pctx_code_mode 0.2.3

TypeScript code execution engine for AI agents with tool schemas, sandboxed Deno runtime, and Rust callbacks
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
# PCTX Code Mode

A TypeScript code execution engine that enables AI agents to dynamically call tools through generated code. Code Mode converts tool schemas (like MCP tools) into TypeScript interfaces, executes LLM-generated code in a sandboxed Deno runtime, and bridges function calls back to your Rust callbacks.

## Quick Start

```rust
use pctx_code_mode::{CodeMode, CallbackRegistry};
use pctx_code_mode::model::CallbackConfig;
use serde_json::json;
use std::sync::Arc;

#[tokio::main]
async fn main() -> anyhow::Result<()> {
    // 1. Define callback tools with JSON schemas
    let callback = CallbackConfig {
        namespace: "Greeter".to_string(),
        name: "greet".to_string(),
        description: Some("Greets a person by name".to_string()),
        input_schema: Some(json!({
            "type": "object",
            "properties": { "name": { "type": "string" } },
            "required": ["name"]
        })),
        output_schema: Some(json!({
            "type": "object",
            "properties": { "message": { "type": "string" } },
            "required": ["message"]
        })),
    };

    // 2. Create CodeMode instance and add callback
    let mut code_mode = CodeMode::default();
    code_mode.add_callback(&callback)?;

    // 3. Register callback functions that execute when tools are called
    let registry = CallbackRegistry::default();
    registry.add(&callback.id(), Arc::new(|args| {
        Box::pin(async move {
            let name = args
                .and_then(|v| v.get("name"))
                .and_then(|v| v.as_str())
                .unwrap_or("World");
            Ok(serde_json::json!({ "message": format!("Hello, {name}!") }))
        })
    }))?;

    // 4. Execute LLM-generated TypeScript code
    let code = r#"
        async function run() {
            const result = await Greeter.greet({ name: "Alice" });
            return result;
        }
    "#;

    let output = code_mode.execute(code, Some(registry)).await?;

    if output.success {
        println!("Result: {}", serde_json::to_string_pretty(&output.output)?);
    } else {
        eprintln!("Error: {}", output.stderr);
    }

    Ok(())
}
```

## Core Concepts

### 1. CodeMode

The [`CodeMode`] struct is the main execution engine. It provides:

**Builder methods** (chainable):

- `with_server()` / `with_servers()` - Add MCP servers
- `with_callback()` / `with_callbacks()` - Add callback tools

**Registration methods** (mutable):

- `add_server()` / `add_servers()` - Add MCP servers
- `add_callback()` / `add_callbacks()` - Add callback tools
- `add_tool_set()` - Add a pre-built ToolSet directly

**Accessor methods**:

- `tool_sets()` - Get registered ToolSets
- `servers()` - Get registered server configurations
- `callbacks()` - Get registered callback configurations

**Execution methods**:

- `list_functions()` - List all available functions with minimal interfaces
- `get_function_details()` - Get full typed interfaces for specific functions
- `execute()` - Execute TypeScript code in the sandbox

```rust
use pctx_code_mode::CodeMode;
use pctx_code_mode::model::{CallbackConfig, GetFunctionDetailsInput, FunctionId};
use serde_json::json;

let mut code_mode = CodeMode::default();

// Add callback tools
code_mode.add_callback(&CallbackConfig {
    namespace: "DataApi".to_string(),
    name: "fetchData".to_string(),
    description: Some("Fetches data from API".to_string()),
    input_schema: Some(json!({
        "type": "object",
        "properties": { "id": { "type": "integer" } },
        "required": ["id"]
    })),
    output_schema: None,
})?;

// List available functions
let list = code_mode.list_functions();
for func in list.functions {
    println!("{}.{}: {:?}", func.namespace, func.name, func.description);
}

// Get detailed type information
let details = code_mode.get_function_details(GetFunctionDetailsInput {
    functions: vec![
        FunctionId { mod_name: "DataApi".into(), fn_name: "fetchData".into() }
    ],
});
println!("TypeScript definitions:\n{}", details.code);
```

