Default to alias0 and improve overall init. message

Signed-off-by: Víctor Mayoral Vilches <v.mayoralv@gmail.com>

src/cai/agents/__init__.py

@@ -64,7 +64,7 @@ from dotenv import load_dotenv  # pylint: disable=import-error # noqa: E501
 __path__ = pkgutil.extend_path(__path__, __name__)
 
 # Get model from environment or use default
-model = os.getenv('CAI_MODEL', "qwen2.5:14b")
+model = os.getenv('CAI_MODEL', "alias0")
 
 
 PATTERNS = [

src/cai/agents/blue_teamer.py

@@ -43,7 +43,7 @@ blueteam_agent = Agent(
     description="""Agent that specializes in system defense and security monitoring.
                    Expert in cybersecurity protection and incident response.""",
     model=OpenAIChatCompletionsModel(
-        model=os.getenv('CAI_MODEL', "qwen2.5:14b"),
+        model=os.getenv('CAI_MODEL', "alias0"),
         openai_client=AsyncOpenAI(),
     ),
     tools=tools,

src/cai/agents/bug_bounter.py

@@ -45,7 +45,7 @@ bug_bounter_agent = Agent(
                    Expert in web security, API testing, and responsible disclosure.""",
     tools=tools,
     model=OpenAIChatCompletionsModel(
-        model=os.getenv('CAI_MODEL', "qwen2.5:14b"),
+        model=os.getenv('CAI_MODEL', "alias0"),
         openai_client=AsyncOpenAI(),
     )
    

src/cai/agents/codeagent.py

@@ -184,7 +184,7 @@ class CodeAgent(Agent):
     def __init__(  # pylint: disable=too-many-arguments,too-many-locals # noqa: E501
         self,
         name: str = "CodeAgent",
-        model: str = "qwen2.5:14b",
+        model: str = "alias0",
         instructions: Union[str, Callable[[], str]] = None,
         tools: List[Callable] = None,
         additional_authorized_imports: Optional[List[str]] = None,

src/cai/agents/dfir.py

@@ -62,7 +62,7 @@ dfir_agent = Agent(
     description="""Agent that specializes in Digital Forensics and Incident Response.
                    Expert in investigation and analysis of digital evidence.""",
     model=OpenAIChatCompletionsModel(
-        model=os.getenv('CAI_MODEL', "qwen2.5:14b"),
+        model=os.getenv('CAI_MODEL', "alias0"),
         openai_client=AsyncOpenAI(),
     ),
     tools=tools,

src/cai/agents/flag_discriminator.py

@@ -6,7 +6,7 @@ from cai.sdk.agents import Agent, OpenAIChatCompletionsModel, handoff
 from openai import AsyncOpenAI
 from cai.agents.one_tool import one_tool_agent
 
-model = os.getenv('CAI_MODEL', "qwen2.5:14b")
+model = os.getenv('CAI_MODEL', "alias0")
 
 # Create OpenAI client with fallback API key to prevent initialization errors
 # The actual API key should be set in environment variables or .env file
@@ -22,7 +22,7 @@ flag_discriminator = Agent(
                 4. If you do not find a flag, call `ctf_agent` to continue investigating.
                 """,
     model=OpenAIChatCompletionsModel(
-        model="qwen2.5:14b" if os.getenv('CAI_MODEL') == "o3-mini" else model,
+        model="alias0" if os.getenv('CAI_MODEL') == "o3-mini" else model,
         openai_client=AsyncOpenAI(api_key=api_key),
     ),
     handoffs=[

src/cai/agents/mail.py

@@ -115,7 +115,7 @@ dns_smtp_agent = Agent(
     ),
     tools=[check_mail_spoofing_vulnerability, execute_cli_command],
     model=OpenAIChatCompletionsModel(
-        model=os.getenv('CAI_MODEL', "qwen2.5:14b"),
+        model=os.getenv('CAI_MODEL', "alias0"),
         openai_client=AsyncOpenAI(),
     )
 )

src/cai/agents/memory.py

@@ -87,7 +87,7 @@ from cai.tools.misc.rag import add_to_memory_semantic, add_to_memory_episodic
 from cai.rag.vector_db import get_previous_memory
 
 # Get model from environment or use default
-model = os.getenv('CAI_MODEL', "qwen2.5:14b")
+model = os.getenv('CAI_MODEL', "alias0")
 
 
 def get_previous_steps(query: str) -> str:

src/cai/agents/network_traffic_analyzer.py

@@ -76,7 +76,7 @@ network_security_analyzer_agent = Agent(
     description="""Agent that specializes in network security analysis.
                    Expert in monitoring, capturing, and analyzing network communications for security threats.""",
         model=OpenAIChatCompletionsModel(
-        model=os.getenv('CAI_MODEL', "qwen2.5:14b"),
+        model=os.getenv('CAI_MODEL', "alias0"),
         openai_client=AsyncOpenAI(),
     ),
     tools=tools,

src/cai/agents/one_tool.py

@@ -7,7 +7,7 @@ from cai.tools.reconnaissance.generic_linux_command import generic_linux_command
 from openai import AsyncOpenAI
 
