anthropic call no arguments function error

[xufanghui]

Request: {"model":"anthropic/claude-3.7-sonnet:thinking","messages":[{"role":"system","content":"这是一个5x5的迷宫网格，起点位于坐标[1, 4]，终点位于坐标[2, 2]。\n 迷宫中分布着若干墙壁，具体位置为： [(0, 1), (0, 2), (0, 3), (1, 0), (1, 1), (1, 2), (1, 3), (3, 0), (3, 1), (3, 3), (3, 4), (4, 2), (4, 3)]。\n 迷宫的字符表示：\n # # # . #\n # # # # s\n . . e . .\n # # . # #\n . . # # .\n 请使用工具走迷宫，从起点(s)走到终点(e),#表示墙，不要走出5x5网格，坐标的最小值位于左上角[0,0],最大值位于右下角[4,4]。\n 详细的规则或步骤如下：\n 1. 首先记录开始点和结束点，调用 record_start_end 工具。\n 2. 依据实际情况，请使用 up 向上走 或者 down 向下走 或者 left 向左走 或者 right 向右走 工具走一步。\n 3. 调用 is_target 工具判断是否到达终点。\n 4. 如果调用 is_target 工具后返回【1】表示到达终点，否则，请重试第2和3步骤，直到重复执行20步后终止。\n "},{"role":"user","content":"让我们走出指令中的迷宫"},{"role":"assistant","content":"我将帮您走出这个迷宫。根据指令，我需要从起点[1, 4]走到终点[2, 2]。\n\n首先，我需要记录开始点和结束点：","tool_calls":[{"id":"toolu_vrtx_011kcCRHX9oVBJfmx8NAaTtR","type":"function","function":{"name":"record_start_end","arguments":"{"json_string": "{\"start_x\":1,\"start_y\":4,\"target_x\":2,\"target_y\":2}"}"}}]},{"role":"tool","content":"nil","name":"record_start_end","tool_call_id":"toolu_vrtx_011kcCRHX9oVBJfmx8NAaTtR"}],"tools":[{"type":"function","function":{"name":"left","description":"向左","parameters":{"type":"object","properties":{},"required":[]}}},{"type":"function","function":{"name":"up","description":"向上","parameters":{"type":"object","properties":{},"required":[]}}},{"type":"function","function":{"name":"record_start_end","description":"记录开始点和结束点, 输入参数为json字符串如：{"start_x":1,"start_y":4,"target_x":2,"target_y":2}","parameters":{"type":"object","properties":{"json_string":{"type":"string"}},"required":["json_string"]}}},{"type":"function","function":{"name":"right","description":"向右","parameters":{"type":"object","properties":{},"required":[]}}},{"type":"function","function":{"name":"down","description":"向下","parameters":{"type":"object","properties":{},"required":[]}}},{"type":"function","function":{"name":"is_target","description":"判断是否到达终点，返回 1 到达；0 未到达","parameters":{"type":"object","properties":{},"required":[]}}}],"parallel_tool_calls":true,"tool_choice":"auto"}
Response: {"role":"assistant","content":"我会继续帮您走出迷宫。现在已经记录了开始点和结束点，我需要开始移动。根据迷宫布局，我将规划一条从[1,4]到[2,2]的路径。\n\n首先，我向右移动一步：","tool_calls":[{"id":"toolu_vrtx_01HxhoZHdBBxtoqaL7PZA6Wd","type":"function","function":{"name":"right","arguments":""}}]}

error:

thread 'tokio-runtime-worker' panicked at aiscript-vm/src/ai/agent.rs:213:18:
called Result::unwrap() on an Err value: Error("EOF while parsing a value", line: 1, column: 0)
note: run with RUST_BACKTRACE=1 environment variable to display a backtrace

thread 'main' panicked at aiscript/src/main.rs:81:18:
called Result::unwrap() on an Err value: JoinError::Panic(Id(11), "called Result::unwrap() on an Err value: Error("EOF while parsing a value", line: 1, column: 0)", ...)

[Folyd]

Can you share your agent code? Thanks.

[xufanghui]

use std.serde;
let t_x = nil;
let t_y = nil;
let x = nil;
let y = nil;

agent Maze {
    instructions: "You're a master of mazes.Please use the tool to navigate the maze, moving from the starting point (s) to the ending point (e). The symbol '#' represents a wall. Do not go beyond the 5x5 grid. The minimum coordinate value is located at the top-left corner [0, 0], and the maximum coordinate value is located at the bottom-right corner [4, 4].
Detailed rules or steps are as follows:
1. First, record the starting and ending points by calling the record_start_end tool.
2. Based on the actual situation, use the up, down, left, or right tool to move one step.
3. Call the is_target tool to determine if the ending point has been reached.
4. If the is_target tool returns [1], it means the ending point has been reached. Otherwise, repeat steps 2 and 3 until a maximum of 20 steps have been executed.
    ",
    fn record_start_end(json_string:str) {
        """Record the starting and ending points, the input parameter is a json string such as：{"start_x":1,"start_y":4,"target_x":2,"target_y":2} """
        print("start_x: ", json_string);
        let p = serde.from_str(json_string);
        x = p.start_x;
        y = p.start_y;
        t_x = p.target_x;
        t_y = p.target_y;
    }

    fn is_target() -> str {
        """Determine if the game is over; Return 1 to reach it; Return 0 if not."""
        print("x: ",x, "y: ", y, "target_x: ", t_x, "target_y: ", t_y);
        if x == t_x and y == t_y  {
            return "1";
        } else {
            return "0";
        }
    }

    fn up() -> str {
        """up x= x-1"""
        print("up: ");
        x = x-1;
    }
    fn down() -> str {
        """down x=x+1"""
        print("down: ");
        x = x+1;
    }
    fn left() -> str {
        """left y=y-1"""
        print("left: ");
        y = y-1;
    }
    fn right() -> str {
        """right y=y+1"""
        print("right: ");
        y = y+1;
    }
}

Maze.run(input="This is a 5x5 maze grid with the starting point at coordinates [1, 4] and the ending point at coordinates [2, 2].
The maze contains several walls located at: [(0, 1), (0, 2), (0, 3), (1, 0), (1, 1), (1, 2), (1, 3), (3, 0), (3, 1), (3, 3), (3, 4), (4, 2), (4, 3)].
The character representation of the maze:
. # # # .
# # # # s
. . e . .
# # . # #
. . # # .
Let’s escape the maze within the instructions.", debug=true);

[xufanghui]

This is maze.ai agent, thanks