### 2. Tools and ToolSets

[`Tool`]s represent individual functions callable from TypeScript.
They are organized into [`ToolSet`]s (namespaces). Tools can be:

- **MCP tools**: Loaded from MCP servers via `add_server()`
- **Callback tools**: Defined via `CallbackConfig` and `add_callback()`

### 3. Callbacks

[`CallbackFn`] are Rust async functions that execute when TypeScript code calls callback tools.
Register them in a [`CallbackRegistry`] and pass it to `execute()`.

```rust
use pctx_code_mode::{CallbackRegistry, CallbackFn};
use std::sync::Arc;

let registry = CallbackRegistry::default();

let callback: CallbackFn = Arc::new(|args| {
    Box::pin(async move {
        // Extract arguments
        let id = args
            .and_then(|v| v.get("id"))
            .and_then(|v| v.as_i64())
            .ok_or("Missing id")?;

        // Do async work
        let data = fetch_from_database(id).await?;

        // Return JSON result
        Ok(serde_json::to_value(data)?)
    })
});

// Register with namespace.function format
registry.add("DataApi.fetchData", callback)?;
```

### 4. Code Execution

Execute LLM-generated TypeScript code that calls your registered tools.

```rust
let code = r#"
    async function run() {
        // Call your registered tools
        const user = await DataApi.fetchData({ id: 123 });
        const greeting = await Greeter.greet({ name: user.name });

        // Chain multiple calls
        const result = await DataApi.saveData({
            id: user.id,
            message: greeting.message
        });

        // Return the final result
        return result;
    }
"#;

let output = code_mode.execute(code, Some(registry)).await?;

match output.success {
    true => println!("Success: {:?}", output.output),
    false => eprintln!("Error: {}", output.stderr),
}
```

## API Reference

### CodeMode

The main execution engine.

#### `default()`

```rust
let code_mode = CodeMode::default();
```

#### Builder Methods

Chainable methods for fluent construction:

```rust
use pctx_code_mode::CodeMode;
use pctx_code_mode::model::CallbackConfig;
use pctx_config::server::ServerConfig;

// Build with callbacks
let code_mode = CodeMode::default()
    .with_callback(&callback_config)?
    .with_callbacks(&[callback1, callback2])?;

// Build with MCP servers (async)
let code_mode = CodeMode::default()
    .with_server(&server_config).await?
    .with_servers(&server_configs, 30).await?;
```

#### `add_callback(config: &CallbackConfig) -> Result<()>`

Adds a callback-based tool to the code mode.

```rust
use pctx_code_mode::model::CallbackConfig;
use serde_json::json;

code_mode.add_callback(&CallbackConfig {
    namespace: "Logger".to_string(),
    name: "logMessage".to_string(),
    description: Some("Logs a message".to_string()),
    input_schema: Some(json!({
        "type": "object",
        "properties": {
            "message": { "type": "string" }
        },
        "required": ["message"]
    })),
    output_schema: None,
})?;
```

#### `add_server(server: &ServerConfig) -> Result<()>`

Connects to an MCP server and registers its tools.

```rust
use pctx_config::server::ServerConfig;

code_mode.add_server(&server_config).await?;

// Or add multiple servers with a timeout (in seconds)
code_mode.add_servers(&server_configs, 30).await?;
```

#### `list_functions() -> ListFunctionsOutput`

Lists all available functions with their TypeScript interface declarations.

```rust
let list = code_mode.list_functions();
println!("Available functions:\n{}", list.code);
for func in list.functions {
    println!("  {}.{}", func.namespace, func.name);
}
```

#### `get_function_details(input: GetFunctionDetailsInput) -> GetFunctionDetailsOutput`

Gets detailed TypeScript type definitions for specific functions.

```rust
use pctx_code_mode::model::{GetFunctionDetailsInput, FunctionId};

let details = code_mode.get_function_details(GetFunctionDetailsInput {
    functions: vec![
        FunctionId {
            mod_name: "DataApi".to_string(),
            fn_name: "fetchData".to_string(),
        }
    ],
});

println!("TypeScript code:\n{}", details.code);
```

