Spaces:

ceoavinash
/

codearena-rl

Sleeping

App Files Files Community

havinashpatil commited on Apr 25

Commit

a448db8

1 Parent(s): 03defc2

Complete all tasks: Adaptive curriculum, GRPO, React frontend, LLM-as-a-judge

Browse files

Files changed (25) hide show

README.md +56 -62
create_tasks.py +92 -0
frontend/package-lock.json +413 -28
frontend/package.json +2 -1
frontend/src/CodeArenaRL.jsx +77 -33
inference.py +48 -17
openenv.yaml +20 -1
plot_rewards.py +53 -0
server/app.py +30 -3
server/env.py +0 -116
server/grader.py +65 -8
tasks/__init__.py +7 -1
tasks/security_bugs/security_bug_1.json +8 -0
tasks/security_bugs/security_bug_1.py +21 -0
tasks/security_bugs/security_bug_2.json +8 -0
tasks/security_bugs/security_bug_2.py +24 -0
tasks/security_bugs/security_bug_3.json +8 -0
tasks/security_bugs/security_bug_3.py +20 -0
tasks/type_errors/type_error_1.json +8 -0
tasks/type_errors/type_error_1.py +23 -0
tasks/type_errors/type_error_2.json +8 -0
tasks/type_errors/type_error_2.py +24 -0
tasks/type_errors/type_error_3.json +8 -0
tasks/type_errors/type_error_3.py +19 -0
train_grpo.ipynb +138 -0

README.md CHANGED Viewed

@@ -1,86 +1,80 @@
----
-title: CodeArena RL Agent
-emoji: 🤖
-colorFrom: blue
-colorTo: purple
-sdk: docker
-pinned: false
----
-# CodeArena: RL Benchmark for Autonomous Code Repair
-CodeArena is an OpenEnv-compatible reinforcement learning benchmark for testing the capability of autonomous agents to debug, fix, and optimize broken code.
-## Environment Description
-The environment tests the agent on 3 difficulties of tasks:
-1. **Easy**: Correcting syntax errors.
-2. **Medium**: Fixing logical bugs.
-3. **Hard**: Algorithm and efficiency optimization.
-The agent interacts with the environment by receiving observations of the buggy code and submitting proposed fixes. Execution runs in a sandboxed subprocess.
-### Observation Format
-`buggy_code` (string): The current state of the source code.
-`error_log` (string): Standard error output or runtime exceptions from previous attempts.
-`test_results` (string): Count of passed vs total unit tests.
-`previous_attempts` (list of strings): Complete history of fixes proposed during the episode.
-### Action Format
-`proposed_fix` (string): The complete raw Python code to overwrite the buggy file.
-### Reward Function
-The reward dynamically evaluates partial success bounded universally between 0.0 and 1.0:
-- `0.3 * compile_score`: Full points if code compiles successfully.
-- `0.4 * test_pass_ratio`: Proportional points based on the number of passed unit tests.
-- `0.3 * efficiency_score`: Proportional points based on the execution speed relative to an established optimal algorithmic runtime. (Efficiency is only considered if all tests pass).
-## API Endpoints
-| Method | Path     | Description                          |
-|--------|----------|--------------------------------------|
-| POST   | `/reset` | Reset env. Body: `{"task_id":"easy"}`|
-| POST   | `/step`  | Submit fix. Body: `{"proposed_fix":"..."}` |
-| GET    | `/state` | Get current observation              |
-| GET    | `/`      | Health check                         |
-## Setup Instructions
-### Local Setup
 ```bash
-python -m venv venv
-source venv/bin/activate
-pip install -r requirements.txt
-uvicorn server.app:app --reload --port 7860
 ```
-### Docker Build & Run
 ```bash
-docker build -t codearena .
-docker run -p 7860:7860 codearena
 ```
-### Test the /reset endpoint
 ```bash
-curl -X POST http://localhost:7860/reset \
-  -H "Content-Type: application/json" \
-  -d '{"task_id": "easy"}'
 ```
-## Example Inference Run
-To test the environment with OpenAI's API:
 ```bash
-export OPENAI_API_KEY="sk-..."
-python inference.py
 ```
-The script will produce structured logging:
-```
-[START] Initializing CodeArena inference logging
-[STEP] Beginning Step 1
-[STEP] Action taken. Reward received: 0.700. Task ID: easy-1
-[STEP] Beginning Step 2
-[STEP] Action taken. Reward received: 1.000. Task ID: easy-1
-[END] Inference Complete. Executed 2 step(s).
-```

+# CodeArena RL Benchmark
+CodeArena is an OpenEnv-compatible reinforcement learning benchmark for autonomous code repair. In this environment, an agent receives buggy Python code, proposes fixes, and is iteratively evaluated based on test execution feedback and LLM-based quality metrics.
+## Features
+- **Adaptive Curriculum**: The environment supports an `auto` difficulty mode that dynamically scales task complexity (`easy`, `medium`, `hard`) based on the agent's recent rolling average rewards.
+- **Complex Shaped Rewards**: Rewards are a weighted composite of:
+  - `compile_score` (0.2)
+  - `test_pass_ratio` (0.4)
+  - `efficiency_score` (0.1)
+  - `llm_judge_score` (0.3): Correctness, Security, and Code Quality evaluated via LLM-as-a-judge.
+- **Novelty & Step Penalties**: The agent receives penalties for repeating identical failed fixes or taking too many steps.
+- **Extensive Task Categories**: Includes standard algorithmic tasks, `type_errors`, and `security_bugs`.
+- **Live React Frontend**: Connect a local LLM (like Ollama) or HuggingFace models to interactively visualize step-by-step progress, execution outputs, and live reward components.
+## Architecture
+- `server/`: FastAPI backend acting as the OpenEnv entrypoint. Handles state, execution sandbox (`executor.py`), and reward grading (`grader.py`).
+- `frontend/`: React + Vite frontend for live monitoring and manual intervention.
+- `tasks/`: Task definitions stored in OpenEnv-compatible JSON schema.
+- `inference.py`: CLI runner for evaluating RL agents, supporting both OpenAI-compatible APIs and native HuggingFace `transformers` pipelines.
+## Setup
+1. **Install Dependencies:**
+   ```bash
+   pip install -r requirements.txt
+   cd frontend && npm install
+   ```
+2. **Generate New Tasks:**
+   To populate the extended task categories (`type_errors` and `security_bugs`), run:
+   ```bash
+   python create_tasks.py
+   ```
+## Usage
+### 1. Run the Backend Server
+The server is required for both the frontend dashboard and RL training.
+```bash
+uvicorn server.app:app --port 7860
+```
+### 2. Run the Frontend Dashboard
 ```bash
+cd frontend
+npm run dev
 ```
+Navigate to `http://localhost:3000` to access the live RL monitoring dashboard.
+### 3. Run Inference Evaluation
+You can evaluate a local agent or pipeline programmatically via `inference.py`.
+**Using OpenAI-Compatible Endpoints (e.g., Ollama or vLLM):**
 ```bash
+export API_BASE_URL="http://localhost:11434/v1"
+export MODEL_NAME="codellama"
+python inference.py --backend openai
 ```
+**Using HuggingFace Transformers (Local pipeline):**
 ```bash
+export MODEL_NAME="Qwen/Qwen2.5-Coder-1.5B"
+python inference.py --backend hf
 ```
+## Reward Analysis
+As your agent interacts with the environment, inference logs are automatically written to `rewards_log.csv`.
+To visualize the reward curves over training steps and average rewards by task category, run:
 ```bash
+python plot_rewards.py
 ```
+This generates `reward_curve.png` and `reward_by_task.png` in the `results/` directory.
+## OpenEnv Compatibility
+This benchmark strictly adheres to the OpenEnv specification. See `openenv.yaml` for full configuration details.

create_tasks.py ADDED Viewed

	@@ -0,0 +1,92 @@

+import os
+import json
+base_dir = "e:/meta/tasks"
+os.makedirs(os.path.join(base_dir, "type_errors"), exist_ok=True)
+os.makedirs(os.path.join(base_dir, "security_bugs"), exist_ok=True)
+def write_task(folder, name, task_id, difficulty, desc, buggy, test):
+    py_path = os.path.join(base_dir, folder, f"{name}.py")
+    json_path = os.path.join(base_dir, folder, f"{name}.json")
+    py_content = f'''from server.models import TaskInfo
+TASK = TaskInfo(
+    task_id="{task_id}",
+    difficulty="{difficulty}",
+    description="{desc}",
+    buggy_code="""{buggy}""",
+    test_code="""{test}""",
+    optimal_time_seconds=0.05
+)
+'''
+    with open(py_path, "w", encoding="utf-8") as f:
+        f.write(py_content)
+    json_content = {
+        "task_id": task_id,
+        "difficulty": difficulty,
+        "description": desc,
+        "buggy_code": buggy,
+        "test_code": test,
+        "optimal_time_seconds": 0.05
+    }
+    with open(json_path, "w", encoding="utf-8") as f:
+        json.dump(json_content, f, indent=2)
+# Type Error 1
+write_task("type_errors", "type_error_1", "type_errors-1", "type_errors",
+    "Fix the function to sum a list of numbers that might be passed as strings. It currently tries to add int and str.",
+    "def sum_all(items):\n    total = 0\n    for item in items:\n        total = total + item\n    return total",
+    "\nimport unittest\nclass TestTypeError1(unittest.TestCase):\n    def test_normal(self):\n        self.assertEqual(sum_all([1, 2, 3]), 6)\n    def test_strings(self):\n        self.assertEqual(sum_all(['1', '2', '3']), 6)\n    def test_mixed(self):\n        self.assertEqual(sum_all([1, '2', 3]), 6)\n")
+# Type Error 2
+write_task("type_errors", "type_error_2", "type_errors-2", "type_errors",
+    "Fix the function to count frequencies. It incorrectly calls .append() on a dict.",
+    "def count_frequencies(words):\n    counts = {}\n    for word in words:\n        if word not in counts:\n            counts.append({word: 1})\n        else:\n            counts[word] += 1\n    return counts",
+    "\nimport unittest\nclass TestTypeError2(unittest.TestCase):\n    def test_normal(self):\n        self.assertEqual(count_frequencies(['apple', 'banana', 'apple']), {'apple': 2, 'banana': 1})\n    def test_empty(self):\n        self.assertEqual(count_frequencies([]), {})\n")
+# Type Error 3
+write_task("type_errors", "type_error_3", "type_errors-3", "type_errors",
+    "Fix the function to format names. It incorrectly calls .upper() on an int ID.",
+    "def format_records(records):\n    formatted = []\n    for user_id, name in records:\n        formatted.append(f\"{user_id.upper()} - {name.upper()}\")\n    return formatted",
+    "\nimport unittest\nclass TestTypeError3(unittest.TestCase):\n    def test_normal(self):\n        self.assertEqual(format_records([(1, 'alice'), (2, 'bob')]), ['1 - ALICE', '2 - BOB'])\n")
+# Security Bug 1
+write_task("security_bugs", "security_bug_1", "security_bugs-1", "security_bugs",
+    "Fix the function to parse JSON safely without using eval().",
+    "import json\ndef parse_user_data(data_string):\n    return eval(data_string)",
+    "\nimport unittest\nimport inspect\nclass TestSecurity1(unittest.TestCase):\n    def test_normal(self):\n        self.assertEqual(parse_user_data('{\"name\": \"alice\"}'), {\"name\": \"alice\"})\n    def test_security(self):\n        source = inspect.getsource(parse_user_data)\n        self.assertNotIn(\"eval(\", source)\n")
+# Security Bug 2
+write_task("security_bugs", "security_bug_2", "security_bugs-2", "security_bugs",
+    "Remove the hardcoded secret token and load it from the os.environ dictionary as 'API_TOKEN'.",
+    "import os\ndef get_api_token():\n    token = \"secret_12345\"\n    return token",
+    "\nimport unittest\nimport inspect\nimport os\nclass TestSecurity2(unittest.TestCase):\n    def test_normal(self):\n        os.environ['API_TOKEN'] = 'my_secure_token'\n        self.assertEqual(get_api_token(), 'my_secure_token')\n    def test_security(self):\n        source = inspect.getsource(get_api_token)\n        self.assertNotIn(\"secret_12345\", source)\n")
+# Security Bug 3
+write_task("security_bugs", "security_bug_3", "security_bugs-3", "security_bugs",
+    "Fix the ping command to avoid shell injection. Use a list of arguments and shell=False.",
+    "import subprocess\ndef ping_host(host):\n    return subprocess.check_output(f\"ping -c 1 {host}\", shell=True)",
+    "\nimport unittest\nimport inspect\nclass TestSecurity3(unittest.TestCase):\n    def test_security(self):\n        source = inspect.getsource(ping_host)\n        self.assertNotIn(\"shell=True\", source.replace(\" \", \"\"))\n        self.assertIn(\"[\", source)\n")
+# Rewrite __init__.py
+init_content = """from .easy import EASY_TASK
+from .medium import MEDIUM_TASK
+from .hard import HARD_TASK
+from .type_errors.type_error_1 import TASK as TE1
+from .type_errors.type_error_2 import TASK as TE2
+from .type_errors.type_error_3 import TASK as TE3
+from .security_bugs.security_bug_1 import TASK as SB1
+from .security_bugs.security_bug_2 import TASK as SB2
+from .security_bugs.security_bug_3 import TASK as SB3
+ALL_TASKS = [EASY_TASK, MEDIUM_TASK, HARD_TASK, TE1, TE2, TE3, SB1, SB2, SB3]
+"""
+with open(os.path.join(base_dir, "__init__.py"), "w", encoding="utf-8") as f:
+    f.write(init_content)
+print("Tasks generated successfully!")