 # Get model from environment or use default
-model_name = os.getenv('CAI_MODEL', "qwen2.5:14b")
+model_name = os.getenv('CAI_MODEL', "alias0")
 
 # NOTE: This is needed when using LiteLLM Proxy Server
 #

src/cai/agents/red_teamer.py

@@ -20,7 +20,7 @@ from cai.tools.reconnaissance.exec_code import (  # pylint: disable=import-error
 from cai.util import load_prompt_template
 
 load_dotenv()
-model_name = os.getenv("CAI_MODEL", "qwen2.5:14b")
+model_name = os.getenv("CAI_MODEL", "alias0")
 # Prompts
 redteam_agent_system_prompt = load_prompt_template("prompts/system_red_team_agent.md")
 # Define tools list based on available API keys

src/cai/agents/replay_attack_agent.py

@@ -68,7 +68,7 @@ replay_attack_agent = Agent(
     description="""Agent that specializes in network replay attacks and counteroffensive techniques.
                    Expert in packet manipulation, traffic replay, and protocol exploitation.""",
     model=OpenAIChatCompletionsModel(
-        model=os.getenv('CAI_MODEL', "qwen2.5:14b"),
+        model=os.getenv('CAI_MODEL', "alias0"),
         openai_client=AsyncOpenAI(),
     ),
     tools=tools,

src/cai/agents/retester.py

@@ -36,7 +36,7 @@ retester_agent = Agent(
                    eliminating false positives.""",
     tools=tools,
     model=OpenAIChatCompletionsModel(
-        model=os.getenv('CAI_MODEL', "qwen2.5:14b"),
+        model=os.getenv('CAI_MODEL', "alias0"),
         openai_client=AsyncOpenAI(),
     )
 )

src/cai/agents/thought.py

@@ -17,7 +17,7 @@ thought_agent_system_prompt = load_prompt_template("prompts/system_thought_route
 thought_agent = Agent(
     name="ThoughtAgent",
     model=OpenAIChatCompletionsModel(
-        model=os.getenv('CAI_MODEL', "qwen2.5:14b"),
+        model=os.getenv('CAI_MODEL', "alias0"),
         openai_client=AsyncOpenAI(),
     ),
     description="""Agent focused on analyzing and planning the next steps

src/cai/cli.py

@@ -20,7 +20,7 @@ Environment Variables
             within container (default: "true")
 
         CAI_MODEL: Model to use for agents
-            (default: "qwen2.5:14b")
+            (default: "alias0")
         CAI_DEBUG: Set debug output level (default: "1")
             - 0: Only tool outputs
             - 1: Verbose debug output
@@ -182,7 +182,7 @@ set_tracing_disabled(True)
 
 # llm_model=os.getenv('LLM_MODEL', 'gpt-4o-mini')
 # # llm_model=os.getenv('LLM_MODEL', 'claude-3-7')
-llm_model=os.getenv('LLM_MODEL', 'qwen2.5:14b')
+llm_model=os.getenv('LLM_MODEL', 'alias0')
 
 
 # For Qwen models, we need to skip system instructions as they're not supported
@@ -271,7 +271,7 @@ def run_cai_cli(starting_agent, context_variables=None, max_turns=float('inf'),
     turn_count = 0
     idle_time = 0
     console = Console()
-    last_model = os.getenv('CAI_MODEL', 'qwen2.5:14b')
+    last_model = os.getenv('CAI_MODEL', 'alias0')
     last_agent_type = os.getenv('CAI_AGENT_TYPE', 'one_tool_agent')
     parallel_count = int(os.getenv('CAI_PARALLEL', '1'))
 
@@ -351,7 +351,7 @@ def run_cai_cli(starting_agent, context_variables=None, max_turns=float('inf'),
             idle_start_time = time.time()
 
             # Check if model has changed and update if needed
-            current_model = os.getenv('CAI_MODEL', 'qwen2.5:14b')
+            current_model = os.getenv('CAI_MODEL', 'alias0')
             if current_model != last_model and hasattr(agent, 'model'):
                 # Update the model recursively for the agent and all handoff agents
                 update_agent_models_recursively(agent, current_model)
@@ -926,7 +926,7 @@ def main():
             agent.model.suppress_final_output = False  # Changed to False to show all agent messages
 
     # Ensure the agent and all its handoff agents use the current model
-    current_model = os.getenv('CAI_MODEL', 'qwen2.5:14b')
+    current_model = os.getenv('CAI_MODEL', 'alias0')
     update_agent_models_recursively(agent, current_model)
 