#### `execute(code: &str, callbacks: Option<CallbackRegistry>) -> Result<ExecuteOutput>`

Executes TypeScript code in a sandboxed Deno runtime.

```rust
let output = code_mode.execute(typescript_code, Some(callback_registry)).await?;

if output.success {
    println!("Return value: {:?}", output.output);
    println!("Stdout: {}", output.stdout);
} else {
    eprintln!("Stderr: {}", output.stderr);
}
```

#### Accessor Methods

```rust
// Get registered tool sets
let tool_sets: &[ToolSet] = code_mode.tool_sets();

// Get registered server configurations
let servers: &[ServerConfig] = code_mode.servers();

// Get registered callback configurations
let callbacks: &[CallbackConfig] = code_mode.callbacks();
```

### CallbackRegistry

Thread-safe registry for managing callback functions.

#### `default() -> CallbackRegistry`

```rust
let registry = CallbackRegistry::default();
```

#### `add(id: &str, callback: CallbackFn) -> Result<()>`

Registers a callback with a specific ID (format: `Namespace.functionName`).

```rust
registry.add("DataApi.fetchData", Arc::new(|args| {
    Box::pin(async move {
        // Your implementation
        Ok(serde_json::json!({"result": "data"}))
    })
}))?;
```

#### `has(id: &str) -> bool`

Checks if a callback is registered.

```rust
if registry.has("DataApi.fetchData") {
    println!("Callback is registered");
}
```

### Types

#### `CallbackConfig`

Configuration for defining callback-based tools:

```rust
pub struct CallbackConfig {
    pub name: String,
    pub namespace: String,
    pub description: Option<String>,
    pub input_schema: Option<serde_json::Value>,
    pub output_schema: Option<serde_json::Value>,
}
```

```rust
use pctx_code_mode::model::CallbackConfig;
use serde_json::json;

let config = CallbackConfig {
    namespace: "MyNamespace".to_string(),
    name: "myFunction".to_string(),
    description: Some("Does something useful".to_string()),
    input_schema: Some(json!({
        "type": "object",
        "properties": { "id": { "type": "integer" } },
        "required": ["id"]
    })),
    output_schema: None,
};
```

#### `Tool` and `ToolSet`

Tools represent individual functions callable from TypeScript. They are organized into ToolSets (namespaces). These are typically created internally when you call `add_callback()` or `add_server()`.

```rust
// Access registered tool sets
for tool_set in code_mode.tool_sets() {
    println!("Namespace: {}", tool_set.namespace);
    for tool in &tool_set.tools {
        println!("  - {}: {:?}", tool.fn_name, tool.description);
    }
}
```

#### `ExecuteOutput`

```rust
pub struct ExecuteOutput {
    pub success: bool,
    pub stdout: String,
    pub stderr: String,
    pub output: Option<serde_json::Value>,
}
```

#### `CallbackFn`

Type alias for callback functions:

```rust
pub type CallbackFn = Arc<
    dyn Fn(Option<serde_json::Value>) -> Pin<Box<dyn Future<Output = Result<serde_json::Value, String>> + Send>>
    + Send
    + Sync
>;
```

## Advanced Usage

### Adding MCP Servers

Connect to MCP (Model Context Protocol) servers to automatically register their tools:

```rust
use pctx_config::server::ServerConfig;

// Create server configuration
let server_config = ServerConfig::new_stdio("my-server", "npx", vec!["-y", "my-mcp-server"]);

// Or for HTTP-based servers
let server_config = ServerConfig::new_http("my-server", "https://api.example.com/mcp");

// Add to CodeMode (connects and registers tools)
code_mode.add_server(&server_config).await?;

// Add multiple servers in parallel with timeout
code_mode.add_servers(&[server1, server2], 30).await?;
```