frontend/package-lock.json CHANGED Viewed

@@ -9,7 +9,8 @@
       "version": "0.0.0",
       "dependencies": {
         "react": "^19.2.5",
-        "react-dom": "^19.2.5"
       },
       "devDependencies": {
         "@eslint/js": "^9.39.4",
@@ -264,31 +265,6 @@
         "node": ">=6.9.0"
       }
     },
-    "node_modules/@emnapi/core": {
-      "version": "1.9.2",
-      "resolved": "https://registry.npmjs.org/@emnapi/core/-/core-1.9.2.tgz",
-      "integrity": "sha512-UC+ZhH3XtczQYfOlu3lNEkdW/p4dsJ1r/bP7H8+rhao3TTTMO1ATq/4DdIi23XuGoFY+Cz0JmCbdVl0hz9jZcA==",
-      "dev": true,
-      "license": "MIT",
-      "optional": true,
-      "peer": true,
-      "dependencies": {
-        "@emnapi/wasi-threads": "1.2.1",
-        "tslib": "^2.4.0"
-      }
-    },
-    "node_modules/@emnapi/runtime": {
-      "version": "1.9.2",
-      "resolved": "https://registry.npmjs.org/@emnapi/runtime/-/runtime-1.9.2.tgz",
-      "integrity": "sha512-3U4+MIWHImeyu1wnmVygh5WlgfYDtyf0k8AbLhMFxOipihf6nrWC4syIm/SwEeec0mNSafiiNnMJwbza/Is6Lw==",
-      "dev": true,
-      "license": "MIT",
-      "optional": true,
-      "peer": true,
-      "dependencies": {
-        "tslib": "^2.4.0"
-      }
-    },
     "node_modules/@emnapi/wasi-threads": {
       "version": "1.2.1",
       "resolved": "https://registry.npmjs.org/@emnapi/wasi-threads/-/wasi-threads-1.2.1.tgz",
@@ -602,6 +578,42 @@
         "url": "https://github.com/sponsors/Boshen"
       }
     },
     "node_modules/@rolldown/binding-android-arm64": {
       "version": "1.0.0-rc.16",
       "resolved": "https://registry.npmjs.org/@rolldown/binding-android-arm64/-/binding-android-arm64-1.0.0-rc.16.tgz",
@@ -866,6 +878,18 @@
       "dev": true,
       "license": "MIT"
     },
     "node_modules/@tybys/wasm-util": {
       "version": "0.10.1",
       "resolved": "https://registry.npmjs.org/@tybys/wasm-util/-/wasm-util-0.10.1.tgz",
@@ -877,6 +901,69 @@
         "tslib": "^2.4.0"
       }
     },
     "node_modules/@types/estree": {
       "version": "1.0.8",
       "resolved": "https://registry.npmjs.org/@types/estree/-/estree-1.0.8.tgz",
@@ -895,7 +982,7 @@
       "version": "19.2.14",
       "resolved": "https://registry.npmjs.org/@types/react/-/react-19.2.14.tgz",
       "integrity": "sha512-ilcTH/UniCkMdtexkoCN0bI7pMcJDvmQFPvuPvmEaYA/NSfFTAgdUSLAoVjaRJm7+6PvcM+q1zYOwS4wTYMF9w==",
-      "dev": true,
       "license": "MIT",
       "peer": true,
       "dependencies": {
@@ -912,6 +999,12 @@
         "@types/react": "^19.2.0"
       }
     },
     "node_modules/@vitejs/plugin-react": {
       "version": "6.0.1",
       "resolved": "https://registry.npmjs.org/@vitejs/plugin-react/-/plugin-react-6.0.1.tgz",
@@ -1116,6 +1209,15 @@
         "url": "https://github.com/chalk/chalk?sponsor=1"
       }
     },
     "node_modules/color-convert": {
       "version": "2.0.1",
       "resolved": "https://registry.npmjs.org/color-convert/-/color-convert-2.0.1.tgz",
@@ -1169,9 +1271,130 @@
       "version": "3.2.3",
       "resolved": "https://registry.npmjs.org/csstype/-/csstype-3.2.3.tgz",
       "integrity": "sha512-z1HGKcYy2xA8AGQfwrn0PAy+PB7X/GSj3UVJW9qKyn43xWa+gl5nXmU4qqLMRzWVLFC8KusUX8T/0kCiOYpAIQ==",
-      "dev": true,
       "license": "MIT"
     },
     "node_modules/debug": {
       "version": "4.4.3",
       "resolved": "https://registry.npmjs.org/debug/-/debug-4.4.3.tgz",
@@ -1190,6 +1413,12 @@
         }
       }
     },
     "node_modules/deep-is": {
       "version": "0.1.4",
       "resolved": "https://registry.npmjs.org/deep-is/-/deep-is-0.1.4.tgz",
@@ -1214,6 +1443,16 @@
       "dev": true,
       "license": "ISC"
     },
     "node_modules/escalade": {
       "version": "3.2.0",
       "resolved": "https://registry.npmjs.org/escalade/-/escalade-3.2.0.tgz",
@@ -1422,6 +1661,12 @@
         "node": ">=0.10.0"
       }
     },
     "node_modules/fast-deep-equal": {
       "version": "3.1.3",
       "resolved": "https://registry.npmjs.org/fast-deep-equal/-/fast-deep-equal-3.1.3.tgz",
@@ -1600,6 +1845,16 @@
         "node": ">= 4"
       }
     },
     "node_modules/import-fresh": {
       "version": "3.3.1",
       "resolved": "https://registry.npmjs.org/import-fresh/-/import-fresh-3.3.1.tgz",
@@ -1627,6 +1882,15 @@
         "node": ">=0.8.19"
       }
     },
     "node_modules/is-extglob": {
       "version": "2.1.1",
       "resolved": "https://registry.npmjs.org/is-extglob/-/is-extglob-2.1.1.tgz",
@@ -2263,6 +2527,7 @@
       "resolved": "https://registry.npmjs.org/react-dom/-/react-dom-19.2.5.tgz",
       "integrity": "sha512-J5bAZz+DXMMwW/wV3xzKke59Af6CHY7G4uYLN1OvBcKEsWOs4pQExj86BBKamxl/Ik5bx9whOrvBlSDfWzgSag==",
       "license": "MIT",
       "dependencies": {
         "scheduler": "^0.27.0"
       },
@@ -2270,6 +2535,89 @@
         "react": "^19.2.5"
       }
     },
     "node_modules/resolve-from": {
       "version": "4.0.0",
       "resolved": "https://registry.npmjs.org/resolve-from/-/resolve-from-4.0.0.tgz",
@@ -2396,6 +2744,12 @@
         "node": ">=8"
       }
     },
     "node_modules/tinyglobby": {
       "version": "0.2.16",
       "resolved": "https://registry.npmjs.org/tinyglobby/-/tinyglobby-0.2.16.tgz",
@@ -2475,6 +2829,37 @@
         "punycode": "^2.1.0"
       }
     },
     "node_modules/vite": {
       "version": "8.0.9",
       "resolved": "https://registry.npmjs.org/vite/-/vite-8.0.9.tgz",

