### Create Task API Request Example (cURL)

Source: https://iterx.deep-reinforce.com/docs/index

Example of creating a new code optimization task using a cURL request. This includes setting task details, initial code, sample size, and the model to be used. It demonstrates the required headers for authentication and content type.

```bash
curl -X POST https://iterx.deep-reinforce.com/api/task/create \
  -H "Authorization: sk-xxxxxxxxxxxxx" \
  -H "Content-Type: application/json" \
  -d '{
    "task_name": "GEMM Optimization",
    "task_description": "Optimize matrix multiplication performance to exceed cuBLAS baseline. Evaluation metrics: execution time (ms) and TFLOPS, using relative scoring.",
    "reward_description": "Performance improvement percentage",
    "initial_code": "def gemm(A, B):\n    return A @ B",
    "sample_size": 4,
    "model": "qwen3-235b-instruct-2507"
  }'
```

--------------------------------

### Get Code by ID

Source: https://iterx.deep-reinforce.com/docs/index

Retrieves the complete code content for a given code ID. This endpoint is used to fetch the source code associated with a specific submission.

```APIDOC
## POST /api/get_code_by_id

### Description
Retrieves the complete code content for a given code ID.

### Method
POST

### Endpoint
/api/get_code_by_id

### Parameters
#### Request Body
- **code_id** (string) - Required - Code ID

### Request Example
```json
{
  "code_id": "a1b2c3d4e5f6g7h8"
}
```

### Response
#### Success Response (200)
- **code** (string) - Code content

#### Response Example
```json
{
	"code": 200,
	"message": "success",
	"data": {
		"code": "import torch\n\ndef optimized_gemm(A, B):\n    return torch.matmul(A, B)"
	},
	"request_id": "550e8400-e29b-41d4-a716-446655440000"
}
```
```

--------------------------------

### Get Code by ID (API)

Source: https://iterx.deep-reinforce.com/docs/index

Retrieves the complete source code content for a specific code ID. This is essential for reviewing or analyzing individual code submissions. The request requires a valid `code_id` and authentication.

```curl
curl -X POST https://iterx.deep-reinforce.com/api/get_code_by_id \
  -H "Authorization: sk-xxxxxxxxxxxxx" \
  -H "Content-Type: application/json" \
  -d '{ \
    "code_id": "a1b2c3d4e5f6g7h8" \
  }'
```

--------------------------------

### POST /api/task/create

Source: https://iterx.deep-reinforce.com/docs/index

Creates a new code optimization task with specified parameters.

```APIDOC
## POST /api/task/create

### Description
Creates a new code optimization task. This endpoint allows users to define task names, descriptions, reward criteria, initial code, sample size, and the model to be used for optimization.

### Method
POST

### Endpoint
/api/task/create

### Parameters
#### Query Parameters
- **task_name** (string) - Required - The name of the task.
- **task_description** (string) - Optional - A detailed description of the task, including evaluation methodology and reward criteria.
- **reward_description** (string) - Optional - Describes the reward mechanism for the task.
- **initial_code** (string) - Optional - The initial code snippet for the task. If provided, a code record with sequence number 0 is automatically created.
- **sample_size** (int) - Optional - The number of samples to use for evaluation. Valid values are 1, 2, 4, or 8. Defaults to 1.
- **model** (string) - Optional - The name of the model to use for the task. Defaults to 'qwen3-235b-instruct-2507'.

### Request Example
```json
{
  "task_name": "GEMM Optimization",
  "task_description": "Optimize matrix multiplication performance to exceed cuBLAS baseline. Evaluation metrics: execution time (ms) and TFLOPS, using relative scoring.",
  "reward_description": "Performance improvement percentage",
  "initial_code": "def gemm(A, B):\n    return A @ B",
  "sample_size": 4,
  "model": "qwen3-235b-instruct-2507"
}
```

### Response
#### Success Response (200)
- **task_id** (string) - The unique identifier (UUID) for the created task.
- **id** (int) - The database ID of the created task.
- **initial_code_id** (string) - The ID of the initial code record, if `initial_code` was provided.
- **message** (string) - A confirmation message indicating successful task creation.

#### Response Example
```json
{
  "code": 200,
  "message": "success",
  "data": {
    "task_id": "550e8400-e29b-41d4-a716-446655440000",
    "id": 1,
    "initial_code_id": "a1b2c3d4e5f6g7h8",
    "message": "Task created successfully"
  },
  "request_id": "550e8400-e29b-41d4-a716-446655440000"
}
```

### Notes
- Task creation requires a token balance greater than 0, but does not consume tokens.
- The `sample_size` parameter only accepts the values 1, 2, 4, or 8.
- If `initial_code` is supplied, a corresponding code record is automatically generated by the system.
```

