- intent_compiler.rs: LLM-to-BASIC translation engine
- auto_task.rs: Auto-executing task data structures
- mcp_client.rs: Model Context Protocol server integration
- safety_layer.rs: Constraints, simulation, audit trail
- autotask_api.rs: HTTP API handlers for Auto Task UI
- Updated mod.rs with new modules and keyword list
- Serve suite UI from botui (dev) or botserver-stack (installed)
- SPA fallback to index.html for client-side routing
- Search paths: ../botui/ui/suite, ./botserver-stack/ui/suite, ./ui/suite
- When Vault unseal fails, only restart Vault - NOT full bootstrap
- Preserve PostgreSQL, Redis, MinIO, etc. when Vault needs re-init
- Simplify shutdown message to 3 lines with pragmatismo.com.br
- Never kill all stack processes just for Vault issues
- Display 'Thank you for using General Bots!' message
- Show version, links, and farewell
- Handle both SIGTERM and SIGINT (Ctrl+C)
- Graceful 10 second timeout for in-flight requests
- Call kill_stack_processes() at the beginning of bootstrap()
- Add port-based killing using fuser for ports 8200, 5432, 9000, etc.
- Use -f (pattern match) instead of -x (exact) for process names
- Fix process names: 'vault server', 'llama-server' instead of 'vault', 'ollama'
- Increase wait time to 1 second for processes to terminate
This fixes issues on dev machines where old processes from a deleted
stack are still running and blocking ports.
Instead of downloading pre-built binaries (which may require AVX2),
compile llama.cpp from source during installation. This ensures:
- Works on older CPUs (Sandy Bridge, Haswell, etc.)
- Uses GGML_NATIVE=ON to optimize for the current CPU
- Binary path updated to build/bin/llama-server
Reverts the AVX2 detection that was incorrectly disabling LLM.
- Add cpu_supports_avx2() function to check /proc/cpuinfo for AVX2 flag
- Skip LLM binary download on CPUs without AVX2 (pre-built llama.cpp requires it)
- Add runtime check for llama-server binary compatibility (catches 'Illegal instruction')
- Provide helpful error messages about alternatives (compile from source or use external API)
- Sandy Bridge (2nd gen i7) and older CPUs now gracefully skip local LLM
- Fix create_conn/establish_pg_connection to return Result instead of panicking
- Fix AppConfig::from_env to not require database access (circular dependency)
- Add #[cfg(test)] to AppState Default impl to prevent accidental panic
- Add extensive debug logging for Vault startup troubleshooting
- Remove Stdio::null() from start() to allow shell redirections to work
- Add direct vault start test in bootstrap for debugging
- Make Vault setup failure fatal (was silently continuing)
When start_all() detects Vault init.json is missing (unseal fails),
it now:
1. Kills all stack processes
2. Cleans the stack directory
3. Runs full bootstrap()
4. Returns (bootstrap starts all services)
Also made SecretsManager init failure in start_all() a hard error
instead of a warning - if Vault is supposed to be working but
SecretsManager can't connect, that's a fatal error.
The core issue was that .env was written at the END of setup_vault(),
but other components needed Vault credentials BEFORE that.
Now:
1. Unseal Vault
2. Write .env with VAULT_ADDR and VAULT_TOKEN immediately
3. Re-initialize SecretsManager so it connects to Vault
4. THEN store secrets in Vault
This ensures SecretsManager is properly configured before any
code tries to use create_conn() or other Vault-dependent functions.
System Logs:
- Add scroll_offset tracking with auto-scroll to bottom on new logs
- Up/Down/j/k keys to scroll line by line
- PageUp/PageDown for page scrolling
- Home/End to jump to top/bottom
- Show scroll indicators in title: [^v], [SCROLL] when not auto-scrolling
- Display log count in title
Editor:
- Fix scroll_offset to follow cursor when moving up/down
- Add PageUp/PageDown for faster navigation
- Add Ctrl+Home/Ctrl+End to jump to start/end of file
- ensure_cursor_visible() keeps cursor in view
Tab Navigation:
- FileTree -> Editor (if open) or Logs -> Chat -> back to start
- Consistent cycling through all panels
When services are already configured (bootstrap completed), the code
calls start_all() or ensure_services_running() which starts Vault
but didn't initialize SecretsManager. This caused create_conn() to
fail with 'Vault not configured' even though Vault was running.