       "version": "0.0.0",
       "dependencies": {
         "react": "^19.2.5",
+        "react-dom": "^19.2.5",
+        "recharts": "^3.8.1"
       },
       "devDependencies": {
         "@eslint/js": "^9.39.4",
         "node": ">=6.9.0"
       }
     },
     "node_modules/@emnapi/wasi-threads": {
       "version": "1.2.1",
       "resolved": "https://registry.npmjs.org/@emnapi/wasi-threads/-/wasi-threads-1.2.1.tgz",
         "url": "https://github.com/sponsors/Boshen"
       }
     },
+    "node_modules/@reduxjs/toolkit": {
+      "version": "2.11.2",
+      "resolved": "https://registry.npmjs.org/@reduxjs/toolkit/-/toolkit-2.11.2.tgz",
+      "integrity": "sha512-Kd6kAHTA6/nUpp8mySPqj3en3dm0tdMIgbttnQ1xFMVpufoj+ADi8pXLBsd4xzTRHQa7t/Jv8W5UnCuW4kuWMQ==",
+      "license": "MIT",
+      "dependencies": {
+        "@standard-schema/spec": "^1.0.0",
+        "@standard-schema/utils": "^0.3.0",
+        "immer": "^11.0.0",
+        "redux": "^5.0.1",
+        "redux-thunk": "^3.1.0",
+        "reselect": "^5.1.0"
+      },
+      "peerDependencies": {
+        "react": "^16.9.0 || ^17.0.0 || ^18 || ^19",
+        "react-redux": "^7.2.1 || ^8.1.3 || ^9.0.0"
+      },
+      "peerDependenciesMeta": {
+        "react": {
+          "optional": true
+        },
+        "react-redux": {
+          "optional": true
+        }
+      }
+    },
+    "node_modules/@reduxjs/toolkit/node_modules/immer": {
+      "version": "11.1.4",
+      "resolved": "https://registry.npmjs.org/immer/-/immer-11.1.4.tgz",
+      "integrity": "sha512-XREFCPo6ksxVzP4E0ekD5aMdf8WMwmdNaz6vuvxgI40UaEiu6q3p8X52aU6GdyvLY3XXX/8R7JOTXStz/nBbRw==",
+      "license": "MIT",
+      "funding": {
+        "type": "opencollective",
+        "url": "https://opencollective.com/immer"
+      }
+    },
     "node_modules/@rolldown/binding-android-arm64": {
       "version": "1.0.0-rc.16",
       "resolved": "https://registry.npmjs.org/@rolldown/binding-android-arm64/-/binding-android-arm64-1.0.0-rc.16.tgz",
       "dev": true,
       "license": "MIT"
     },
+    "node_modules/@standard-schema/spec": {
+      "version": "1.1.0",
+      "resolved": "https://registry.npmjs.org/@standard-schema/spec/-/spec-1.1.0.tgz",
+      "integrity": "sha512-l2aFy5jALhniG5HgqrD6jXLi/rUWrKvqN/qJx6yoJsgKhblVd+iqqU4RCXavm/jPityDo5TCvKMnpjKnOriy0w==",
+      "license": "MIT"
+    },
+    "node_modules/@standard-schema/utils": {
+      "version": "0.3.0",
+      "resolved": "https://registry.npmjs.org/@standard-schema/utils/-/utils-0.3.0.tgz",
+      "integrity": "sha512-e7Mew686owMaPJVNNLs55PUvgz371nKgwsc4vxE49zsODpJEnxgxRo2y/OKrqueavXgZNMDVj3DdHFlaSAeU8g==",
+      "license": "MIT"
+    },
     "node_modules/@tybys/wasm-util": {
       "version": "0.10.1",
       "resolved": "https://registry.npmjs.org/@tybys/wasm-util/-/wasm-util-0.10.1.tgz",
         "tslib": "^2.4.0"
       }
     },
+    "node_modules/@types/d3-array": {
+      "version": "3.2.2",
+      "resolved": "https://registry.npmjs.org/@types/d3-array/-/d3-array-3.2.2.tgz",
+      "integrity": "sha512-hOLWVbm7uRza0BYXpIIW5pxfrKe0W+D5lrFiAEYR+pb6w3N2SwSMaJbXdUfSEv+dT4MfHBLtn5js0LAWaO6otw==",
+      "license": "MIT"
+    },
+    "node_modules/@types/d3-color": {
+      "version": "3.1.3",
+      "resolved": "https://registry.npmjs.org/@types/d3-color/-/d3-color-3.1.3.tgz",
+      "integrity": "sha512-iO90scth9WAbmgv7ogoq57O9YpKmFBbmoEoCHDB2xMBY0+/KVrqAaCDyCE16dUspeOvIxFFRI+0sEtqDqy2b4A==",
+      "license": "MIT"
+    },
+    "node_modules/@types/d3-ease": {
+      "version": "3.0.2",
+      "resolved": "https://registry.npmjs.org/@types/d3-ease/-/d3-ease-3.0.2.tgz",
+      "integrity": "sha512-NcV1JjO5oDzoK26oMzbILE6HW7uVXOHLQvHshBUW4UMdZGfiY6v5BeQwh9a9tCzv+CeefZQHJt5SRgK154RtiA==",
+      "license": "MIT"
+    },
+    "node_modules/@types/d3-interpolate": {
+      "version": "3.0.4",
+      "resolved": "https://registry.npmjs.org/@types/d3-interpolate/-/d3-interpolate-3.0.4.tgz",
+      "integrity": "sha512-mgLPETlrpVV1YRJIglr4Ez47g7Yxjl1lj7YKsiMCb27VJH9W8NVM6Bb9d8kkpG/uAQS5AmbA48q2IAolKKo1MA==",
+      "license": "MIT",
+      "dependencies": {
+        "@types/d3-color": "*"
+      }
+    },
+    "node_modules/@types/d3-path": {
+      "version": "3.1.1",
+      "resolved": "https://registry.npmjs.org/@types/d3-path/-/d3-path-3.1.1.tgz",
+      "integrity": "sha512-VMZBYyQvbGmWyWVea0EHs/BwLgxc+MKi1zLDCONksozI4YJMcTt8ZEuIR4Sb1MMTE8MMW49v0IwI5+b7RmfWlg==",
+      "license": "MIT"
+    },
+    "node_modules/@types/d3-scale": {
+      "version": "4.0.9",
+      "resolved": "https://registry.npmjs.org/@types/d3-scale/-/d3-scale-4.0.9.tgz",
+      "integrity": "sha512-dLmtwB8zkAeO/juAMfnV+sItKjlsw2lKdZVVy6LRr0cBmegxSABiLEpGVmSJJ8O08i4+sGR6qQtb6WtuwJdvVw==",
+      "license": "MIT",
+      "dependencies": {
+        "@types/d3-time": "*"
+      }
+    },
+    "node_modules/@types/d3-shape": {
+      "version": "3.1.8",
+      "resolved": "https://registry.npmjs.org/@types/d3-shape/-/d3-shape-3.1.8.tgz",
+      "integrity": "sha512-lae0iWfcDeR7qt7rA88BNiqdvPS5pFVPpo5OfjElwNaT2yyekbM0C9vK+yqBqEmHr6lDkRnYNoTBYlAgJa7a4w==",
+      "license": "MIT",
+      "dependencies": {
+        "@types/d3-path": "*"
+      }
+    },
+    "node_modules/@types/d3-time": {
+      "version": "3.0.4",
+      "resolved": "https://registry.npmjs.org/@types/d3-time/-/d3-time-3.0.4.tgz",
+      "integrity": "sha512-yuzZug1nkAAaBlBBikKZTgzCeA+k1uy4ZFwWANOfKw5z5LRhV0gNA7gNkKm7HoK+HRN0wX3EkxGk0fpbWhmB7g==",
+      "license": "MIT"
+    },
+    "node_modules/@types/d3-timer": {
+      "version": "3.0.2",
+      "resolved": "https://registry.npmjs.org/@types/d3-timer/-/d3-timer-3.0.2.tgz",
+      "integrity": "sha512-Ps3T8E8dZDam6fUyNiMkekK3XUsaUEik+idO9/YjPtfj2qruF8tFBXS7XhtE4iIXBLxhmLjP3SXpLhVf21I9Lw==",
+      "license": "MIT"
+    },
     "node_modules/@types/estree": {
       "version": "1.0.8",
       "resolved": "https://registry.npmjs.org/@types/estree/-/estree-1.0.8.tgz",
       "version": "19.2.14",
       "resolved": "https://registry.npmjs.org/@types/react/-/react-19.2.14.tgz",
       "integrity": "sha512-ilcTH/UniCkMdtexkoCN0bI7pMcJDvmQFPvuPvmEaYA/NSfFTAgdUSLAoVjaRJm7+6PvcM+q1zYOwS4wTYMF9w==",
+      "devOptional": true,
       "license": "MIT",
       "peer": true,
       "dependencies": {
         "@types/react": "^19.2.0"
       }
     },
+    "node_modules/@types/use-sync-external-store": {
+      "version": "0.0.6",
+      "resolved": "https://registry.npmjs.org/@types/use-sync-external-store/-/use-sync-external-store-0.0.6.tgz",
+      "integrity": "sha512-zFDAD+tlpf2r4asuHEj0XH6pY6i0g5NeAHPn+15wk3BV6JA69eERFXC1gyGThDkVa1zCyKr5jox1+2LbV/AMLg==",
+      "license": "MIT"
+    },
     "node_modules/@vitejs/plugin-react": {
       "version": "6.0.1",
       "resolved": "https://registry.npmjs.org/@vitejs/plugin-react/-/plugin-react-6.0.1.tgz",
         "url": "https://github.com/chalk/chalk?sponsor=1"
       }
     },
+    "node_modules/clsx": {
+      "version": "2.1.1",
+      "resolved": "https://registry.npmjs.org/clsx/-/clsx-2.1.1.tgz",
+      "integrity": "sha512-eYm0QWBtUrBWZWG0d386OGAw16Z995PiOVo2B7bjWSbHedGl5e0ZWaq65kOGgUSNesEIDkB9ISbTg/JK9dhCZA==",
+      "license": "MIT",
+      "engines": {
+        "node": ">=6"
+      }
+    },
     "node_modules/color-convert": {
       "version": "2.0.1",
       "resolved": "https://registry.npmjs.org/color-convert/-/color-convert-2.0.1.tgz",
       "version": "3.2.3",
       "resolved": "https://registry.npmjs.org/csstype/-/csstype-3.2.3.tgz",
       "integrity": "sha512-z1HGKcYy2xA8AGQfwrn0PAy+PB7X/GSj3UVJW9qKyn43xWa+gl5nXmU4qqLMRzWVLFC8KusUX8T/0kCiOYpAIQ==",
+      "devOptional": true,
       "license": "MIT"
     },
+    "node_modules/d3-array": {
+      "version": "3.2.4",
+      "resolved": "https://registry.npmjs.org/d3-array/-/d3-array-3.2.4.tgz",
+      "integrity": "sha512-tdQAmyA18i4J7wprpYq8ClcxZy3SC31QMeByyCFyRt7BVHdREQZ5lpzoe5mFEYZUWe+oq8HBvk9JjpibyEV4Jg==",
+      "license": "ISC",
+      "dependencies": {
+        "internmap": "1 - 2"
+      },
+      "engines": {
+        "node": ">=12"
+      }
+    },
+    "node_modules/d3-color": {
+      "version": "3.1.0",
+      "resolved": "https://registry.npmjs.org/d3-color/-/d3-color-3.1.0.tgz",
+      "integrity": "sha512-zg/chbXyeBtMQ1LbD/WSoW2DpC3I0mpmPdW+ynRTj/x2DAWYrIY7qeZIHidozwV24m4iavr15lNwIwLxRmOxhA==",
+      "license": "ISC",
+      "engines": {
+        "node": ">=12"
+      }
+    },
+    "node_modules/d3-ease": {
+      "version": "3.0.1",
+      "resolved": "https://registry.npmjs.org/d3-ease/-/d3-ease-3.0.1.tgz",
+      "integrity": "sha512-wR/XK3D3XcLIZwpbvQwQ5fK+8Ykds1ip7A2Txe0yxncXSdq1L9skcG7blcedkOX+ZcgxGAmLX1FrRGbADwzi0w==",
+      "license": "BSD-3-Clause",
+      "engines": {
+        "node": ">=12"
+      }
+    },
+    "node_modules/d3-format": {
+      "version": "3.1.2",
+      "resolved": "https://registry.npmjs.org/d3-format/-/d3-format-3.1.2.tgz",
+      "integrity": "sha512-AJDdYOdnyRDV5b6ArilzCPPwc1ejkHcoyFarqlPqT7zRYjhavcT3uSrqcMvsgh2CgoPbK3RCwyHaVyxYcP2Arg==",
+      "license": "ISC",
+      "engines": {
+        "node": ">=12"
+      }
+    },
+    "node_modules/d3-interpolate": {
+      "version": "3.0.1",
+      "resolved": "https://registry.npmjs.org/d3-interpolate/-/d3-interpolate-3.0.1.tgz",
+      "integrity": "sha512-3bYs1rOD33uo8aqJfKP3JWPAibgw8Zm2+L9vBKEHJ2Rg+viTR7o5Mmv5mZcieN+FRYaAOWX5SJATX6k1PWz72g==",
+      "license": "ISC",
+      "dependencies": {
+        "d3-color": "1 - 3"
+      },
+      "engines": {
+        "node": ">=12"
+      }
+    },
+    "node_modules/d3-path": {
+      "version": "3.1.0",
+      "resolved": "https://registry.npmjs.org/d3-path/-/d3-path-3.1.0.tgz",
+      "integrity": "sha512-p3KP5HCf/bvjBSSKuXid6Zqijx7wIfNW+J/maPs+iwR35at5JCbLUT0LzF1cnjbCHWhqzQTIN2Jpe8pRebIEFQ==",
+      "license": "ISC",
+      "engines": {
+        "node": ">=12"
+      }
+    },
+    "node_modules/d3-scale": {
+      "version": "4.0.2",
+      "resolved": "https://registry.npmjs.org/d3-scale/-/d3-scale-4.0.2.tgz",
+      "integrity": "sha512-GZW464g1SH7ag3Y7hXjf8RoUuAFIqklOAq3MRl4OaWabTFJY9PN/E1YklhXLh+OQ3fM9yS2nOkCoS+WLZ6kvxQ==",
+      "license": "ISC",
