes:
- Check current profile before starting gateways
- Execute 'hermes profile use default' if needed
- Wait 2 seconds for profile switch to take effect
- Update internal GatewayManager state
- Add detailed logging for the switching process
* feat(bridge): refactor compression to use DB history and add structured logging
- Extract buildDbHistory() to share message loading between buildCompressedHistory and forceCompressBridgeHistory
- forceCompressBridgeHistory now reads from local DB instead of using Python-provided messages, ensuring consistency with api_server path
- Pass sessionId to compressor for snapshot-aware compression
- Add force_compress flag to bridge chat requests
- Add bridgeLogger structured logging for compression lifecycle
- Simplify schemas, session-sync, and providers
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
* fix bridge compression history handling
---------
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
* feat: add CLI chat sessions with Python agent bridge
Introduce a new CLI chat mode that connects Web UI directly to Hermes
Agent's AIAgent via a Python bridge subprocess and Socket.IO, bypassing
the API Server /v1/responses path. Supports streaming, slash commands
(/new, /undo, /retry, /branch, /compress, /save, /title), interrupt,
and steer.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
* feat: update CLI chat session bridge
* fix: extend agent bridge startup timeouts
* docs: update bridge chat session design
* feat: align bridge compression and provider registry
* chore: bump version to 0.5.20
---------
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
The regex `^(\w[\w\s]*?)` could not match keys containing dots (`.env`,
`SOUL.md`), so `hasEnv` and `hasSoulMd` were always `undefined` → false.
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
Add comprehensive debug logging throughout the gateway lifecycle to
help troubleshoot nodemon restart issues on Windows, where SIGTERM
is used instead of SIGUSR2.
Changes:
- Enhanced shutdown handler to log all signals and env var states
- Gateway manager now logs process detachment mode explicitly
- Added environment variable confirmation on bootstrap
- Updated gateway-development.md with new debug logs and troubleshooting steps
Benefits:
- Easier troubleshooting of gateway lifecycle issues
- Clear visibility into signal handling during nodemon restarts
- Better cross-platform development experience
- Production behavior remains unchanged
Testing:
- ✅ Windows: Gateways persist across nodemon restarts
- ✅ macOS/Linux: Existing SIGUSR2 behavior preserved
- ✅ Production: Default shutdown cleanup unchanged
- ✅ Backward compatibility: No breaking changes
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Set HERMES_WEB_UI_STOP_GATEWAYS_ON_SHUTDOWN=0 in nodemon.json so
nodemon restarts don't kill running gateways. Production behavior
unchanged (stops owned gateways by default).
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
- Refactor port detection into reusable getListeningPids/killListeningPids
- Add ss command fallback when lsof is unavailable
- Extract readPidFile for cleaner PID file reading
- Remove unnecessary PowerShell candidate from Windows shell detection
- Add PID validity check in gateway_state.json fallback (Number.isFinite + > 0)
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
* refactor(docker): merge two-container setup into single container
The Web UI already manages the Hermes Agent gateway lifecycle internally
via GatewayManager (spawn hermes gateway run --replace), making the
separate hermes-agent container redundant. The Dockerfile is built on
the hermes-agent base image, so all CLI tooling is already included.
- Remove hermes-agent service and shared volume from docker-compose.yml
- Remove gateway port mapping (8642-8670) — internal-only now
- Update docs/docker.md, README.md, README_zh.md for single-container setup
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
* docs: remove version tag from image references
Use ekkoye8888/hermes-web-ui instead of ekkoye8888/hermes-web-ui:latest
to avoid pinning a specific version in documentation.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
- Emit tool.started on response.output_item.added instead of .done
- Track startedAt timestamp for each tool call to calculate duration
- Include duration (2 decimal places) and error status in tool.completed
- Fix response.completed fallback to emit tool.started before tool.completed
- Update website license from MIT to BSL-1.1
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
Prevents infinite retry loop on /api/hermes/available-models when
accessing the app without a token, which triggers server-side 429
rate limiting. Both loadModels() and fetchProviders() now bail out
early if hasApiKey() returns false.
Closes#606
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
- copilot-models.ts: add windowsHide to `gh auth token` call
- file-provider.ts: add shared execOpts with windowsHide for all
docker/ssh/singularity calls
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
Add missing keys: removeCustomModel, aliasEdit/Title/Placeholder/etc
(10 alias keys), and visibilitySelectOne to match zh.ts.
