feat: orchestrator enhancements, agent model refactor, dashboard UX improvements (#10)

Author: partarstu

.agent/skills/prepare-pr/SKILL.md

@@ -3,6 +3,8 @@ name: adding-orchestrator-workflow
 description: Adds new workflow endpoints to the QuAIA orchestrator. Use when creating new FastAPI endpoints that coordinate agent tasks, handle webhooks, or expose new API functionality.
 ---
 
+// turbo-all
+
 # Adding a New Orchestrator Workflow
 
 This skill provides a comprehensive guide for adding new workflow endpoints to the QuAIA™ orchestrator. Workflows are FastAPI endpoints that trigger and coordinate agent tasks.
@@ -15,6 +17,14 @@ The orchestrator (`orchestrator/main.py`) exposes HTTP endpoints that:
 3. Coordinate multi-agent workflows
 4. Handle results and trigger follow-up actions
 
+## ⚡ Auto-Run Policy
+
+This skill is designed to run with minimal interruption.
+
+- **Always set `SafeToAutoRun: true`** for all `run_command` calls.
+- **Exceptions:** Only set `SafeToAutoRun: false` if a step specifically instructs you to "Ask" the user, "Wait" for approval, or "Verify" a destructive action before proceeding.
+- **Restricted Symbols:** Never use the redirection operator (`>`) or `2>` in commands. Use alternatives (e.g., `Set-Content`, `Out-File`, or ignoring errors explicitly).
+
 ## Workflow Architecture
 
 A typical orchestrator workflow:

…ls/adding-orchestrator-workflow/SKILL.md …ls/adding-orchestrator-workflow/SKILL.md.agent/skills/adding-orchestrator-workflow/SKILL.md renamed to .agents/skills/adding-orchestrator-workflow/SKILL.md .agent/skills/adding-orchestrator-workflow/SKILL.md renamed to .agents/skills/adding-orchestrator-workflow/SKILL.md

@@ -3,6 +3,8 @@ name: creating-new-agent
 description: Creates new A2A-compliant agents in the QuAIA framework. Use when adding a new specialized agent with custom tools, prompts, and MCP server integrations.
 ---
 
+// turbo-all
+
 # Creating a New Agent
 
 This skill provides a comprehensive guide for creating a new specialized agent in the QuAIA™ framework. Agents are A2A-compliant (Agent-to-Agent protocol) services that handle specific QA-related tasks.
@@ -17,6 +19,14 @@ Each agent in QuAIA consists of:
 5. **Configuration** - Class in `config.py` for agent-specific settings
 6. **Unit tests** - Test file in `tests/agents/`
 
+## ⚡ Auto-Run Policy
+
+This skill is designed to run with minimal interruption.
+
+- **Always set `SafeToAutoRun: true`** for all `run_command` calls.
+- **Exceptions:** Only set `SafeToAutoRun: false` if a step specifically instructs you to "Ask" the user, "Wait" for approval, or "Verify" a destructive action before proceeding.
+- **Restricted Symbols:** Never use the redirection operator (`>`) or `2>` in commands. Use alternatives (e.g., `Set-Content`, `Out-File`, or ignoring errors explicitly).
+
 ## Step-by-Step Instructions
 
 ### Step 1: Create the Agent Directory Structure
@@ -45,7 +55,7 @@ Add a configuration class in `config.py` using the template:
 📄 **Template:** [resources/config_template.py](resources/config_template.py)
 
 **Configuration field descriptions:**
-- `THINKING_BUDGET`: Token budget for chain-of-thought reasoning (0 disables it)
+- `THINKING_LEVEL`: Thinking level for chain-of-thought reasoning ("MINIMAL" disables it or keeps it to minimum)
 - `OWN_NAME`: Human-readable name displayed in the orchestrator dashboard
 - `PORT`: Internal container port the agent listens on
 - `EXTERNAL_PORT`: Externally accessible port (usually same as PORT)
@@ -86,7 +96,7 @@ Create `agents/<agent_name>/main.py`:
 
 **Key points:**
 - The agent class MUST inherit from `AgentBase`
-- Implement `get_thinking_budget()` and `get_max_requests_per_task()`
+- Implement `get_thinking_level()` and `get_max_requests_per_task()`
 - Custom tools are defined as methods with full docstrings (LLM uses these)
 - The `app` variable exposes the A2A-compliant FastAPI application
 - `start_as_server()` runs the agent standalone with uvicorn

…r-workflow/examples/complete_workflow.py …r-workflow/examples/complete_workflow.py.agent/skills/adding-orchestrator-workflow/examples/complete_workflow.py renamed to .agents/skills/adding-orchestrator-workflow/examples/complete_workflow.py .agent/skills/adding-orchestrator-workflow/examples/complete_workflow.py renamed to .agents/skills/adding-orchestrator-workflow/examples/complete_workflow.py

@@ -24,7 +24,7 @@ def mock_config(monkeypatch):
     monkeypatch.setattr(config.<AgentName>AgentConfig, "EXTERNAL_PORT", 8099)
     monkeypatch.setattr(config.<AgentName>AgentConfig, "PROTOCOL", "http")
     monkeypatch.setattr(config.<AgentName>AgentConfig, "MODEL_NAME", "test")
-    monkeypatch.setattr(config.<AgentName>AgentConfig, "THINKING_BUDGET", 100)
+    monkeypatch.setattr(config.<AgentName>AgentConfig, "THINKING_LEVEL", "LOW")
     monkeypatch.setattr(config.<AgentName>AgentConfig, "MAX_REQUESTS_PER_TASK", 5)
     monkeypatch.setattr(config, "AGENT_BASE_URL", "http://localhost")
 
@@ -43,5 +43,5 @@ def test_agent_init(mock_super_init, mock_prompt_cls, mock_config):
     assert kwargs["agent_name"] == "Test Agent"
     assert kwargs["instructions"] == "system prompt"
 
-    assert agent.get_thinking_budget() == 100
+    assert agent.get_thinking_level() == "LOW"
     assert agent.get_max_requests_per_task() == 5