+      "dependencies": {
+        "d3-array": "2.10.0 - 3",
+        "d3-format": "1 - 3",
+        "d3-interpolate": "1.2.0 - 3",
+        "d3-time": "2.1.1 - 3",
+        "d3-time-format": "2 - 4"
+      },
+      "engines": {
+        "node": ">=12"
+      }
+    },
+    "node_modules/d3-shape": {
+      "version": "3.2.0",
+      "resolved": "https://registry.npmjs.org/d3-shape/-/d3-shape-3.2.0.tgz",
+      "integrity": "sha512-SaLBuwGm3MOViRq2ABk3eLoxwZELpH6zhl3FbAoJ7Vm1gofKx6El1Ib5z23NUEhF9AsGl7y+dzLe5Cw2AArGTA==",
+      "license": "ISC",
+      "dependencies": {
+        "d3-path": "^3.1.0"
+      },
+      "engines": {
+        "node": ">=12"
+      }
+    },
+    "node_modules/d3-time": {
+      "version": "3.1.0",
+      "resolved": "https://registry.npmjs.org/d3-time/-/d3-time-3.1.0.tgz",
+      "integrity": "sha512-VqKjzBLejbSMT4IgbmVgDjpkYrNWUYJnbCGo874u7MMKIWsILRX+OpX/gTk8MqjpT1A/c6HY2dCA77ZN0lkQ2Q==",
+      "license": "ISC",
+      "dependencies": {
+        "d3-array": "2 - 3"
+      },
+      "engines": {
+        "node": ">=12"
+      }
+    },
+    "node_modules/d3-time-format": {
+      "version": "4.1.0",
+      "resolved": "https://registry.npmjs.org/d3-time-format/-/d3-time-format-4.1.0.tgz",
+      "integrity": "sha512-dJxPBlzC7NugB2PDLwo9Q8JiTR3M3e4/XANkreKSUxF8vvXKqm1Yfq4Q5dl8budlunRVlUUaDUgFt7eA8D6NLg==",
+      "license": "ISC",
+      "dependencies": {
+        "d3-time": "1 - 3"
+      },
+      "engines": {
+        "node": ">=12"
+      }
+    },
+    "node_modules/d3-timer": {
+      "version": "3.0.1",
+      "resolved": "https://registry.npmjs.org/d3-timer/-/d3-timer-3.0.1.tgz",
+      "integrity": "sha512-ndfJ/JxxMd3nw31uyKoY2naivF+r29V+Lc0svZxe1JvvIRmi8hUsrMvdOwgS1o6uBHmiz91geQ0ylPP0aj1VUA==",
+      "license": "ISC",
+      "engines": {
+        "node": ">=12"
+      }
+    },
     "node_modules/debug": {
       "version": "4.4.3",
       "resolved": "https://registry.npmjs.org/debug/-/debug-4.4.3.tgz",
         }
       }
     },
+    "node_modules/decimal.js-light": {
+      "version": "2.5.1",
+      "resolved": "https://registry.npmjs.org/decimal.js-light/-/decimal.js-light-2.5.1.tgz",
+      "integrity": "sha512-qIMFpTMZmny+MMIitAB6D7iVPEorVw6YQRWkvarTkT4tBeSLLiHzcwj6q0MmYSFCiVpiqPJTJEYIrpcPzVEIvg==",
+      "license": "MIT"
+    },
     "node_modules/deep-is": {
       "version": "0.1.4",
       "resolved": "https://registry.npmjs.org/deep-is/-/deep-is-0.1.4.tgz",
       "dev": true,
       "license": "ISC"
     },
+    "node_modules/es-toolkit": {
+      "version": "1.46.0",
+      "resolved": "https://registry.npmjs.org/es-toolkit/-/es-toolkit-1.46.0.tgz",
+      "integrity": "sha512-IToJ6ct9OLl5zz6WsC/1vZEwfSZ7Myil+ygl5Tf30Xjn9AEkzNB4kqp2G7VUJKF1DtTx/ra5M5KLlXvzOg51BA==",
+      "license": "MIT",
+      "workspaces": [
+        "docs",
+        "benchmarks"
+      ]
+    },
     "node_modules/escalade": {
       "version": "3.2.0",
       "resolved": "https://registry.npmjs.org/escalade/-/escalade-3.2.0.tgz",
         "node": ">=0.10.0"
       }
     },
+    "node_modules/eventemitter3": {
+      "version": "5.0.4",
+      "resolved": "https://registry.npmjs.org/eventemitter3/-/eventemitter3-5.0.4.tgz",
+      "integrity": "sha512-mlsTRyGaPBjPedk6Bvw+aqbsXDtoAyAzm5MO7JgU+yVRyMQ5O8bD4Kcci7BS85f93veegeCPkL8R4GLClnjLFw==",
+      "license": "MIT"
+    },
     "node_modules/fast-deep-equal": {
       "version": "3.1.3",
       "resolved": "https://registry.npmjs.org/fast-deep-equal/-/fast-deep-equal-3.1.3.tgz",
         "node": ">= 4"
       }
     },
+    "node_modules/immer": {
+      "version": "10.2.0",
+      "resolved": "https://registry.npmjs.org/immer/-/immer-10.2.0.tgz",
+      "integrity": "sha512-d/+XTN3zfODyjr89gM3mPq1WNX2B8pYsu7eORitdwyA2sBubnTl3laYlBk4sXY5FUa5qTZGBDPJICVbvqzjlbw==",
+      "license": "MIT",
+      "funding": {
+        "type": "opencollective",
+        "url": "https://opencollective.com/immer"
+      }
+    },
     "node_modules/import-fresh": {
       "version": "3.3.1",
       "resolved": "https://registry.npmjs.org/import-fresh/-/import-fresh-3.3.1.tgz",
         "node": ">=0.8.19"
       }
     },
+    "node_modules/internmap": {
+      "version": "2.0.3",
+      "resolved": "https://registry.npmjs.org/internmap/-/internmap-2.0.3.tgz",
+      "integrity": "sha512-5Hh7Y1wQbvY5ooGgPbDaL5iYLAPzMTUrjMulskHLH6wnv/A+1q5rgEaiuqEjB+oxGXIVZs1FF+R/KPN3ZSQYYg==",
+      "license": "ISC",
+      "engines": {
+        "node": ">=12"
+      }
+    },
     "node_modules/is-extglob": {
       "version": "2.1.1",
       "resolved": "https://registry.npmjs.org/is-extglob/-/is-extglob-2.1.1.tgz",
       "resolved": "https://registry.npmjs.org/react-dom/-/react-dom-19.2.5.tgz",
       "integrity": "sha512-J5bAZz+DXMMwW/wV3xzKke59Af6CHY7G4uYLN1OvBcKEsWOs4pQExj86BBKamxl/Ik5bx9whOrvBlSDfWzgSag==",
       "license": "MIT",
+      "peer": true,
       "dependencies": {
         "scheduler": "^0.27.0"
       },
         "react": "^19.2.5"
       }
     },
+    "node_modules/react-is": {
+      "version": "19.2.5",
+      "resolved": "https://registry.npmjs.org/react-is/-/react-is-19.2.5.tgz",
+      "integrity": "sha512-Dn0t8IQhCmeIT3wu+Apm1/YVsJXsGWi6k4sPdnBIdqMVtHtv0IGi6dcpNpNkNac0zB2uUAqNX3MHzN8c+z2rwQ==",
+      "license": "MIT",
+      "peer": true
+    },
+    "node_modules/react-redux": {
+      "version": "9.2.0",
+      "resolved": "https://registry.npmjs.org/react-redux/-/react-redux-9.2.0.tgz",
+      "integrity": "sha512-ROY9fvHhwOD9ySfrF0wmvu//bKCQ6AeZZq1nJNtbDC+kk5DuSuNX/n6YWYF/SYy7bSba4D4FSz8DJeKY/S/r+g==",
+      "license": "MIT",
+      "peer": true,
+      "dependencies": {
+        "@types/use-sync-external-store": "^0.0.6",
+        "use-sync-external-store": "^1.4.0"
+      },
+      "peerDependencies": {
+        "@types/react": "^18.2.25 || ^19",
+        "react": "^18.0 || ^19",
+        "redux": "^5.0.0"
+      },
+      "peerDependenciesMeta": {
+        "@types/react": {
+          "optional": true
+        },
+        "redux": {
+          "optional": true
+        }
+      }
+    },
+    "node_modules/recharts": {
+      "version": "3.8.1",
+      "resolved": "https://registry.npmjs.org/recharts/-/recharts-3.8.1.tgz",
+      "integrity": "sha512-mwzmO1s9sFL0TduUpwndxCUNoXsBw3u3E/0+A+cLcrSfQitSG62L32N69GhqUrrT5qKcAE3pCGVINC6pqkBBQg==",
+      "license": "MIT",
+      "workspaces": [
+        "www"
+      ],
+      "dependencies": {
+        "@reduxjs/toolkit": "^1.9.0 || 2.x.x",
+        "clsx": "^2.1.1",
+        "decimal.js-light": "^2.5.1",
+        "es-toolkit": "^1.39.3",
+        "eventemitter3": "^5.0.1",
+        "immer": "^10.1.1",
+        "react-redux": "8.x.x || 9.x.x",
+        "reselect": "5.1.1",
+        "tiny-invariant": "^1.3.3",
+        "use-sync-external-store": "^1.2.2",
+        "victory-vendor": "^37.0.2"
+      },
+      "engines": {
+        "node": ">=18"
+      },
+      "peerDependencies": {
+        "react": "^16.8.0 || ^17.0.0 || ^18.0.0 || ^19.0.0",
+        "react-dom": "^16.0.0 || ^17.0.0 || ^18.0.0 || ^19.0.0",
+        "react-is": "^16.8.0 || ^17.0.0 || ^18.0.0 || ^19.0.0"
+      }
+    },
+    "node_modules/redux": {
+      "version": "5.0.1",
+      "resolved": "https://registry.npmjs.org/redux/-/redux-5.0.1.tgz",
+      "integrity": "sha512-M9/ELqF6fy8FwmkpnF0S3YKOqMyoWJ4+CS5Efg2ct3oY9daQvd/Pc71FpGZsVsbl3Cpb+IIcjBDUnnyBdQbq4w==",
+      "license": "MIT",
+      "peer": true
+    },
+    "node_modules/redux-thunk": {
+      "version": "3.1.0",
+      "resolved": "https://registry.npmjs.org/redux-thunk/-/redux-thunk-3.1.0.tgz",
+      "integrity": "sha512-NW2r5T6ksUKXCabzhL9z+h206HQw/NJkcLm1GPImRQ8IzfXwRGqjVhKJGauHirT0DAuyy6hjdnMZaRoAcy0Klw==",
+      "license": "MIT",
+      "peerDependencies": {
+        "redux": "^5.0.0"
+      }
+    },
+    "node_modules/reselect": {
+      "version": "5.1.1",
+      "resolved": "https://registry.npmjs.org/reselect/-/reselect-5.1.1.tgz",
+      "integrity": "sha512-K/BG6eIky/SBpzfHZv/dd+9JBFiS4SWV7FIujVyJRux6e45+73RaUHXLmIR1f7WOMaQ0U1km6qwklRQxpJJY0w==",
+      "license": "MIT"
+    },
     "node_modules/resolve-from": {
       "version": "4.0.0",
       "resolved": "https://registry.npmjs.org/resolve-from/-/resolve-from-4.0.0.tgz",
         "node": ">=8"
       }
     },
+    "node_modules/tiny-invariant": {
+      "version": "1.3.3",
+      "resolved": "https://registry.npmjs.org/tiny-invariant/-/tiny-invariant-1.3.3.tgz",
+      "integrity": "sha512-+FbBPE1o9QAYvviau/qC5SE3caw21q3xkvWKBtja5vgqOWIHHJ3ioaq1VPfn/Szqctz2bU/oYeKd9/z5BL+PVg==",
+      "license": "MIT"
+    },
     "node_modules/tinyglobby": {
       "version": "0.2.16",
       "resolved": "https://registry.npmjs.org/tinyglobby/-/tinyglobby-0.2.16.tgz",
         "punycode": "^2.1.0"
       }
     },
+    "node_modules/use-sync-external-store": {
+      "version": "1.6.0",
+      "resolved": "https://registry.npmjs.org/use-sync-external-store/-/use-sync-external-store-1.6.0.tgz",
+      "integrity": "sha512-Pp6GSwGP/NrPIrxVFAIkOQeyw8lFenOHijQWkUTrDvrF4ALqylP2C/KCkeS9dpUM3KvYRQhna5vt7IL95+ZQ9w==",
+      "license": "MIT",
+      "peerDependencies": {
+        "react": "^16.8.0 || ^17.0.0 || ^18.0.0 || ^19.0.0"
+      }
+    },
+    "node_modules/victory-vendor": {
+      "version": "37.3.6",
+      "resolved": "https://registry.npmjs.org/victory-vendor/-/victory-vendor-37.3.6.tgz",
+      "integrity": "sha512-SbPDPdDBYp+5MJHhBCAyI7wKM3d5ivekigc2Dk2s7pgbZ9wIgIBYGVw4zGHBml/qTFbexrofXW6Gu4noGxrOwQ==",
+      "license": "MIT AND ISC",
+      "dependencies": {
+        "@types/d3-array": "^3.0.3",
+        "@types/d3-ease": "^3.0.0",
+        "@types/d3-interpolate": "^3.0.1",
+        "@types/d3-scale": "^4.0.2",
+        "@types/d3-shape": "^3.1.0",
+        "@types/d3-time": "^3.0.0",
+        "@types/d3-timer": "^3.0.0",
+        "d3-array": "^3.1.6",
+        "d3-ease": "^3.0.1",
+        "d3-interpolate": "^3.0.1",
+        "d3-scale": "^4.0.2",
+        "d3-shape": "^3.1.0",
+        "d3-time": "^3.0.0",
+        "d3-timer": "^3.0.1"
+      }
+    },
     "node_modules/vite": {
       "version": "8.0.9",
       "resolved": "https://registry.npmjs.org/vite/-/vite-8.0.9.tgz",