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
Expand changelog from 5 to 25 entries covering all changes:
comic theme, kanban parity, model visibility, TTS settings,
Windows/Termux compat, YAML parsing, gateway ownership, etc.
Also backfill zh-TW locale with missing keys (changelog, model
visibility, TTS rate/pitch).
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
Use independent OPENCODE_ZEN_API_KEY and OPENCODE_GO_API_KEY
instead of shared OPENCODE_API_KEY, preventing cross-provider config coupling.
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
* fix: comprehensive Windows compatibility and gateway management improvements
This commit addresses multiple Windows compatibility issues and improves
gateway management across all platforms.
## Windows Compatibility Fixes
- Add hermes-path.ts with cross-platform Hermes home/bin detection
- Fix Windows native installation paths (%LOCALAPPDATA%\hermes)
- Update terminal.ts to use PowerShell instead of /bin/bash on Windows
- Fix upload.ts path construction to use path.join() for cross-platform paths
- Fix download.ts to use isAbsolute() for Windows absolute path detection
- Update auth.ts to skip file mode 0o600 on Windows (unsupported)
- Add nodemon.json for cross-platform environment variable handling
## Gateway Management Improvements
- Simplify gateway startup: all platforms use 'run' mode uniformly
- Remove complex init system detection and platform-specific code paths
- Improve PID file validation: use health check instead of port detection
- Remove getPortByPid() method (too complex and error-prone)
- Remove checkPortAvailable() TCP bind test (TIME_WAIT false positives)
- Trust gateway --replace flag to handle real port conflicts
- Add smart PID validation: check if stale process via health check
- Fix port allocation to avoid incrementing when gateway restarts
- Add allocatedPorts.clear() on each startAll() call
- Add clearPidFile() method to clean up stale PID files
## Process Management
- Remove detached:true and unref() from gateway spawn
- Gateway processes now follow parent process lifecycle
- Add process reference storage in ManagedGateway interface
- Improve shutdown logic: call gatewayManager.stopAll() before exit
- Fix Windows process killing: use process.kill(pid) for Windows
- Remove PowerShell command for lock file cleanup (use Node.js fs.unlinkSync)
## Frontend Theme Fixes
- Fix main.ts localStorage key mismatch (hermes_theme → hermes_brightness)
- Add inline script in index.html to prevent FOUC (Flash of Unstyled Content)
- Apply theme classes before Vue mount to avoid visual glitches
## Developer Experience
- Fix nodemon windows-kill popup on Windows by removing signal config
- Add delay and environment variables to nodemon.json
- Add windowsHide: true to all child process spawns
## Breaking Changes
- Gateway management now exclusively uses 'run' mode on all platforms
- systemd/launchd integration removed (use --replace flag instead)
This fix ensures hermes-web-ui works correctly on Windows native
installations while maintaining compatibility with Linux/macOS/WSL2.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* Fix gateway lifecycle port handling
* fix: comprehensive Windows compatibility and gateway management improvements
- Simplified hermes CLI binary resolution logic
- Fixed Windows line ending compatibility in profile list parsing
- Migrated gateway restart logic from CLI to GatewayManager
- Added gateway restart to updateCredentials method
- Removed unnecessary gateway restarts from provider operations
- Fixed configuration preservation when switching profiles
- Added nodemon quiet mode and legacy watch to reduce Windows popups
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* revert: change back to nodemon due to tsx compatibility issues
- tsx has compatibility issues with Koa generator functions
- Restored nodemon with simplified configuration
- Added cross-env package for future Windows environment variable needs
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* feat: replace nodemon with ts-node-dev to eliminate Windows popup windows
- Installed ts-node-dev as nodemon replacement
- ts-node-dev has better Windows compatibility without console popups
- Supports respawning, inspector debugging, and TypeScript compilation
- Uses cross-env for Windows environment variable support
- Removed nodemon.json configuration file (no longer needed)
Benefits:
- No more Windows console popup windows during development
- Faster restart times compared to nodemon
- Built-in TypeScript compilation without ts-node overhead
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix: improve log parsing and Windows compatibility for agent/error logs
- Fixed Pino JSON log parsing bug where logger field incorrectly used obj.msg
- Changed logger field to use obj.name to properly display log source
- Added Windows line ending support (\r\n) for log file listing
- Added support for 'error' log type in addition to 'errors'
- Improved error message extraction from obj.err when available
This fixes the missing agent and error logs issue on Windows.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* Fix gateway health checks and shutdown ownership
* Refine auth lock window and dev shutdown
* fix: improve Hermes plugin discovery on Windows by fixing Python path resolution