--------------------------------

### Query Training Status

Source: https://iterx.deep-reinforce.com/docs/index

Retrieves the current training progress, evaluation statistics, and code leaderboard for a specified task.

```APIDOC
## POST /api/get_training_status

### Description
Retrieves training progress, evaluation statistics, and code leaderboard for a task.

### Method
POST

### Endpoint
/api/get_training_status

### Parameters
#### Query Parameters
- None

#### Request Body
- **task_id** (string) - Required - Task ID (UUID)

### Request Example
```json
{
  "task_id": "550e8400-e29b-41d4-a716-446655440000"
}
```

### Response
#### Success Response (200)
- **progress_percentage** (float) - Task progress percentage (0-100)
- **total_evaluated** (int) - Total number of evaluated code submissions
- **best** (object) - Best performing code information
  - **code_id** (string) - Best code ID
  - **reward** (float) - Highest reward score
- **top5** (array) - Top 5 code submissions (sorted by score descending)
  - **code_id** (string) - Code ID
  - **reward** (float) - Reward score
- **all_evaluated** (array) - All evaluated code submissions (sorted by score descending)
  - **code_id** (string) - Code ID
  - **reward** (float) - Reward score
- **task_is_finished** (bool) - Whether the task is complete

#### Response Example
```json
{
	"code": 200,
	"message": "success",
	"data": {
		"progress_percentage": 32.0,
		"total_evaluated": 50,
		"best": {
			"code_id": "a1b2c3d4e5f6g7h8",
			"reward": 98.5
		},
		"top5": [
			{
				"code_id": "a1b2c3d4e5f6g7h8",
				"reward": 98.5
			},
			{
				"code_id": "b2c3d4e5f6g7h8i9",
				"reward": 95.2
			},
			{
				"code_id": "c3d4e5f6g7h8i9j0",
				"reward": 92.8
			},
			{
				"code_id": "d4e5f6g7h8i9j0k1",
				"reward": 90.1
			},
			{
				"code_id": "e5f6g7h8i9j0k1l2",
				"reward": 88.5
			}
		],
		"all_evaluated": [
			{
				"code_id": "a1b2c3d4e5f6g7h8",
				"reward": 98.5
			}
		],
		"task_is_finished": false
	},
	"request_id": "550e8400-e29b-41d4-a716-446655440000"
}
```

### Notes
- Only returns evaluated code submissions (where `has_reward = true`).
- Leaderboard is sorted by `reward` in descending order.
- `top5` returns a maximum of 5 code submissions.
```

--------------------------------

### Query Training Status using cURL

Source: https://iterx.deep-reinforce.com/docs/index

Retrieves the current training progress, evaluation statistics, and code leaderboard for a specified task. This endpoint requires a task ID and an authorization header. The response includes detailed metrics about the task's performance and completion status.

```curl
curl -X POST https://iterx.deep-reinforce.com/api/get_training_status \
  -H "Authorization: sk-xxxxxxxxxxxxx" \
  -H "Content-Type: application/json" \
  -d '{
    "task_id": "550e8400-e29b-41d4-a716-446655440000"
  }'
```

--------------------------------

### Submit Code Reward (API)

Source: https://iterx.deep-reinforce.com/docs/index

Evaluates a code submission and submits a reward score along with optional details. This endpoint is crucial for the reinforcement learning loop, allowing the system to learn from the submitted code. `reward` must be non-negative, and if `correctness_check` is false, `reward` must be 0.

```curl
curl -X POST https://iterx.deep-reinforce.com/api/push_code_reward_by_id \
  -H "Authorization: sk-xxxxxxxxxxxxx" \
  -H "Content-Type: application/json" \
  -d '{ \
    "task_id": "550e8400-e29b-41d4-a716-446655440000", \
    "code_id": "a1b2c3d4e5f6g7h8", \
    "reward": 85.5, \
    "correctness_check": true, \
    "details": "", \
    "code_error_msg": "" \
  }'
```