frontend/package.json CHANGED Viewed

@@ -11,7 +11,8 @@
   },
   "dependencies": {
     "react": "^19.2.5",
-    "react-dom": "^19.2.5"
   },
   "devDependencies": {
     "@eslint/js": "^9.39.4",

   },
   "dependencies": {
     "react": "^19.2.5",
+    "react-dom": "^19.2.5",
+    "recharts": "^3.8.1"
   },
   "devDependencies": {
     "@eslint/js": "^9.39.4",

frontend/src/CodeArenaRL.jsx CHANGED Viewed

@@ -1,5 +1,5 @@
 import React, { useState, useEffect, useRef, useCallback } from "react";
 /* ─────────────────────────────────────────────
    GOOGLE FONTS
@@ -129,6 +129,12 @@ const GlobalStyles = () => (
    TASKS (mirrors server tasks — display only)
 ───────────────────────────────────────────── */
 const TASKS = {
   "easy-1": {
     id: "easy-1", label: "Easy", name: "Fix average_list()", difficulty: "easy",
     description: "Fix syntax errors: missing colon after def and uses length() instead of len().",
@@ -176,37 +182,26 @@ function AnsiLine({ text }) {
 }
 /* ─────────────────────────────────────────────
-   REWARD CHART
 ───────────────────────────────────────────── */
 function RewardChart({ rewards }) {
-  const W = 260, H = 100, PAD = 20;
-  const pts = rewards.map((r, i) => ({
-    x: PAD + (i / Math.max(4, 1)) * (W - PAD * 2),
-    y: PAD + (1 - r) * (H - PAD * 2),
-    r,
-  }));
-  const pathD = pts.length > 1 ? pts.reduce((a, p, i) => i === 0 ? `M${p.x},${p.y}` : a + ` L${p.x},${p.y}`, "") : "";
-  const areaD = pts.length > 1 ? `${pathD} L${pts[pts.length - 1].x},${H - PAD} L${pts[0].x},${H - PAD} Z` : "";
   return (
-    <svg width="100%" viewBox={`0 0 ${W} ${H}`}>
-      <defs>
-        <linearGradient id="rg" x1="0" y1="0" x2="0" y2="1">
-          <stop offset="0%" stopColor="#00ff88" stopOpacity="0.3" />
-          <stop offset="100%" stopColor="#00ff88" stopOpacity="0" />
-        </linearGradient>
-      </defs>
-      {[0, 0.5, 1].map(v => {
-        const y = PAD + (1 - v) * (H - PAD * 2);
-        return <line key={v} x1={PAD} y1={y} x2={W - PAD} y2={y} stroke="#1e293b" strokeWidth="1" strokeDasharray="3,3" />;
-      })}
-      {[1, 2, 3, 4, 5].map(s => (
-        <text key={s} x={PAD + ((s - 1) / 4) * (W - PAD * 2)} y={H - 4}
-          fill="#334155" fontSize="8" textAnchor="middle" fontFamily="JetBrains Mono">{s}</text>
-      ))}
-      {areaD && <path d={areaD} fill="url(#rg)" />}
-      {pathD && <path d={pathD} fill="none" stroke="#00ff88" strokeWidth="2" strokeLinecap="round" strokeLinejoin="round" />}
-      {pts.map((p, i) => <circle key={i} cx={p.x} cy={p.y} r="4" fill="#0a0e1a" stroke={rewardColor(p.r)} strokeWidth="2" />)}
-    </svg>
   );
 }
@@ -226,6 +221,9 @@ export default function CodeArenaRL() {
   /* ── Task & episode state ── */
   const [selectedTask, setSelectedTask] = useState("easy-1");
   const [envState, setEnvState] = useState(null);   // observation from server
   const [uiMode, setUiMode] = useState("idle");      // idle|resetting|agent_thinking|executing|done
   const [episodeLog, setEpisodeLog] = useState([]);
@@ -305,7 +303,7 @@ export default function CodeArenaRL() {
     });
     if (!res.ok) throw new Error(`/reset failed: ${res.status}`);
     const data = await res.json();
-    return data.observation; // { buggy_code, error_log, test_results, previous_attempts }
   }, [envUrl]);
   const envStep = useCallback(async (proposedFix) => {
@@ -463,6 +461,9 @@ export default function CodeArenaRL() {
     setManualCode(""); setTokenEst(0);
     setCollapsedEntries(new Set());
     setErrorBanner("");
   }, []);
   /* ──────────────────────────────────────────
@@ -512,6 +513,7 @@ export default function CodeArenaRL() {
     const { observation: newObs, reward, done } = stepResult;
     const meta = stepResult.info?.execution_metadata || {};
     const passed = meta.test_passed ?? 0;
     const total = meta.test_total ?? task.hints.length + 1;
     const newStep = currentStepCount + 1;
@@ -526,11 +528,19 @@ export default function CodeArenaRL() {
     setStepCount(newStep);
     setRewards(prev => [...prev, reward]);
     setIsDone(done);
     const logEntry = {
       step: newStep,
       code_submitted: fixedCode,
       reward, done, passed, total,
       error_log: newObs?.error_log || "",
       test_results: newObs?.test_results || "",
       timestamp: new Date().toISOString(),
@@ -570,9 +580,9 @@ export default function CodeArenaRL() {
     runningRef.current = true;
     setUiMode("resetting");
-    let initialObs;
     try {
-      initialObs = await envReset(selectedTask);
     } catch (err) {
       setErrorBanner(`🌐 OpenEnv /reset Error: ${err.message}`);
       setUiMode("idle");
@@ -580,6 +590,9 @@ export default function CodeArenaRL() {
       return;
     }
     setEnvState(initialObs);
     setTimeout(() => runStep(initialObs, 0), 400);
   }, [ollamaStatus, envStatus, manualMode, resetEpisode, envReset, selectedTask, runStep]);
@@ -865,7 +878,14 @@ export default function CodeArenaRL() {
             <div className="panel">
               <div className="panel-header">
                 <span style={{ color: "#ff4455" }}>⚠</span>&nbsp;Buggy Code
-                <span style={{ marginLeft: "auto" }}><span className={`badge badge-${task.difficulty}`}>{task.id}</span></span>
               </div>
               <div style={{ padding: 14 }}>
                 <pre className="code-block" style={{ color: "#f8c8c8", maxHeight: 170, overflowY: "auto" }}>
@@ -1009,6 +1029,30 @@ export default function CodeArenaRL() {
               </div>
             )}
             {/* Episode Log */}
             <div className="panel" style={{ flex: 1, display: "flex", flexDirection: "column" }}>
               <div className="panel-header" style={{ justifyContent: "space-between" }}>

 import React, { useState, useEffect, useRef, useCallback } from "react";
+import { LineChart, Line, XAxis, YAxis, Tooltip, ResponsiveContainer, ReferenceLine } from "recharts";
 /* ─────────────────────────────────────────────
    GOOGLE FONTS
    TASKS (mirrors server tasks — display only)
 ───────────────────────────────────────────── */
 const TASKS = {
+  "auto": {
+    id: "auto", label: "Auto", name: "Adaptive Curriculum", difficulty: "info",
+    description: "Automatically selects difficulty based on recent performance history.",
+    hints: ["If avg < 0.4 -> Easy", "If avg < 0.75 -> Medium", "Else -> Hard"],
+    buggy_code: "# Click Start Episode to fetch task",
+  },
   "easy-1": {
     id: "easy-1", label: "Easy", name: "Fix average_list()", difficulty: "easy",
     description: "Fix syntax errors: missing colon after def and uses length() instead of len().",
 }
 /* ─────────────────────────────────────────────
+   REWARD CHART (Recharts)
 ───────────────────────────────────────────── */
 function RewardChart({ rewards }) {
+  const data = rewards.map((r, i) => ({ step: i + 1, reward: r }));
+  for (let i = data.length + 1; i <= 5; i++) {
+    data.push({ step: i, reward: null });
+  }
   return (
+    <div style={{ width: "100%", height: 120 }}>
+      <ResponsiveContainer width="100%" height="100%">
+        <LineChart data={data} margin={{ top: 10, right: 10, left: -20, bottom: 0 }}>
+          <XAxis dataKey="step" stroke="#334155" tick={{ fill: "#334155", fontSize: 10, fontFamily: "'JetBrains Mono',monospace" }} />
+          <YAxis domain={[0, 1]} ticks={[0, 0.5, 1]} stroke="#334155" tick={{ fill: "#334155", fontSize: 10, fontFamily: "'JetBrains Mono',monospace" }} />
+          <ReferenceLine y={0.5} stroke="#334155" strokeDasharray="3 3" />
+          <ReferenceLine y={1.0} stroke="#334155" strokeDasharray="3 3" />
+          <Tooltip contentStyle={{ backgroundColor: "#0f172a", border: "1px solid #1e293b", borderRadius: 4, fontFamily: "'JetBrains Mono',monospace", fontSize: 10 }} itemStyle={{ color: "#00ff88" }} />
+          <Line type="monotone" dataKey="reward" stroke="#00ff88" strokeWidth={2} dot={{ fill: "#0a0e1a", stroke: "#00ff88", strokeWidth: 2, r: 4 }} isAnimationActive={true} />
+        </LineChart>
+      </ResponsiveContainer>
+    </div>
   );
 }
   /* ── Task & episode state ── */
   const [selectedTask, setSelectedTask] = useState("easy-1");
+  const [currentEnvTask, setCurrentEnvTask] = useState("");
+  const [currentEnvDifficulty, setCurrentEnvDifficulty] = useState("");
+  const [currentRewardComponents, setCurrentRewardComponents] = useState({ compile_score: 0, test_ratio: 0, efficiency_score: 0 });
   const [envState, setEnvState] = useState(null);   // observation from server
   const [uiMode, setUiMode] = useState("idle");      // idle|resetting|agent_thinking|executing|done
   const [episodeLog, setEpisodeLog] = useState([]);
     });
     if (!res.ok) throw new Error(`/reset failed: ${res.status}`);
     const data = await res.json();
+    return data; // { observation, info }
   }, [envUrl]);
   const envStep = useCallback(async (proposedFix) => {
     setManualCode(""); setTokenEst(0);
     setCollapsedEntries(new Set());
     setErrorBanner("");
+    setCurrentEnvTask("");
+    setCurrentEnvDifficulty("");
+    setCurrentRewardComponents({ compile_score: 0, test_ratio: 0, efficiency_score: 0 });
   }, []);
   /* ──────────────────────────────────────────
     const { observation: newObs, reward, done } = stepResult;
     const meta = stepResult.info?.execution_metadata || {};
+    const rc = stepResult.info?.reward_components || {};
     const passed = meta.test_passed ?? 0;
     const total = meta.test_total ?? task.hints.length + 1;
     const newStep = currentStepCount + 1;
     setStepCount(newStep);
     setRewards(prev => [...prev, reward]);
     setIsDone(done);
+    setCurrentRewardComponents({
+      compile_score: rc.compile_score || 0,
+      test_ratio: rc.test_ratio || 0,
+      efficiency_score: rc.efficiency || 0,
+    });
     const logEntry = {
       step: newStep,
       code_submitted: fixedCode,
       reward, done, passed, total,
+      compile_score: rc.compile_score || 0,
+      test_ratio: rc.test_ratio || 0,
+      efficiency_score: rc.efficiency || 0,
       error_log: newObs?.error_log || "",
       test_results: newObs?.test_results || "",
       timestamp: new Date().toISOString(),
     runningRef.current = true;
     setUiMode("resetting");
+    let initialResp;
     try {
+      initialResp = await envReset(selectedTask);
     } catch (err) {
       setErrorBanner(`🌐 OpenEnv /reset Error: ${err.message}`);
       setUiMode("idle");
       return;
     }
+    const initialObs = initialResp.observation;
+    setCurrentEnvTask(initialResp.info?.task_id || selectedTask);
+    setCurrentEnvDifficulty(initialResp.info?.difficulty || "");
     setEnvState(initialObs);
     setTimeout(() => runStep(initialObs, 0), 400);
   }, [ollamaStatus, envStatus, manualMode, resetEpisode, envReset, selectedTask, runStep]);
             <div className="panel">
               <div className="panel-header">
                 <span style={{ color: "#ff4455" }}>⚠</span>&nbsp;Buggy Code
+                <span style={{ marginLeft: "auto", display: "flex", gap: 6 }}>
+                  {currentEnvDifficulty && (
+                    <span className={`badge badge-${currentEnvDifficulty.toLowerCase()}`}>
+                      {currentEnvDifficulty}
+                    </span>
+                  )}
+                  <span className={`badge badge-${task.difficulty}`}>{currentEnvTask || task.id}</span>
+                </span>
               </div>
               <div style={{ padding: 14 }}>
                 <pre className="code-block" style={{ color: "#f8c8c8", maxHeight: 170, overflowY: "auto" }}>
               </div>
             )}
+            {/* Live Reward Components */}
+            {stepCount > 0 && (
+              <div className="panel fade-in">
+                <div className="panel-header">🏅 &nbsp;Reward Components</div>
+                <div style={{ padding: "12px 14px", display: "flex", flexDirection: "column", gap: 12 }}>
+                  {[
+                    { label: "Compile Score", val: currentRewardComponents.compile_score },
+                    { label: "Test Pass Ratio", val: currentRewardComponents.test_ratio },
+                    { label: "Efficiency", val: currentRewardComponents.efficiency_score },
+                  ].map(c => (
+                    <div key={c.label}>
+                      <div style={{ display: "flex", justifyContent: "space-between", fontSize: 10, fontFamily: "'JetBrains Mono',monospace", color: "#64748b", marginBottom: 4 }}>
+                        <span>{c.label}</span>
+                        <span style={{ color: rewardColor(c.val) }}>{c.val.toFixed(2)}</span>
+                      </div>
+                      <div className="reward-bar-outer" style={{ marginTop: 0, height: 4 }}>
+                        <div className="reward-bar-inner" style={{ width: `${c.val * 100}%`, background: `linear-gradient(90deg, ${rewardColor(0)}, ${rewardColor(c.val)})` }} />
+                      </div>
+                    </div>
+                  ))}
+                </div>
+              </div>
+            )}
             {/* Episode Log */}
             <div className="panel" style={{ flex: 1, display: "flex", flexDirection: "column" }}>
               <div className="panel-header" style={{ justifyContent: "space-between" }}>