- Added support for Windows venv Scripts directory structure
- Fixed Python executable path detection for hermes.exe in venv/Scripts/
- Added Windows LOCALAPPDATA hermes-agent directory to search paths
- Improved cross-platform compatibility for plugin discovery
This fixes the "No module named 'hermes_cli'" error on Windows by correctly
locating the Python virtual environment that contains the Hermes modules.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* refactor: improve cross-platform compatibility for Hermes plugin discovery
- Added platform detection to only add Windows-specific paths on Windows
- Prevents potential issues on Unix/Linux/macOS systems
- Ensures LOCALAPPDATA path is only used when available on Windows
- Maintains existing behavior for all platforms
This makes the Windows plugin discovery fix safer for cross-platform usage.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* chore: remove unused development dependencies
- Removed nodemon (replaced by ts-node-dev)
- Removed tsx (had compatibility issues with Koa)
- Removed nodemon.json configuration file
- Cleaned up development tools to only what's actually used
This reduces dependency size and eliminates the windows-kill popup
source that was part of nodemon.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* chore: remove memory system files
- Removed MEMORY.md index file
- Removed memory/ directory and windows-compatibility.md
- Cleaned up unused memory persistence system
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix: resolve TypeScript compilation error in plugins.ts
- Added type assertion 'as string[]' after filter(Boolean)
- Fixes TS2769 error: No overload matches this call
- Ensures type compatibility with hasHermesPluginModule function
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix: comprehensive Windows compatibility and gateway management improvements
- Fix gateway detection after nodemon restart by adding health check-based detection
- Prevent port conflicts by detecting already-running gateways without PID files
- Switch to serial gateway startup to avoid lock file race conditions
- Return to nodemon from ts-node-dev for development stability
- Always stop gateways on shutdown to prevent orphan processes
- Prevent project root config files from being committed to git
- Fix syntax issues in plugins.ts
Resolves issues where default profile gateway failed to start after
nodemon restart and gateways were incorrectly marked as stopped
despite running on correct ports.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* feat: comic theme multilingual fonts, sidebar collapse fix, plugin discovery for Termux, and cron history
- Add Chinese (ZCOOL KuaiLe), Japanese (Zen Maru Gothic), Korean (Gaegu) handwritten fonts for Comic theme
- Fix collapsed sidebar: hide language switch, stack theme icons vertically
- Add hermes shebang parsing as fallback Python discovery for Termux
- Remove cron source filter from history sessions
- Add 0.5.17 changelog entries for all 8 locales
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
* fix: tolerate duplicate YAML keys in config parsing (closes#628)
Add `{ json: true }` to all 7 `yaml.load()` calls so duplicated mapping
keys (e.g. multiple `mcp_servers:` blocks) no longer crash the API.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
* fix: gateway ownership check requires PID file to prevent cross-profile port hijacking
Remove fallback that assumed ownership of healthy gateways without PID
verification. Now only claims a gateway if PID file exists and process
is alive, preventing one profile from hijacking another's port.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Reduce the message count threshold that triggers LLM-based context
compression to avoid excessively long histories before compression kicks in.
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
* fix: lower context compression message threshold from 200 to 150
Reduce the message count threshold that triggers LLM-based context
compression to avoid excessively long histories before compression kicks in.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
* feat: convert image uploads to base64 multimodal format for API
Images sent by users are now read from disk, converted to base64 data
URLs, and sent as input_image parts in the /v1/responses API request
instead of being replaced with text placeholders. File attachments remain
as text mentions.
- convertContentBlocks returns multimodal array instead of plain text
- Input is wrapped in [{role:"user", content:[...]}] format for gateway
- History conversion extracts text only (no base64 in conversation_history)
- Add debug logging for request input preview
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
* chore: remove debug console.log from chat-run-socket
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
- Switch from MIT to Business Source License 1.1 (non-commercial use,
auto-converts to Apache 2.0 on 2029-05-10)
- Remove third-party brand keywords from package.json (chatgpt, claude,
openai, gemini, deepseek, telegram, discord, slack, whatsapp, etc.)
- Remove afdian sponsor links from README.md and README_zh.md
- Update license references in both READMEs to BSL-1.1
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>