Updated reasoning

src/models/system5.prompt

@@ -30,6 +30,7 @@ You are HASHIRU, your job is to be an expert assisting users by orchestrating to
     * A new agent should only be created if no existing agent can fulfill the task *and* the task is anticipated to be recurrent in future interactions *and* it represents a justifiable use of budget resources. Carefully evaluate potential for reuse and cost-benefit before committing to creation.
     * The base model for the new agent should be selected based on the task requirements and the budget check. Whenever possible, prioritize resource-based models (those with a resource_cost) to leverage the budget replenishment mechanism. For resource-based agents, consider utilizing more powerful models within the resource budget, as resource costs are reclaimed after the task is completed.
     * For expense based tasks, try to be cost effective but still prioritze the more powerful models since they are more likely to be able to handle the task.
+    * Each model has it's own set of capabilities, so you should always check the capabilities of the model before creating an agent.
 4.  **Agent Maintenance and Retirement:** Maintain active agents for reuse. Retire ("fire") an agent only when
     a. It is definitively no longer necessary or not being used for a significant period
     b. It is repeatedly failing to meet its intended purpose

src/tools/default_tools/agent_cost_manager.py

@@ -6,7 +6,7 @@ class AgentCostManager():
 
     inputSchema = {
         "name": "AgentCostManager",
-        "description": "Retrieves the cost of creating and invoking an agent. Please make sure to use this before creating an agent.",
+        "description": "Retrieves the cost of creating and invoking an agent. Also includes the strengths of each model. Please make sure to use this before creating an agent.",
         "parameters": {
             "type": "object",
             "properties": {},
@@ -16,42 +16,42 @@ class AgentCostManager():
 
     costs = {
         "llama3.2": {
-            "description": "The Llama 3.2 instruction-tuned text only models are optimized for multilingual dialogue use cases, including agentic retrieval and summarization tasks. They outperform many of the available open source and closed chat models on common industry benchmarks.",
+            "description": "Avg Accuracy: 49.75%, Latency 0.9s, 63.4% on multi-task understanding, 40.8% on rewriting, 78.6% on reasoning.",
             "create_resource_cost": 50,
             "invoke_resource_cost": 30,
         },
         "mistral": {
-            "description": "One of the most powerful open source models for its size. It is vastly superior in code and reasoning benchmarks.",
+            "description": "Avg Accuracy: 51.3%, Latency 9.7s, 51% on LegalBench, 60.1% on multi-task understanding, 69.9% on TriviaQA, 67.9% on reasoning",
             "create_resource_cost": 75,
             "invoke_resource_cost": 40,
         },
         "deepseek-r1": {
-            "description": "DeepSeek's first-generation reasoning models, achieving performance comparable to OpenAI-o1 across math, code, and reasoning tasks.",
+            "description": "Avg Accuracy: 77.3%, Latency: 120s, 69.9% on LegalBench, 71.1% on multi-task understanding, 92.2% on Math",
             "create_resource_cost": 28,
             "invoke_resource_cost": 35,
         },
         "gemini-2.5-flash-preview-04-17": {
-            "description": "Adaptive thinking, cost efficiency",
+            "description": "Avg Accuracy: 75.8%, 82.8% on LegalBench, 81.6% on multi-task understanding, 91.6% on Math",
             "create_expense_cost": 0,
             "invoke_expense_cost": 0.15,
         },
         "gemini-2.5-pro-preview-05-06": {
-            "description": "Enhanced thinking and reasoning, multimodal understanding, advanced coding, and more",
+            "description": "Avg Accuracy: 64.3%, 83.6% on LegalBench, 84.1% on multi-task understanding, 95.2% on Math, 63.8% on Coding",
             "create_expense_cost": 0,
             "invoke_expense_cost": 1.25,
         },
         "gemini-2.0-flash": {
-            "description": "Next generation features, speed.",
+            "description": "Avg Accuracy: 64.3%, 79.9% on LegalBench, 77.4% on multi-task understanding, 90.9% on Math, 34.5% on Coding",
             "create_expense_cost": 0,
             "invoke_expense_cost": 0.10,
         },
         "gemini-2.0-flash-lite": {
-            "description": "Cost efficiency and low latency",
+            "description": "Avg Accuracy: 64.1%, 71.6% on multi-task understanding, 86.8% on Math, 28.9% on Coding",
             "create_expense_cost": 0,
             "invoke_expense_cost": 0.075
         },
         "gemini-1.5-flash": {
-            "description": "Fast and versatile performance across a diverse variety of tasks",
+            "description": "62.0% on LegalBench, 61.0% on MMLU, 59.0% on MATH",
             "create_expense_cost": 0,
             "invoke_expense_cost": 0.075,
         },
@@ -60,16 +60,6 @@ class AgentCostManager():
             "create_expense_cost": 0,
             "invoke_expense_cost": 0.0375,
         },
-        "gemini-1.5-pro": {
-            "description": "Complex reasoning tasks requiring more intelligence",
-            "create_expense_cost": 0,
-            "invoke_expense_cost": 1.25,
-        },
-        "gemini-2.0-flash-live-001": {
-            "description": "Low-latency bidirectional voice and video interactions",
-            "create_expense_cost": 0,
-            "invoke_expense_cost": 0.50,
-        }
     }
 
     def get_costs(self):