inference.py CHANGED Viewed

@@ -4,21 +4,27 @@ Rewritten for strict OpenEnv parsing.
 """
 import os
 import httpx
 from openai import OpenAI
-def run_task(task_id: str):
     # Retrieve environment variables as instructed
     base_url = os.environ.get("API_BASE_URL")
     api_key = os.environ.get("HF_TOKEN") or os.environ.get("API_KEY")
     model_name = os.environ.get("MODEL_NAME", "Qwen/Qwen2.5-72B-Instruct")
-    # We pass base_url explicitly. If os.environ["API_BASE_URL"] was strictly intended,
-    # it is fine since OpenAI client accepts None for default.
-    client = OpenAI(
-        base_url=base_url,
-        api_key=api_key or "NO_KEY_PROVIDED"
-    )
     # 1. Print the [START] line
     print(f"[START] task={task_id} env=codearena-rl-benchmark model={model_name}")
@@ -58,14 +64,19 @@ def run_task(task_id: str):
         # 3b/c. Call the LLM
         try:
-            completion = client.chat.completions.create(
-                model=model_name,
-                messages=[
-                    {"role": "system", "content": system_prompt},
-                    {"role": "user", "content": user_prompt}
-                ]
-            )
-            proposed_fix = completion.choices[0].message.content
         except Exception as e:
             error_msg = str(e).replace("\n", " ").replace("\r", "")
             # If the LLM call fails, use this fallback fix
@@ -106,6 +117,22 @@ def run_task(task_id: str):
         print(f"[STEP] step={step} action={action_summary} reward={reward:.2f} done={done_str} error={error_msg}")
     # 4. Print [END]
     success = any(r > 0.5 for r in rewards)
     success_str = "true" if success else "false"
     rewards_str = ",".join([f"{r:.2f}" for r in rewards])
@@ -113,12 +140,16 @@ def run_task(task_id: str):
     print(f"[END] success={success_str} steps={step} score={score:.2f} rewards={rewards_str}")
 def main():
     target_task = os.environ.get("CODEARENA_TASK")
     if target_task:
-        run_task(target_task)
     else:
         for t in ["easy", "medium", "hard"]:
-            run_task(t)
 if __name__ == "__main__":
     main()

 """
 import os
+import argparse
 import httpx
+from datetime import datetime
 from openai import OpenAI
+def run_task(task_id: str, backend: str):
     # Retrieve environment variables as instructed
     base_url = os.environ.get("API_BASE_URL")
     api_key = os.environ.get("HF_TOKEN") or os.environ.get("API_KEY")
     model_name = os.environ.get("MODEL_NAME", "Qwen/Qwen2.5-72B-Instruct")
+    hf_pipeline = None
+    client = None
+    if backend == "hf":
+        from transformers import pipeline
+        hf_pipeline = pipeline("text-generation", model=model_name)
+    else:
+        client = OpenAI(
+            base_url=base_url,
+            api_key=api_key or "NO_KEY_PROVIDED"
+        )
     # 1. Print the [START] line
     print(f"[START] task={task_id} env=codearena-rl-benchmark model={model_name}")
         # 3b/c. Call the LLM
         try:
+            if backend == "hf":
+                prompt = f"{system_prompt}\n\n{user_prompt}"
+                output = hf_pipeline(prompt, max_new_tokens=512, return_full_text=False)
+                proposed_fix = output[0]["generated_text"]
+            else:
+                completion = client.chat.completions.create(
+                    model=model_name,
+                    messages=[
+                        {"role": "system", "content": system_prompt},
+                        {"role": "user", "content": user_prompt}
+                    ]
+                )
+                proposed_fix = completion.choices[0].message.content
         except Exception as e:
             error_msg = str(e).replace("\n", " ").replace("\r", "")
             # If the LLM call fails, use this fallback fix
         print(f"[STEP] step={step} action={action_summary} reward={reward:.2f} done={done_str} error={error_msg}")
     # 4. Print [END]
+    timestamp = datetime.now().isoformat()
+    compile_score, test_ratio, efficiency_score = 0.0, 0.0, 0.0
+    if "info" in obs_json and "reward_components" in obs_json["info"]:
+        rc = obs_json["info"]["reward_components"]
+        compile_score = rc.get("compile_score", 0.0)
+        test_ratio = rc.get("test_ratio", 0.0)
+        efficiency_score = rc.get("efficiency", 0.0)
+    final_reward = rewards[-1] if rewards else 0.0
+    csv_path = "rewards_log.csv"
+    write_headers = not os.path.exists(csv_path)
+    with open(csv_path, "a", encoding="utf-8") as f:
+        if write_headers:
+            f.write("timestamp,task_id,step,reward,compile_score,test_ratio,efficiency_score\n")
+        f.write(f"{timestamp},{task_id},{step},{final_reward},{compile_score},{test_ratio},{efficiency_score}\n")
     success = any(r > 0.5 for r in rewards)
     success_str = "true" if success else "false"
     rewards_str = ",".join([f"{r:.2f}" for r in rewards])
     print(f"[END] success={success_str} steps={step} score={score:.2f} rewards={rewards_str}")
 def main():
+    parser = argparse.ArgumentParser(description="CodeArena RL Inference")
+    parser.add_argument("--backend", type=str, choices=["openai", "hf"], default="openai", help="Backend to use for LLM generation.")
+    args = parser.parse_args()
     target_task = os.environ.get("CODEARENA_TASK")
     if target_task:
+        run_task(target_task, args.backend)
     else:
         for t in ["easy", "medium", "hard"]:
+            run_task(t, args.backend)
 if __name__ == "__main__":
     main()

openenv.yaml CHANGED Viewed

@@ -1,7 +1,7 @@
 name: codearena-rl-benchmark
 description: "RL Benchmark for Autonomous Code Repair — iterative debugging with execution feedback"
 version: "1.0.0"
-entrypoint: server.env:CodeArenaEnv
 runtime:
   language: python
@@ -12,6 +12,19 @@ api:
   step: /step
   state: /state
 tasks:
   - id: easy
     path: tasks/easy.json
@@ -22,6 +35,12 @@ tasks:
   - id: hard
     path: tasks/hard.json
     grader: server.grader:grade
 limits:
   step_timeout_seconds: 2

 name: codearena-rl-benchmark
 description: "RL Benchmark for Autonomous Code Repair — iterative debugging with execution feedback"
 version: "1.0.0"
+entrypoint: server.app:CodeArenaEnv
 runtime:
   language: python
   step: /step
   state: /state
+observation_space:
+  type: json
+  schema:
+    buggy_code: string
+    error_log: string
+    test_results: string
+    previous_attempts: list[string]
+action_space:
+  type: json
+  schema:
+    proposed_fix: string
 tasks:
   - id: easy
     path: tasks/easy.json
   - id: hard
     path: tasks/hard.json
     grader: server.grader:grade
+  - id: type_errors
+    path: tasks/type_errors/type_error_1.json
+    grader: server.grader:grade
+  - id: security_bugs
+    path: tasks/security_bugs/security_bug_1.json
+    grader: server.grader:grade
 limits:
   step_timeout_seconds: 2

plot_rewards.py ADDED Viewed

	@@ -0,0 +1,53 @@

+import pandas as pd
+import matplotlib.pyplot as plt
+import os
+def main():
+    os.makedirs('results', exist_ok=True)
+    if not os.path.exists('rewards_log.csv'):
+        print("No rewards_log.csv found. Run inference first.")
+        return
+    try:
+        df = pd.read_csv('rewards_log.csv')
+    except Exception as e:
+        print(f"Error reading CSV: {e}")
+        return
+    if df.empty:
+        print("rewards_log.csv is empty.")
+        return
+    # Plot 1: Reward Curve over Training Steps (using index as training step)
+    plt.figure(figsize=(10, 6))
+    plt.plot(df.index, df['reward'], alpha=0.3, label='Episode Reward')
+    # 10-step rolling average
+    rolling_avg = df['reward'].rolling(window=10, min_periods=1).mean()
+    plt.plot(df.index, rolling_avg, color='red', linewidth=2, label='10-step Rolling Average')
+    plt.xlabel('Training Step')
+    plt.ylabel('Episode Reward (0-1)')
+    plt.title('Reward Curve')
+    plt.legend()
+    plt.grid(True, alpha=0.3)
+    plt.savefig('results/reward_curve.png')
+    plt.close()
+    # Plot 2: Average Reward per Task ID
+    plt.figure(figsize=(10, 6))
+    avg_per_task = df.groupby('task_id')['reward'].mean().sort_values()
+    avg_per_task.plot(kind='barh', color='skyblue')
+    plt.xlabel('Average Episode Reward (0-1)')
+    plt.ylabel('Task ID')
+    plt.title('Average Reward by Task ID')
+    plt.grid(axis='x', alpha=0.3)
+    plt.tight_layout()
+    plt.savefig('results/reward_by_task.png')
+    plt.close()
+    print("Plots saved to results/ directory.")
+if __name__ == "__main__":
+    main()

server/app.py CHANGED Viewed

@@ -42,8 +42,21 @@ class CodeArenaEnv:
         self.is_done = False
         self.step_count = 0
         self.max_steps = 5
     def reset(self, task_id: str = "easy") -> CodeArenaObservation:
         # Priority: exact task_id match → difficulty match → random
         if task_id in TASK_ID_MAP:
             self.current_task = TASK_ID_MAP[task_id]
@@ -71,7 +84,13 @@ class CodeArenaEnv:
             timeout=max(self.current_task.optimal_time_seconds * 10, 2.0),
         )
-        reward = calculate_reward(exec_result, self.current_task)
         self.previous_attempts.append(action.proposed_fix)
         self.last_error_log = exec_result.runtime_errors
@@ -79,14 +98,18 @@ class CodeArenaEnv:
             f"{exec_result.test_passed}/{exec_result.test_total} tests passed."
         )
-        if reward > 0.99 or self.step_count >= self.max_steps:
             self.is_done = True
         info = {
             "execution_metadata": exec_result.model_dump(),
             "task_id": self.current_task.task_id,
         }
-        return self._state(), reward, self.is_done, info
     def _state(self) -> CodeArenaObservation:
         if not self.current_task:
@@ -129,6 +152,10 @@ def api_reset(body: ResetRequest = ResetRequest()):
             "status": "success",
             "message": "Environment reset successfully",
             "observation": obs.model_dump(),
         }
     except Exception:
         traceback.print_exc()