--------------------------------

### Submit Code Reward

Source: https://iterx.deep-reinforce.com/docs/index

Evaluates code and submits a reward score. This endpoint is used to provide feedback on code submissions, including a reward, correctness check, and optional details.

```APIDOC
## POST /api/push_code_reward_by_id

### Description
Evaluates code and submits a reward score.

### Method
POST

### Endpoint
/api/push_code_reward_by_id

### Parameters
#### Request Body
- **task_id** (string) - Required - Task ID
- **code_id** (string) - Required - Code ID
- **reward** (float) - Required - Reward score (must be >= 0)
- **correctness_check** (bool) - Optional - Whether the code passes correctness validation
- **details** (string) - Optional - Evaluation details
- **code_error_msg** (string) - Optional - Code error message

### Request Example
```json
{
  "task_id": "550e8400-e29b-41d4-a716-446655440000",
  "code_id": "a1b2c3d4e5f6g7h8",
  "reward": 85.5,
  "correctness_check": true,
  "details": "",
  "code_error_msg": ""
}
```

### Response
#### Success Response (200)
- **task_id** (string) - Task ID
- **code_id** (string) - Code ID
- **reward** (float) - Reward score
- **correctness_check** (bool) - Correctness check result
- **status** (string) - Task status after reward submission

#### Response Example
```json
{
	"code": 200,
	"message": "success",
	"data": {
		"message": "Scoring successful",
		"task_id": "550e8400-e29b-41d4-a716-446655440000",
		"code_id": "a1b2c3d4e5f6g7h8",
		"reward": 85.5,
		"correctness_check": true,
		"status": "ongoing"
	},
	"request_id": "550e8400-e29b-41d4-a716-446655440000"
}
```

### Reward Rules
- `reward` must be non-negative.
- When `correctness_check = false`, `reward` must be 0.
```

--------------------------------

### Fetch Unevaluated Code IDs (API)

Source: https://iterx.deep-reinforce.com/docs/index

Retrieves a list of unevaluated code IDs for a given task. This endpoint is useful for batch processing or monitoring the status of code submissions. It only returns submissions where `has_reward` is false and is restricted to tasks created by the current user.

```curl
curl -X POST https://iterx.deep-reinforce.com/api/fetch_unevaluated_code_ids \
  -H "Authorization: sk-xxxxxxxxxxxxx" \
  -H "Content-Type: application/json" \
  -d '{ \
    "task_id": "550e8400-e29b-41d4-a716-446655440000" \
  }'
```

--------------------------------

### Fetch Unevaluated Code IDs

Source: https://iterx.deep-reinforce.com/docs/index

Retrieves a list of all unevaluated code IDs for a specified task. This endpoint is useful for batch processing or identifying code submissions that require evaluation.

```APIDOC
## POST /api/fetch_unevaluated_code_ids

### Description
Retrieves a list of all unevaluated code IDs for a specified task.

### Method
POST

### Endpoint
/api/fetch_unevaluated_code_ids

### Parameters
#### Request Body
- **task_id** (string) - Required - Task ID (UUID)

### Request Example
```json
{
  "task_id": "550e8400-e29b-41d4-a716-446655440000"
}
```

### Response
#### Success Response (200)
- **total_code_count** (int) - Total number of unevaluated code submissions
- **code_id_list** (array) - List of unevaluated code IDs
- **task_is_finished** (bool) - Whether the task is complete

#### Response Example
```json
{
	"code": 200,
	"message": "success",
	"data": {
		"total_code_count": 5,
		"code_id_list": ["a1b2c3d4e5f6g7h8", "b2c3d4e5f6g7h8i9", "c3d4e5f6g7h8i9j0", "d4e5f6g7h8i9j0k1", "e5f6g7h8i9j0k1l2"],
		"task_is_finished": false
	},
	"request_id": "550e8400-e29b-41d4-a716-446655440000"
}
```

### Notes
- Only returns unevaluated code submissions (where `has_reward = false`).
- Only accessible for tasks created by the current user.
- Returns an empty array when no pending code submissions exist.
```

=== COMPLETE CONTENT === This response contains all available snippets from this library. No additional content exists. Do not make further requests.