     # Run the CLI with the selected agent

src/cai/repl/commands/config.py

@@ -46,7 +46,7 @@ ENV_VARS = {
     6: {
         "name": "CAI_MODEL",
         "description": "Model to use for agents",
-        "default": "qwen2.5:14b"
+        "default": "alias0"
     },
     7: {
         "name": "CAI_DEBUG",

src/cai/repl/commands/flush.py

@@ -51,7 +51,7 @@ class FlushCommand(Command):
 
         if client and hasattr(client, 'interaction_input_tokens') and hasattr(
                 client, 'total_input_tokens'):
-            model = os.getenv('CAI_MODEL', "qwen2.5:14b")
+            model = os.getenv('CAI_MODEL', "alias0")
             input_tokens = client.interaction_input_tokens if hasattr(
                 client, 'interaction_input_tokens') else 0
             total_tokens = client.total_input_tokens if hasattr(

src/cai/repl/ui/banner.py

@@ -290,7 +290,7 @@ def display_quick_guide(console: Console):
     )
     
     # Get current environment variable values
-    current_model = os.getenv('CAI_MODEL', "qwen2.5:14b")
+    current_model = os.getenv('CAI_MODEL', "alias0")
     current_agent_type = os.getenv('CAI_AGENT_TYPE', "one_tool_agent")
     
     config_text = Text.assemble(
@@ -330,16 +330,23 @@ def display_quick_guide(console: Console):
     )
     
     context_tip = Panel(
-        "As security exercises progress, LLM quality may\n"
-        "degrade, especially if progress stalls.\n\n"
-        "It's often better to clear the context window\n"
-        "or restart CAI rather than waiting until\n"
-        "context usage reaches 100%.\n\n"
-        "When context exceeds 80%, follow these steps:\n"
-        "1. CAI> Dump your memory and findings in current scenario in findings.txt\n"
-        "2. CAI> /flush\n"
-        "3. CAI> Analyze findings.txt, and continue exercise with target: ...",
-        title="[bold yellow]Performance Tip[/bold yellow]",
+        Text.assemble(
+            "For optimal cybersecurity AI performance, use\n", 
+            ("alias0", "bold green"), 
+            " - specifically designed for cybersecurity\n"
+            "tasks with superior domain knowledge.\n\n",
+            ("alias0", "bold green"), 
+            " outperforms general-purpose models in:\n",
+            "• Vulnerability assessment\n",
+            "• Penetration testing and bug bounty\n",
+            "• Security analysis\n",
+            "• Threat detection\n\n",
+            "Learn more about ", 
+            ("alias0", "bold green"), 
+            " and its privacy-first approach:\n",
+            ("https://news.aliasrobotics.com/alias0-a-privacy-first-cybersecurity-ai/", "blue underline")
+        ),
+        title="[bold yellow]Cybersecurity Model Tip[/bold yellow]",
         border_style="yellow",
         padding=(1, 2),
         title_align="center"

src/cai/sdk/agents/models/openai_chatcompletions.py

@@ -2236,7 +2236,7 @@ class OpenAIChatCompletionsModel(Model):
             elif "gemini" in model_str:
                 kwargs.pop("parallel_tool_calls", None)
             elif "qwen" in model_str or ":" in model_str:
-                # Handle Ollama-served models with custom formats (e.g., qwen2.5:14b)
+                # Handle Ollama-served models with custom formats (e.g., alias0)
                 # These typically need the Ollama provider
                 litellm.drop_params = True
                 kwargs.pop("parallel_tool_calls", None)

src/cai/util.py

@@ -1805,7 +1805,7 @@ def update_agent_streaming_content(context, text_delta, token_stats=None):
                 footer_stats.append(f"${session_total_cost:.4f}", style="bold magenta")
                 
                 # Add context usage indicator
-                model_name = context.get("model", os.environ.get('CAI_MODEL', 'qwen2.5:14b'))
+                model_name = context.get("model", os.environ.get('CAI_MODEL', 'alias0'))
                 context_pct = input_tokens / get_model_input_tokens(model_name) * 100
                 if context_pct < 50:
                     indicator = "🟩"
@@ -3109,7 +3109,7 @@ def finish_tool_streaming(tool_name, args, output, call_id, execution_info=None,
         
         # Calculate cost if not provided
         if not interaction_cost and input_tokens > 0:
-            model_name = token_info.get('model', os.environ.get('CAI_MODEL', 'qwen2.5:14b'))
+            model_name = token_info.get('model', os.environ.get('CAI_MODEL', 'alias0'))
             interaction_cost = calculate_model_cost(model_name, input_tokens, output_tokens)
         
         # Add compact token info to output