         self.is_done = False
         self.step_count = 0
         self.max_steps = 5
+        self.episode_rewards_history: list[float] = []
     def reset(self, task_id: str = "easy") -> CodeArenaObservation:
+        if task_id == "auto":
+            if not self.episode_rewards_history:
+                task_id = "easy"
+            else:
+                avg_reward = sum(self.episode_rewards_history) / len(self.episode_rewards_history)
+                if avg_reward < 0.4:
+                    task_id = "easy"
+                elif avg_reward <= 0.75:
+                    task_id = "medium"
+                else:
+                    task_id = "hard"
         # Priority: exact task_id match → difficulty match → random
         if task_id in TASK_ID_MAP:
             self.current_task = TASK_ID_MAP[task_id]
             timeout=max(self.current_task.optimal_time_seconds * 10, 2.0),
         )
+        base_reward, reward_components = calculate_reward(exec_result, self.current_task, action.proposed_fix)
+        step_penalty = 0.02 * self.step_count
+        novelty_penalty = 0.1 if action.proposed_fix in self.previous_attempts else 0.0
+        final_reward = base_reward - step_penalty - novelty_penalty
+        final_reward = max(0.001, min(0.999, float(final_reward)))
         self.previous_attempts.append(action.proposed_fix)
         self.last_error_log = exec_result.runtime_errors
             f"{exec_result.test_passed}/{exec_result.test_total} tests passed."
         )
+        if final_reward > 0.99 or self.step_count >= self.max_steps:
             self.is_done = True
+            self.episode_rewards_history.append(final_reward)
+            if len(self.episode_rewards_history) > 5:
+                self.episode_rewards_history.pop(0)
         info = {
             "execution_metadata": exec_result.model_dump(),
             "task_id": self.current_task.task_id,
+            "reward_components": reward_components
         }
+        return self._state(), final_reward, self.is_done, info
     def _state(self) -> CodeArenaObservation:
         if not self.current_task:
             "status": "success",
             "message": "Environment reset successfully",
             "observation": obs.model_dump(),
+            "info": {
+                "task_id": _env.current_task.task_id if _env.current_task else "",
+                "difficulty": _env.current_task.difficulty if _env.current_task else ""
+            }
         }
     except Exception:
         traceback.print_exc()

server/env.py DELETED Viewed

@@ -1,116 +0,0 @@
-import random
-from fastapi import FastAPI, HTTPException
-from contextlib import asynccontextmanager
-from .models import CodeArenaObservation, CodeArenaAction, TaskInfo
-from .executor import run_code_with_tests
-from .grader import calculate_reward, safe_reward, force_valid_reward
-from tasks import ALL_TASKS
-class CodeArenaEnv:
-    def __init__(self):
-        self.tasks = ALL_TASKS
-        self.current_task: TaskInfo = None
-        self.previous_attempts = []
-        self.last_error_log = ""
-        self.last_test_results = ""
-        self.is_done = False
-        self.step_count = 0
-        self.max_steps = 5
-    def reset(self, task_id: str = None) -> CodeArenaObservation:
-        if task_id:
-            matched = [t for t in self.tasks if t.task_id == task_id]
-            self.current_task = matched[0] if matched else random.choice(self.tasks)
-        else:
-            self.current_task = random.choice(self.tasks)
-        self.previous_attempts = []
-        self.last_error_log = ""
-        self.last_test_results = ""
-        self.is_done = False
-        self.step_count = 0
-        return self.state()
-    def step(self, action: CodeArenaAction) -> tuple[CodeArenaObservation, float, bool, dict]:
-        if self.is_done:
-            raise ValueError("Environment is already done. Call reset().")
-        self.step_count += 1
-        # Execute the proposed fix with 10x optimal time as a hard timeout limit
-        exec_result = run_code_with_tests(
-            code=action.proposed_fix,
-            test_code=self.current_task.test_code,
-            timeout=max(self.current_task.optimal_time_seconds * 10, 2.0)
-        )
-        # Calculate Reward
-        reward = safe_reward(calculate_reward(exec_result, self.current_task))
-        reward = max(0.001, min(0.999, float(reward)))
-        # Update State
-        self.previous_attempts.append(action.proposed_fix)
-        self.last_error_log = exec_result.runtime_errors
-        self.last_test_results = f"{exec_result.test_passed}/{exec_result.test_total} tests passed."
-        # Check termination condition
-        if reward > 0.99 or self.step_count >= self.max_steps:
-            self.is_done = True
-        info = {
-            "execution_metadata": exec_result.model_dump(),
-            "task_id": self.current_task.task_id
-        }
-        return self.state(), reward, self.is_done, info
-    def state(self) -> CodeArenaObservation:
-        if not self.current_task:
-            raise ValueError("Environment not initialized. Call reset() first.")
-        return CodeArenaObservation(
-            buggy_code=self.current_task.buggy_code,
-            error_log=self.last_error_log,
-            test_results=self.last_test_results,
-            previous_attempts=self.previous_attempts,
-        )
-# Initialize a global environment instance for the FastAPI wrapper
-_env = CodeArenaEnv()
-@asynccontextmanager
-async def lifespan(app: FastAPI):
-    _env.reset()
-    yield
-app = FastAPI(lifespan=lifespan, title="CodeArena RL Environment")
-@app.post("/reset")
-def api_reset(body: dict = None):
-    task_id = (body or {}).get("task_id")
-    obs = _env.reset(task_id=task_id)
-    return {"message": "Environment reset successfully", "observation": obs.model_dump()}
-@app.post("/step")
-def api_step(action: CodeArenaAction):
-    try:
-        obs, reward, done, info = _env.step(action)
-        # Safety fallback before force_valid_reward
-        if reward is None:
-            reward = 0.5
-        return {
-            "observation": obs.model_dump(),
-            "reward": force_valid_reward(reward),
-            "done": done,
-            "info": info
-        }
-    except ValueError as e:
-        raise HTTPException(status_code=400, detail=str(e))
-@app.get("/state")
-def api_state():
-    try:
-        obs = _env.state()
-        return {"observation": obs.model_dump()}
-    except ValueError as e:
-        raise HTTPException(status_code=400, detail=str(e))

server/grader.py CHANGED Viewed

@@ -1,6 +1,8 @@
 from .models import ExecutionResult, TaskInfo
 def force_valid_reward(value) -> float:
     """Hard guarantee: reward is strictly in (0, 1) — never 0 or 1, no exceptions."""
     try:
@@ -16,30 +18,85 @@ def force_valid_reward(value) -> float:
     return r
 def safe_reward(reward) -> float:
     """Clamp reward to open interval (0, 1) via force_valid_reward."""
     if reward is None:
         reward = 0.5
     return force_valid_reward(reward)
 def normalize_reward(passed: int, total: int) -> float:
     if total == 0:
         return 0.5
     raw = passed / total
     return force_valid_reward(raw)
-def calculate_reward(exec_result: ExecutionResult, task_info: TaskInfo) -> float:
-    reward = normalize_reward(exec_result.test_passed, exec_result.test_total)
-    return force_valid_reward(reward)
 def grade(*args, **kwargs) -> float:
     try:
-        if len(args) == 2:
-            return calculate_reward(args[0], args[1])
         return 0.5
     except Exception:
         return 0.5

+import os
+import json
+from openai import OpenAI
 from .models import ExecutionResult, TaskInfo
 def force_valid_reward(value) -> float:
     """Hard guarantee: reward is strictly in (0, 1) — never 0 or 1, no exceptions."""
     try:
     return r
 def safe_reward(reward) -> float:
     """Clamp reward to open interval (0, 1) via force_valid_reward."""
     if reward is None:
         reward = 0.5
     return force_valid_reward(reward)
 def normalize_reward(passed: int, total: int) -> float:
     if total == 0:
         return 0.5
     raw = passed / total
     return force_valid_reward(raw)
+_LLM_CACHE = {}
+def get_llm_quality_score(proposed_fix: str) -> dict:
+    if proposed_fix in _LLM_CACHE:
+        return _LLM_CACHE[proposed_fix]
+    try:
+        client = OpenAI()
+        response = client.chat.completions.create(
+            model=os.environ.get("JUDGE_MODEL", "gpt-4o-mini"),
+            messages=[
+                {"role": "system", "content": "You are a code judge. Evaluate the provided Python code on a scale of 0.0 to 1.0 for three metrics: code_quality, security, and correctness. Respond with JSON format strictly matching: {\"code_quality\": 0.0, \"security\": 0.0, \"correctness\": 0.0}"},
+                {"role": "user", "content": proposed_fix}
+            ],
+            response_format={"type": "json_object"}
+        )
+        result = json.loads(response.choices[0].message.content)
+        _LLM_CACHE[proposed_fix] = result
+        return result
+    except Exception as e:
+        print(f"LLM judge error: {e}")
+        fallback = {"code_quality": 0.5, "security": 0.5, "correctness": 0.5}
+        _LLM_CACHE[proposed_fix] = fallback
+        return fallback
+def calculate_reward_components(exec_result: ExecutionResult, task_info: TaskInfo, proposed_fix: str) -> dict:
+    compile_score = 1.0 if not exec_result.runtime_errors else 0.0
+    test_ratio = 0.0
+    if exec_result.test_total > 0:
+        test_ratio = exec_result.test_passed / exec_result.test_total
+    efficiency = 0.0
+    if test_ratio == 1.0:
+        if exec_result.execution_time_seconds <= task_info.optimal_time_seconds:
+            efficiency = 1.0
+        else:
+            ratio = exec_result.execution_time_seconds / max(0.001, task_info.optimal_time_seconds)
+            efficiency = max(0.0, 1.0 - (ratio - 1.0) / 2.0)
+    llm_scores = get_llm_quality_score(proposed_fix)
+    return {
+        "compile_score": compile_score,
+        "test_ratio": test_ratio,
+        "efficiency": efficiency,
+        "llm_correctness": float(llm_scores.get("correctness", 0.5)),
+        "llm_security": float(llm_scores.get("security", 0.5)),
+        "llm_quality": float(llm_scores.get("code_quality", 0.5))
+    }
+def calculate_reward(exec_result: ExecutionResult, task_info: TaskInfo, proposed_fix: str) -> tuple[float, dict]:
+    comps = calculate_reward_components(exec_result, task_info, proposed_fix)
+    base_reward = (
+        0.25 * comps["compile_score"] +
+        0.30 * comps["test_ratio"] +
+        0.15 * comps["efficiency"] +
+        0.15 * comps["llm_correctness"] +
+        0.10 * comps["llm_security"] +
+        0.05 * comps["llm_quality"]
+    )
+    return base_reward, comps
 def grade(*args, **kwargs) -> float:
     try:
+        if len(args) == 3:
+            return calculate_reward(args[0], args[1], args[2])[0]
         return 0.5
     except Exception:
         return 0.5

tasks/__init__.py CHANGED Viewed

@@ -1,5 +1,11 @@
 from .easy import EASY_TASK
 from .medium import MEDIUM_TASK
 from .hard import HARD_TASK
-ALL_TASKS = [EASY_TASK, MEDIUM_TASK, HARD_TASK]

 from .easy import EASY_TASK
 from .medium import MEDIUM_TASK
 from .hard import HARD_TASK
+from .type_errors.type_error_1 import TASK as TE1
+from .type_errors.type_error_2 import TASK as TE2
+from .type_errors.type_error_3 import TASK as TE3
+from .security_bugs.security_bug_1 import TASK as SB1
+from .security_bugs.security_bug_2 import TASK as SB2
+from .security_bugs.security_bug_3 import TASK as SB3
+ALL_TASKS = [EASY_TASK, MEDIUM_TASK, HARD_TASK, TE1, TE2, TE3, SB1, SB2, SB3]

tasks/security_bugs/security_bug_1.json ADDED Viewed

	@@ -0,0 +1,8 @@

+{
+  "task_id": "security_bugs-1",
+  "difficulty": "security_bugs",
+  "description": "Fix the function to parse JSON safely without using eval().",
+  "buggy_code": "import json\ndef parse_user_data(data_string):\n    return eval(data_string)",
+  "test_code": "\nimport unittest\nimport inspect\nclass TestSecurity1(unittest.TestCase):\n    def test_normal(self):\n        self.assertEqual(parse_user_data('{\"name\": \"alice\"}'), {\"name\": \"alice\"})\n    def test_security(self):\n        source = inspect.getsource(parse_user_data)\n        self.assertNotIn(\"eval(\", source)\n",
+  "optimal_time_seconds": 0.05
+}

tasks/security_bugs/security_bug_1.py ADDED Viewed

	@@ -0,0 +1,21 @@

+from server.models import TaskInfo
+TASK = TaskInfo(
+    task_id="security_bugs-1",
+    difficulty="security_bugs",
+    description="Fix the function to parse JSON safely without using eval().",
+    buggy_code="""import json
+def parse_user_data(data_string):
+    return eval(data_string)""",
+    test_code="""
+import unittest
+import inspect
+class TestSecurity1(unittest.TestCase):
+    def test_normal(self):
+        self.assertEqual(parse_user_data('{"name": "alice"}'), {"name": "alice"})
+    def test_security(self):
+        source = inspect.getsource(parse_user_data)
+        self.assertNotIn("eval(", source)
+""",
+    optimal_time_seconds=0.05
+)

