add project level management

Author: TimeLovercc

README.md

@@ -11,13 +11,13 @@ A 24/7 AI assistant daemon that continuously works on tasks via Slack. Uses Clau
 
 ## Full Precedure
 
-  1. Command arrives via CLI (sle task "...") or Slack (/task ...)
+  1. Command arrives via CLI (sle think -p project "..." or sle think "...") or Slack (/think --project=project ... or /think ...)
   2. TaskQueue stores it in SQLite database with status PENDING
   3. Daemon polls every 5 seconds, calling _process_tasks()
   4. SmartScheduler decides if task can execute:
     - Checks budget (time-based quotas: 90% night, 10% day)
     - Checks available parallel slots
-    - Prioritizes SERIOUS tasks over RANDOM thoughts
+    - Prioritizes SERIOUS tasks (with -p/--project) over THOUGHT tasks (without -p)
     - Returns list of tasks to execute
   5. Daemon executes via _execute_task():
     - Marks task as IN_PROGRESS

src/sleepless_agent/config.yaml

@@ -3,13 +3,13 @@ claude_code:
   model: claude-sonnet-4-5-20250929
   night_start_hour: 1
   night_end_hour: 9
-  threshold_day: 20.0
-  threshold_night: 80.0
+  threshold_day: 95.0
+  threshold_night: 96.0
   usage_command: claude /usage
 
 git:
   use_remote_repo: true
-  remote_repo_url: [email protected]:TimeLovercc/test1.git
+  remote_repo_url: [email protected]:TimeLovercc/test4.git # !!!CHANGE THIS!!!
   auto_create_repo: true
 
 agent:
@@ -38,9 +38,18 @@ auto_generation:
         - Active tasks: {task_count} ({pending_count} pending, {in_progress_count} in progress)
         - Many tasks are already in progress or pending
 
+        ## Available Tasks to Refine
+        {available_tasks}
+
         ## Recent Work & Context
         {recent_work}
 
+        ## Workspace Constraints
+        - Tasks execute in isolated directories: workspace/tasks/<task_id>/
+        - Each task only has access to its own workspace folder and workspace/shared/
+        - Do NOT reference workspace/data/ or other system directories in task descriptions
+        - REFINE tasks reuse existing task workspaces (no new folder created)
+
         ## Task Generation
         Generate ONE REFINE task to continue or improve existing work:
         - Complete partial/incomplete tasks mentioned above
@@ -50,9 +59,13 @@ auto_generation:
         - Add missing components to existing implementations
         - Expand documentation or analysis from previous tasks
 
-        IMPORTANT: Your response MUST start with [REFINE] followed by a specific, actionable task description in 1-2 sentences.
+        IMPORTANT Format Requirements:
+        - To refine a specific task: Use [REFINE:#<task_id>] followed by description
+          Example: [REFINE:#2] Improve error handling and add edge case tests
+        - For general refinements: Use [REFINE] followed by description
+        - Your response MUST be 1-2 sentences max
 
-        Focus on completing or improving what already exists in the workspace rather than starting new projects.
+        Focus on completing or improving what already exists rather than starting new projects.
       weight: 0.45
     - name: balanced
       prompt: |-
@@ -61,9 +74,18 @@ auto_generation:
         ## Current State
         - Active tasks: {task_count} ({pending_count} pending, {in_progress_count} in progress)
 
+        ## Available Tasks to Refine
+        {available_tasks}
+
         ## Recent Work & Context
         {recent_work}
 
+        ## Workspace Constraints
+        - Tasks execute in isolated directories: workspace/tasks/<task_id>/
+        - Each task only has access to its own workspace folder and workspace/shared/
+        - Do NOT reference workspace/data/ or other system directories in task descriptions
+        - REFINE tasks reuse existing task workspaces (no new folder created)
+
         ## Task Generation
         Generate ONE valuable task (NEW or REFINE):
         - For REFINE: improve existing work, complete partial tasks, enhance current projects
@@ -79,7 +101,12 @@ auto_generation:
         - Automation and productivity improvements
         - Analysis and evaluation reports
 
-        IMPORTANT: Prefix your response with [NEW] or [REFINE] followed by a specific, actionable task description in 1-2 sentences.
+        IMPORTANT Format Requirements:
+        - To refine a specific task: Use [REFINE:#<task_id>] followed by description
+          Example: [REFINE:#1] Add comprehensive testing suite with edge cases
+        - For general refinements: Use [REFINE] followed by description
+        - For new tasks: Use [NEW] followed by description
+        - Your response MUST be 1-2 sentences max
       weight: 0.35
     - name: new_friendly
       prompt: |-
@@ -89,6 +116,15 @@ auto_generation:
         - Active tasks: {task_count} ({pending_count} pending, {in_progress_count} in progress)
         - Few tasks in queue - good time for new projects!
 
+        ## Available Tasks to Refine
+        {available_tasks}
+
+        ## Workspace Constraints
+        - Tasks execute in isolated directories: workspace/tasks/<task_id>/
+        - Each task only has access to its own workspace folder and workspace/shared/
+        - Do NOT reference workspace/data/ or other system directories in task descriptions
+        - REFINE tasks reuse existing task workspaces (no new folder created)
+
         ## Task Generation
         Generate ONE innovative task that creates value.
 
@@ -106,6 +142,11 @@ auto_generation:
 
         Can be NEW (fresh project) or REFINE (improve existing work) - choose what would be most valuable.
 
-        IMPORTANT: Prefix your response with [NEW] or [REFINE] followed by a specific, actionable task description in 1-2 sentences.
+        IMPORTANT Format Requirements:
+        - To refine a specific task: Use [REFINE:#<task_id>] followed by description
+          Example: [REFINE:#3] Expand analysis with performance benchmarks
+        - For general refinements: Use [REFINE] followed by description
+        - For new tasks: Use [NEW] followed by description
+        - Your response MUST be 1-2 sentences max
       weight: 0.20
 

src/sleepless_agent/core/daemon.py

@@ -206,7 +206,7 @@ def _signal_handler(self, sig, _frame) -> None:
         self.running = False
         if self.live_status_tracker:
             self.live_status_tracker.clear_all()
-        self.bot.stop()
+        # Don't call bot.stop() here - let the finally block handle cleanup
         sys.exit(0)
 
     async def run(self) -> None: