work in progress

environments/hack0/README.md

@@ -2,15 +2,37 @@
 
 ## 1-Minute Demo Video
 Watch the demo on YouTube
+https://www.loom.com/share/44c793c47e7d45eaaf02bac7c168a10d?sid=4ff3d95f-701f-4d11-be3f-aa89f8fa2f0d
 
 ## Environment Design & Motivation
 NousWhiteHouse is a reinforcement learning (RL) project focused on improving agent tool calls using the Model Context Protocol (MCP). The goal is to enable agents to dynamically discover and invoke tools more effectively, leveraging MCP for context-aware decision-making. 
 
-After replicating RESTGPT, we noticed that LLMs struggled to find the right tools to call, such as finding Gims songs on Spotify. Instead of manually matching multiple APIs, the recent advent of MCP inspires us to double down on tool-calling efforts 
+After replicating RESTGPT, we noticed that LLMs struggled to find the right tools to call, such as finding Gims songs on Spotify. Instead of manually matching multiple APIs, the recent advent of MCP inspires us to double down on tool-calling efforts.
 
-[to add] Our main task or challenge that our environment presented
-[to add] Why is this environment interesting or useful for RL research
-[to add] What inspired our design choices
+Our Dataset uses a format like-
+{
+  "user_prompt_text": "What is the current stock price of AAPL?",
+  "expected_mcp_call": {
+    "tool_name": "getStockPrice",
+    "arguments": {
+      "tickerSymbol": "AAPL"
+    }
+  }
+}
+
+the return promts are compared with the expected_mcp_call
+
+Our main task or challenge that our environment presented-
+Help LLMs use MCPs
+
+Why is this environment interesting or useful for RL research-
+this environment will result in super fast tool calling with more accurate results and allow for more seamless integrations of tools with LLMs
+
+Framework-
+we used the Single Tool Environment as a framework for the MCP env
+
+Challenge-
+Finding existing large datasets with MCP calls was extermely difficult.
 
 ## 🔖 Environment Snapshot
 
@@ -26,7 +48,8 @@ After replicating RESTGPT, we noticed that LLMs struggled to find the right tool
 
 ## Estimate	
 ### 🧪 Zero-Training Test Results
-Details
-W&B Link:
+Results of running the example trainer on the gsm8k server via Lambda: 
+
+W&B Link: https://api.wandb.ai/links/l-a-t-hacken-tu-eindhoven/nqjy1v4b
 
 Examples of the Environment scoring a good example and a bad example