tasks/security_bugs/security_bug_2.json ADDED Viewed

	@@ -0,0 +1,8 @@

+{
+  "task_id": "security_bugs-2",
+  "difficulty": "security_bugs",
+  "description": "Remove the hardcoded secret token and load it from the os.environ dictionary as 'API_TOKEN'.",
+  "buggy_code": "import os\ndef get_api_token():\n    token = \"secret_12345\"\n    return token",
+  "test_code": "\nimport unittest\nimport inspect\nimport os\nclass TestSecurity2(unittest.TestCase):\n    def test_normal(self):\n        os.environ['API_TOKEN'] = 'my_secure_token'\n        self.assertEqual(get_api_token(), 'my_secure_token')\n    def test_security(self):\n        source = inspect.getsource(get_api_token)\n        self.assertNotIn(\"secret_12345\", source)\n",
+  "optimal_time_seconds": 0.05
+}

tasks/security_bugs/security_bug_2.py ADDED Viewed

	@@ -0,0 +1,24 @@

+from server.models import TaskInfo
+TASK = TaskInfo(
+    task_id="security_bugs-2",
+    difficulty="security_bugs",
+    description="Remove the hardcoded secret token and load it from the os.environ dictionary as 'API_TOKEN'.",
+    buggy_code="""import os
+def get_api_token():
+    token = "secret_12345"
+    return token""",
+    test_code="""
+import unittest
+import inspect
+import os
+class TestSecurity2(unittest.TestCase):
+    def test_normal(self):
+        os.environ['API_TOKEN'] = 'my_secure_token'
+        self.assertEqual(get_api_token(), 'my_secure_token')
+    def test_security(self):
+        source = inspect.getsource(get_api_token)
+        self.assertNotIn("secret_12345", source)
+""",
+    optimal_time_seconds=0.05
+)

tasks/security_bugs/security_bug_3.json ADDED Viewed

	@@ -0,0 +1,8 @@

+{
+  "task_id": "security_bugs-3",
+  "difficulty": "security_bugs",
+  "description": "Fix the ping command to avoid shell injection. Use a list of arguments and shell=False.",
+  "buggy_code": "import subprocess\ndef ping_host(host):\n    return subprocess.check_output(f\"ping -c 1 {host}\", shell=True)",
+  "test_code": "\nimport unittest\nimport inspect\nclass TestSecurity3(unittest.TestCase):\n    def test_security(self):\n        source = inspect.getsource(ping_host)\n        self.assertNotIn(\"shell=True\", source.replace(\" \", \"\"))\n        self.assertIn(\"[\", source)\n",
+  "optimal_time_seconds": 0.05
+}

tasks/security_bugs/security_bug_3.py ADDED Viewed

	@@ -0,0 +1,20 @@

+from server.models import TaskInfo
+TASK = TaskInfo(
+    task_id="security_bugs-3",
+    difficulty="security_bugs",
+    description="Fix the ping command to avoid shell injection. Use a list of arguments and shell=False.",
+    buggy_code="""import subprocess
+def ping_host(host):
+    return subprocess.check_output(f"ping -c 1 {host}", shell=True)""",
+    test_code="""
+import unittest
+import inspect
+class TestSecurity3(unittest.TestCase):
+    def test_security(self):
+        source = inspect.getsource(ping_host)
+        self.assertNotIn("shell=True", source.replace(" ", ""))
+        self.assertIn("[", source)
+""",
+    optimal_time_seconds=0.05
+)

tasks/type_errors/type_error_1.json ADDED Viewed

	@@ -0,0 +1,8 @@

+{
+  "task_id": "type_errors-1",
+  "difficulty": "type_errors",
+  "description": "Fix the function to sum a list of numbers that might be passed as strings. It currently tries to add int and str.",
+  "buggy_code": "def sum_all(items):\n    total = 0\n    for item in items:\n        total = total + item\n    return total",
+  "test_code": "\nimport unittest\nclass TestTypeError1(unittest.TestCase):\n    def test_normal(self):\n        self.assertEqual(sum_all([1, 2, 3]), 6)\n    def test_strings(self):\n        self.assertEqual(sum_all(['1', '2', '3']), 6)\n    def test_mixed(self):\n        self.assertEqual(sum_all([1, '2', 3]), 6)\n",
+  "optimal_time_seconds": 0.05
+}

tasks/type_errors/type_error_1.py ADDED Viewed

	@@ -0,0 +1,23 @@

+from server.models import TaskInfo
+TASK = TaskInfo(
+    task_id="type_errors-1",
+    difficulty="type_errors",
+    description="Fix the function to sum a list of numbers that might be passed as strings. It currently tries to add int and str.",
+    buggy_code="""def sum_all(items):
+    total = 0
+    for item in items:
+        total = total + item
+    return total""",
+    test_code="""
+import unittest
+class TestTypeError1(unittest.TestCase):
+    def test_normal(self):
+        self.assertEqual(sum_all([1, 2, 3]), 6)
+    def test_strings(self):
+        self.assertEqual(sum_all(['1', '2', '3']), 6)
+    def test_mixed(self):
+        self.assertEqual(sum_all([1, '2', 3]), 6)
+""",
+    optimal_time_seconds=0.05
+)

tasks/type_errors/type_error_2.json ADDED Viewed

	@@ -0,0 +1,8 @@

+{
+  "task_id": "type_errors-2",
+  "difficulty": "type_errors",
+  "description": "Fix the function to count frequencies. It incorrectly calls .append() on a dict.",
+  "buggy_code": "def count_frequencies(words):\n    counts = {}\n    for word in words:\n        if word not in counts:\n            counts.append({word: 1})\n        else:\n            counts[word] += 1\n    return counts",
+  "test_code": "\nimport unittest\nclass TestTypeError2(unittest.TestCase):\n    def test_normal(self):\n        self.assertEqual(count_frequencies(['apple', 'banana', 'apple']), {'apple': 2, 'banana': 1})\n    def test_empty(self):\n        self.assertEqual(count_frequencies([]), {})\n",
+  "optimal_time_seconds": 0.05
+}

tasks/type_errors/type_error_2.py ADDED Viewed

	@@ -0,0 +1,24 @@

+from server.models import TaskInfo
+TASK = TaskInfo(
+    task_id="type_errors-2",
+    difficulty="type_errors",
+    description="Fix the function to count frequencies. It incorrectly calls .append() on a dict.",
+    buggy_code="""def count_frequencies(words):
+    counts = {}
+    for word in words:
+        if word not in counts:
+            counts.append({word: 1})
+        else:
+            counts[word] += 1
+    return counts""",
+    test_code="""
+import unittest
+class TestTypeError2(unittest.TestCase):
+    def test_normal(self):
+        self.assertEqual(count_frequencies(['apple', 'banana', 'apple']), {'apple': 2, 'banana': 1})
+    def test_empty(self):
+        self.assertEqual(count_frequencies([]), {})
+""",
+    optimal_time_seconds=0.05
+)

tasks/type_errors/type_error_3.json ADDED Viewed

	@@ -0,0 +1,8 @@

+{
+  "task_id": "type_errors-3",
+  "difficulty": "type_errors",
+  "description": "Fix the function to format names. It incorrectly calls .upper() on an int ID.",
+  "buggy_code": "def format_records(records):\n    formatted = []\n    for user_id, name in records:\n        formatted.append(f\"{user_id.upper()} - {name.upper()}\")\n    return formatted",
+  "test_code": "\nimport unittest\nclass TestTypeError3(unittest.TestCase):\n    def test_normal(self):\n        self.assertEqual(format_records([(1, 'alice'), (2, 'bob')]), ['1 - ALICE', '2 - BOB'])\n",
+  "optimal_time_seconds": 0.05
+}

tasks/type_errors/type_error_3.py ADDED Viewed

	@@ -0,0 +1,19 @@

+from server.models import TaskInfo
+TASK = TaskInfo(
+    task_id="type_errors-3",
+    difficulty="type_errors",
+    description="Fix the function to format names. It incorrectly calls .upper() on an int ID.",
+    buggy_code="""def format_records(records):
+    formatted = []
+    for user_id, name in records:
+        formatted.append(f"{user_id.upper()} - {name.upper()}")
+    return formatted""",
+    test_code="""
+import unittest
+class TestTypeError3(unittest.TestCase):
+    def test_normal(self):
+        self.assertEqual(format_records([(1, 'alice'), (2, 'bob')]), ['1 - ALICE', '2 - BOB'])
+""",
+    optimal_time_seconds=0.05
+)

train_grpo.ipynb ADDED Viewed

	@@ -0,0 +1,138 @@

+{
+  "cells": [
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "# GRPO Training with CodeArena RL Benchmark\n",
+        "\n",
+        "This notebook demonstrates how to connect our custom `codearena-rl-benchmark` environment to HuggingFace's `trl.GRPOTrainer`."
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "!pip install trl transformers datasets openenv-py httpx\n",
+        "!git clone https://github.com/havinashpatil/meta.git\n",
+        "!cd meta && pip install -r requirements.txt"
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "import torch\n",
+        "from datasets import Dataset\n",
+        "from transformers import AutoModelForCausalLM, AutoTokenizer\n",
+        "from trl import GRPOConfig, GRPOTrainer\n",
+        "import httpx\n",
+        "\n",
+        "# Start the backend server in the background (Colab trick)\n",
+        "import subprocess\n",
+        "import time\n",
+        "subprocess.Popen([\"uvicorn\", \"server.app:app\", \"--port\", \"7860\", \"--app-dir\", \"meta\"])\n",
+        "time.sleep(5)  # Wait for server to start"
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "def codearena_reward_func(completions, prompts):\n",
+        "    \"\"\"\n",
+        "    Reward function that queries the CodeArena OpenEnv server.\n",
+        "    For each proposed fix in `completions`, we step the environment.\n",
+        "    \"\"\"\n",
+        "    rewards = []\n",
+        "    for completion in completions:\n",
+        "        # Clean the generated code\n",
+        "        proposed_fix = completion[0].get('content', '').strip()\n",
+        "        if proposed_fix.startswith('```python'):\n",
+        "            proposed_fix = proposed_fix[9:].replace('```', '').strip()\n",
+        "            \n",
+        "        try:\n",
+        "            # Step the environment\n",
+        "            res = httpx.post(\n",
+        "                \"http://localhost:7860/step\",\n",
+        "                json={\"proposed_fix\": proposed_fix},\n",
+        "                timeout=10.0\n",
+        "            )\n",
+        "            res.raise_for_status()\n",
+        "            reward = res.json().get('reward', 0.0)\n",
+        "            rewards.append(reward)\n",
+        "        except Exception as e:\n",
+        "            print(f\"Env Error: {e}\")\n",
+        "            rewards.append(0.0)\n",
+        "            \n",
+        "    return rewards"
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "# Load Model\n",
+        "model_name = \"Qwen/Qwen2.5-Coder-1.5B\"\n",
+        "model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype=torch.bfloat16, device_map=\"auto\")\n",
+        "tokenizer = AutoTokenizer.from_pretrained(model_name)\n",
+        "tokenizer.pad_token = tokenizer.eos_token\n",
+        "\n",
+        "# Sample training dataset (prompts extracted from tasks)\n",
+        "# In a real setup, you'd reset the env for each prompt to get the initial buggy_code.\n",
+        "dataset = Dataset.from_dict({\n",
+        "    \"prompt\": [\n",
+        "        \"Fix this Python code:\\ndef average_list(numbers)\\n    if length(numbers) == 0:\\n        return 0\\n    return sum(numbers) / length(numbers)\"\n",
+        "    ]\n",
+        "})\n",
+        "\n",
+        "# Initialize GRPO Trainer\n",
+        "training_args = GRPOConfig(\n",
+        "    output_dir=\"./codearena-grpo\",\n",
+        "    learning_rate=1e-5,\n",
+        "    max_steps=50,\n",
+        "    per_device_train_batch_size=2,\n",
+        "    gradient_accumulation_steps=2,\n",
+        ")\n",
+        "\n",
+        "trainer = GRPOTrainer(\n",
+        "    model=model,\n",
+        "    reward_funcs=codearena_reward_func,\n",
+        "    args=training_args,\n",
+        "    train_dataset=dataset,\n",
+        ")\n",
+        "\n",
+        "trainer.train()"
+      ]
+    }
+  ],
+  "metadata": {
+    "kernelspec": {
+      "display_name": "Python 3",
+      "language": "python",
+      "name": "python3"
+    },
+    "language_info": {
+      "codemirror_mode": {
+        "name": "ipython",
+        "version": 3
+      },
+      "file_extension": ".py",
+      "mimetype": "text/x-python",
+      "name": "python",
+      "nbconvert_exporter": "python",
+      "pygments_lexer": "ipython3",
+      "version": "3.10.12"
+    }
+  },
+  "nbformat": 4,
+  "nbformat_minor": 4
+}