Now both functions initialize SecretsManager after Vault is unsealed,
ensuring database connections can retrieve credentials from Vault.
Replace emoji icons with plain text/ASCII equivalents to prevent
encoding issues and improve console compatibility:
- Replace checkmarks with *
- Replace x marks with x
- Replace status indicators with text [OK], [ERR], etc.
- Remove decorative emojis from info/debug log messages
- Keep functional emojis in user-facing chat/sentiment features
The bootstrap order is vault -> tables -> directory. When tables
component was processed, it tried to call establish_pg_connection()
which requires Vault, but Vault setup only happens when processing
the vault component (which is earlier in the loop but the if-block
for tables runs for the tables component, not vault).
Fix: Run migrations directly with the generated password during the
tables component setup, before Vault is configured. This avoids the
dependency on Vault being set up.
- Redirect Vault exec_cmd output to logs/vault.log
- Add stdout/stderr null pipes to component spawn in installer.rs
- Suppress output in run_commands() in facade.rs
- All component output now goes to respective log files in logs/
When llm-server=false in bot_configuration, the code now skips
attempting to start local llama-server processes. This prevents
the 60-attempt timeout error when using external LLM endpoints
or when local LLM serving is intentionally disabled.
- Add state_channel field to XtreeUI for receiving AppState updates
- Add set_state_channel() method to enable async state communication
- Poll for AppState in event loop to initialize panels when ready
- UI now shows loading state instantly, logs stream in real-time
- Transitions to full interactive mode when AppState is received
- Don't delete archive files from cache after tar/zip extraction
- Copy binaries from cache instead of moving them
- Cached files are now preserved for offline installation reuse
- Disable TLS on Vault for local development (HTTP instead of HTTPS)
- Fix bot_configuration id column type mismatch (TEXT -> UUID)
- Add migration 6.1.1 to convert config table id columns to UUID
- Fix sync_config_csv_to_db to use UUID binding for id column
- Make start_all async with proper Vault startup sequence
- Sync default.gbai config.csv to existing 'Default Bot' from migrations
- Add diagnostic logging for config loading
- Change default LLM/embedding URLs from https to http for local dev
- Rename service names to component names:
- postgres -> tables
- minio -> drive
- redis -> cache
- qdrant -> vectordb
- mail -> email
- Update bootstrap certificate generation with new hostnames
- Update CoreDNS zone file with component-based DNS entries
- Update CA service certificates to use component names
- Update CA directory creation to use component names
All certificates include 127.0.0.1 as SAN for IP-based mTLS access
- Add create_vault_config() function to generate config.hcl with mTLS settings
- Configure Vault to require client certificate verification
- Generate client certificate for botserver in bootstrap
- Update .env to include mTLS paths (VAULT_CACERT, VAULT_CLIENT_CERT, VAULT_CLIENT_KEY)
- Remove unused import in tls.rs
- Add Default impl for AppState with mock services for testing
- Add MockLLMProvider for tests that need LLM without real API
- Add create_mock_auth_service() for Zitadel testing
- Add test_utils.rs with TestAppStateBuilder, MockChannelAdapter
- Enable rhai 'sync' feature to fix Send+Sync for Dynamic types
- Fix attendance.rs tests to use pure functions (no DB required)
- Fix on_error.rs tests to use String instead of i32
- Remove unused imports in attendance.rs
All tests pass, 0 warnings, 0 errors.
- Add email, teams, google columns to attendant.csv
- Add find_attendant_by_identifier() for multi-channel lookup
- Add find_attendants_by_channel() and find_attendants_by_department()
- Implement handle_status_command with database persistence
- Implement handle_transfer_command with actual transfer logic
- Update AttendantCSV struct with all new fields