### Dynamic Tool Registration

Register tools at runtime based on configuration:

```rust
use pctx_code_mode::model::CallbackConfig;

for config in tool_configs {
    code_mode.add_callback(&CallbackConfig {
        namespace: config.namespace,
        name: config.name,
        description: Some(config.description),
        input_schema: Some(config.input_schema),
        output_schema: config.output_schema,
    })?;

    // Register the corresponding callback function
    let callback_id = format!("{}.{}", config.namespace, config.name);
    registry.add(&callback_id, create_callback_for_config(&config))?;
}
```

### Async Tool Execution

Callbacks support full async operations:

```rust
registry.add("Database.query", Arc::new(|args| {
    Box::pin(async move {
        let query = args
            .and_then(|v| v.get("sql"))
            .and_then(|v| v.as_str())
            .ok_or("Missing sql parameter")?;

        // Perform async database query
        let pool = get_db_pool().await;
        let rows = sqlx::query(query)
            .fetch_all(&pool)
            .await
            .map_err(|e| e.to_string())?;

        Ok(serde_json::to_value(rows)?)
    })
}))?;
```

### Error Handling

```rust
let output = code_mode.execute(code, Some(registry)).await?;

if !output.success {
    // Check stderr for execution errors
    if output.stderr.contains("TypeError") {
        eprintln!("Type error in generated code: {}", output.stderr);
    } else if output.stderr.contains("not found") {
        eprintln!("Tool not found: {}", output.stderr);
    } else {
        eprintln!("Execution failed: {}", output.stderr);
    }
}
```

### TypeScript Code Requirements

LLM-generated code must follow this pattern:

```typescript
async function run() {
  // Your code that calls registered tools
  const result = await Namespace.toolName({ param: value });

  // MUST return a value
  return result;
}
```

The code execution engine:

- Wraps your code with namespace implementations
- Automatically calls `run()` and captures its return value
- Provides the return value in `ExecuteOutput.output`

## Architecture

1. **Tool Definition**: Tools are defined with JSON Schemas for inputs/outputs
2. **Code Generation**: TypeScript interface definitions are generated from schemas
3. **Code Execution**: User code is wrapped with namespace implementations and executed in Deno
4. **Callback Routing**: Function calls in TypeScript are routed to Rust callbacks or MCP servers
5. **Result Marshaling**: JSON values are passed between TypeScript and Rust

## Sandbox Security

Code is executed in a Deno runtime with:

- Network access restricted to allowed hosts (from registered MCP servers)
- No file system access
- No subprocess spawning
- Isolated V8 context per execution

```rust
// Add servers
code_mode.add_server(&server_config).await?;
```

## Examples

### Multi-Tool Workflow

```rust
let code = r#"
    async function run() {
        // Fetch user data
        const user = await UserApi.getUser({ id: 123 });

        // Process the data
        const processed = await DataProcessor.transform({
            input: user.data,
            format: "normalized"
        });

        // Save results
        const saved = await Storage.save({
            key: `user_${user.id}`,
            value: processed
        });

        return {
            userId: user.id,
            saved: saved.success,
            location: saved.url
        };
    }
"#;

let output = code_mode.execute(code, Some(registry)).await?;
```

### Error Recovery

```rust
let code = r#"
    async function run() {
        try {
            return await RiskyApi.operation({ id: 1 });
        } catch (error) {
            console.error("Operation failed:", error);
            // Fall back to safe default
            return await SafeApi.getDefault();
        }
    }
"#;

let output = code_mode.execute(code, Some(registry)).await?;

// Check console output
if !output.stdout.is_empty() {
    println!("Console output: {}", output.stdout);
}
```

### Parallel Execution

```rust
let code = r#"
    async function run() {
        // Execute multiple operations in parallel
        const [users, posts, comments] = await Promise.all([
            UserApi.listUsers(),
            PostApi.listPosts(),
            CommentApi.listComments()
        ]);

        return { users, posts, comments };
    }
"#;
```

## Related Crates

- `pctx_config`: Server configuration types (`ServerConfig`)
- `pctx_codegen`: TypeScript code generation from JSON schemas
- `pctx_executor`: Deno runtime execution engine
- `pctx_code_execution_runtime`: Runtime environment and callback system

## License

MIT