Renamed PostgreSQL references to "Tables" for clarity in bootstrap logs, changed config sync logging from info to trace for reduced verbosity, and made session message clearing method private to limit external access.
Added a check in `BootstrapManager` to detect if PostgreSQL is running and attempt to start the "tables" component automatically if not. Also prefixed unused variables and struct fields with underscores in compiler, session, and model modules to suppress warnings and improve code clarity.
Added `once_cell` and `scopeguard` dependencies to implement thread-safe compaction lock mechanism. Modified `compact_prompt_for_bot` to:
- Prevent concurrent compaction for the same bot using a global lock
- Add proper tracing and error handling
- Improve summarization with content filtering
- Clean up locks automatically using scopeguard
- Remove redundant threshold check and compact entire history
The changes ensure thread safety during prompt compaction and provide better observability through tracing.
- Added initial 30s delay to compact prompt scheduler
- Implemented async LLM summarization for conversation history
- Reduced lock contention by minimizing critical sections
- Added fallback to original text if summarization fails
- Updated README with guidance for failed requirements
- Added new `summarize` method to LLMProvider trait
- Improved session manager query with proper DSL usage
The changes optimize the prompt compaction process by:
1. Reducing lock contention through better resource management
2. Adding LLM-based summarization for better conversation compression
3. Making the system more resilient with proper error handling
4. Improving documentation for development practices
- Simplified auth module by removing unused imports and code
- Cleaned up shared models by removing unused structs (Organization, User, Bot, etc.)
- Updated add-req.sh to comment out unused directories
- Modified LLM fallback strategy in README with additional notes about model behaviors
The changes focus on removing unused code and improving documentation while maintaining existing functionality. The auth module was significantly reduced by removing redundant code, and similar cleanup was applied to shared models. The build script was adjusted to reflect currently used directories.
Enable automation, basic, config, context, session, kb, and tools directories while disabling bootstrap, file, and drive_monitor to refine which modules are processed by add-req.sh.
Added interaction count tracking for sessions with Redis or in-memory fallback. Implemented conversation history replacement functionality to compact and update message history. The changes include:
- New AtomicUsize counter in SessionManager for interaction tracking
- increment_and_get_interaction_count method with Redis support
- replace_conversation_history to update and compact message history
- Maintains existing functionality while adding new features
- Deduplicate consecutive messages with same role in conversation history
- Add n_predict configuration option for LLM server
- Prevent duplicate message storage in session manager
- Update announcement schedule timing from 37 to 55 minutes
- Add default n_predict value in default bot config
- Added `trace!` logging in `bot_memory.rs` to record retrieved memory values for easier debugging.
- Refactored `BotOrchestrator` in `bot/mod.rs`:
- Removed duplicate session save block and consolidated message persistence.
- Replaced low‑level LLM streaming with a structured `UserMessage` and `stream_response` workflow, improving error handling and readability.
- Updated configuration loading in `config/mod.rs`:
- Imported `get_default_bot` and enhanced `get_config` to fall back to the default bot configuration when the primary query fails.
- Established a fresh DB connection for the fallback path to avoid borrowing issues.
- Remove `scripts_dir` from `AutomationService` and its constructor, as it was unused.
- Drop LLM and embedding server readiness checks; the service now only schedules periodic tasks.
- Increase the health‑check interval from 5 seconds to 15 seconds for reduced load.
- Streamline the LLM keyword prompt to a concise `"User: {}"` format, removing verbose boilerplate.
- Remove unnecessary logging and LLM cache handling code from the bot orchestrator, cleaning up unused environment variable checks and cache queries.
- Added LLM server readiness check in AutomationService before starting tasks
- Renamed `user` parameter to `user_session` in execute_talk for clarity
- Updated BotResponse fields to use user_session data instead of hardcoded values
- Improved Redis key generation in execute_talk to use user_session fields
- Removed commented Redis code in set_current_context_keyword
The changes ensure proper initialization of automation tasks by checking LLM server availability first, and improve code clarity by using more descriptive variable names for user session data.
Changed async Redis operations to synchronous in add_suggestion_keyword function. Removed unnecessary async/await and tokio::spawn since the operations are now blocking. This simplifies the code while maintaining the same functionality of storing suggestions and context state in Redis. Error handling remains robust with proper early returns.
- Remove comments from launch.json configuration
- Simplify Redis key format in set_current_context_keyword by removing context_name
- Remove unused /api/start endpoint and related session start logic
- Change log level from info to trace for config sync messages
- Add trace logging to drive_monitor module
The changes focus on code cleanup and simplification, particularly around session handling and logging. The Redis key format was simplified as the context_name was deemed unnecessary for uniqueness.
- Added new ADD_SUGGESTION keyword handler to support sending suggestions in responses
- Removed unused env import in hear_talk module
- Simplified bot_id assignment to use static string
- Added suggestions field to BotResponse struct
- Improved SET_CONTEXT keyword to take both name and value parameters
- Fixed whitespace in auth handler
- Enhanced error handling for suggestion sending
The changes improve the suggestion system functionality while cleaning up unused code and standardizing response handling.
Introduce bot mounting logic in `BotOrchestrator` to load and manage active bots from the database. Enhance `BootstrapManager` by refining Diesel query usage and connection handling for better reliability and maintainability.
- Extract LLM generation into `execute_llm_generation` and simplify
keyword handling.
- Prepend system prompt and session context to LLM prompts in
`BotOrchestrator`.
- Parse incoming WebSocket messages as JSON and use the `content` field.
- Add async `get_session_context` and stop injecting Redis context into
conversation history.
- Change default LLM URL to `http://48.217.66.81:8080` throughout the
project.
- Use the existing DB pool instead of creating a separate custom
connection.
- Update `start.bas` to call LLM and set a new context string.
- Refactor web client message handling: separate event processing,
improve streaming logic, reset streaming state on thinking end, and
remove unused test functions.
- Use `tokio::spawn` to run Redis SET for `SET_CONTEXT` in a background
task with detailed tracing.
- Move context retrieval from `BotOrchestrator` to `SessionManager`,
inserting it as a system message in conversation history.
- Remove redundant Redis fetch logic from `BotOrchestrator`.
- Update `DEV.md` to install `valkey-cli`, reorder cargo tools, and
adjust apt commands.
- Add a `SET_CONTEXT "azul bolinha"` example to the announcements
template.
