File size: 29,326 Bytes
183b3b6 | 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 | #!/usr/bin/env python3
"""
Synthetic Tool-Calling Training Data Generator for Stack 2.9
Generates training examples in Qwen2.5-Coder format with tool_calls.
"""
import json
import random
import argparse
from pathlib import Path
from typing import Dict, List, Any
from datetime import datetime
# ============================================================================
# Tool Definitions (Qwen2.5-Coder format)
# ============================================================================
TOOL_DEFINITIONS = [
{
"type": "function",
"function": {
"name": "Bash",
"description": "Execute bash commands in the terminal. Use for running shell commands, scripts, git operations, package managers, and system commands.",
"parameters": {
"type": "object",
"properties": {
"command": {
"type": "string",
"description": "The bash command to execute"
},
"timeout": {
"type": "integer",
"description": "Timeout in seconds (default: 30)"
}
},
"required": ["command"]
}
}
},
{
"type": "function",
"function": {
"name": "FileRead",
"description": "Read the contents of a file from the filesystem. Use for viewing source code, configuration files, documentation, or any text-based files.",
"parameters": {
"type": "object",
"properties": {
"path": {
"type": "string",
"description": "Path to the file to read"
},
"offset": {
"type": "integer",
"description": "Line number to start reading from (1-indexed)"
},
"limit": {
"type": "integer",
"description": "Maximum number of lines to read"
}
},
"required": ["path"]
}
}
},
{
"type": "function",
"function": {
"name": "FileWrite",
"description": "Create or overwrite a file with content. Use for creating new files, updating existing files, or writing code, configuration, or documentation.",
"parameters": {
"type": "object",
"properties": {
"path": {
"type": "string",
"description": "Path where the file should be created or written"
},
"content": {
"type": "string",
"description": "The content to write to the file"
},
"append": {
"type": "boolean",
"description": "Append to existing file instead of overwriting (default: false)"
}
},
"required": ["path", "content"]
}
}
},
{
"type": "function",
"function": {
"name": "WebSearch",
"description": "Search the web for information. Use for finding documentation, looking up error messages, researching libraries, or getting up-to-date information.",
"parameters": {
"type": "object",
"properties": {
"query": {
"type": "string",
"description": "The search query to look up on the web"
},
"count": {
"type": "integer",
"description": "Number of results to return (default: 5)"
}
},
"required": ["query"]
}
}
},
{
"type": "function",
"function": {
"name": "Grep",
"description": "Search for patterns in files. Use for finding specific code, function definitions, imports, TODO comments, error patterns, or any text across the codebase.",
"parameters": {
"type": "object",
"properties": {
"pattern": {
"type": "string",
"description": "The search pattern or regex to match"
},
"path": {
"type": "string",
"description": "Directory or file path to search in (default: current directory)"
},
"recursive": {
"type": "boolean",
"description": "Search recursively in subdirectories (default: true)"
},
"file_pattern": {
"type": "string",
"description": "File pattern to filter results (e.g., '*.py', '*.js')"
}
},
"required": ["pattern"]
}
}
}
]
# ============================================================================
# Template Data for Generation
# ============================================================================
FILE_PATHS = [
"src/main.py", "src/utils.py", "src/config.py", "src/models.py",
"src/api.py", "src/handlers.py", "src/middleware.py",
"tests/test_main.py", "tests/test_utils.py", "tests/conftest.py",
"README.md", "LICENSE", "package.json", "requirements.txt",
"config.yaml", "config.json", ".env.example",
"src/components/Button.tsx", "src/components/Header.jsx",
"src/styles.css", "src/index.js", "src/app.js",
"docs/API.md", "docs/ARCHITECTURE.md", "docs/CONTRIBUTING.md",
"scripts/setup.sh", "scripts/deploy.py", "Makefile"
]
CODE_SNIPPETS = {
"python": [
"def hello():\n print('Hello, World!')",
"class MyClass:\n def __init__(self):\n self.value = 42",
"import os\nos.path.join('a', 'b')",
"async def fetch_data():\n async with aiohttp.ClientSession() as session:\n return await session.get(url)",
],
"javascript": [
"const fetch = require('node-fetch');\nconst data = await fetch(url);",
"function handleClick() {\n setCount(count + 1);\n}",
"export default function App() {\n return <div>Hello</div>;\n}",
"const [state, setState] = useState(null);",
],
"bash": [
"npm install",
"git status",
"pytest -v",
"python -m pytest tests/",
"make build",
"docker build -t myapp .",
"ls -la",
"curl -X GET https://api.example.com",
]
}
WEB_SEARCH_QUERIES = [
"python async await best practices",
"javascript array methods map filter reduce",
"TypeScript generics tutorial",
"React hooks useEffect dependency array",
"Node.js error handling best practices",
"Docker vs Kubernetes differences",
"Git rebase vs merge",
"SQL join types explained",
"REST API design principles",
"Python list comprehension examples",
"JavaScript promise async await",
"CSS flexbox vs grid",
"MongoDB vs PostgreSQL",
"AWS Lambda cold start optimization",
"Docker compose tutorial",
]
GREP_PATTERNS = [
("TODO", "*.py", "Find TODO comments in Python files"),
("function ", "*.js", "Find function definitions in JS"),
("import ", "*.py", "Find imports in Python"),
("class ", "*.ts", "Find class definitions in TypeScript"),
("def ", "*.py", "Find function definitions in Python"),
("const ", "*.js", "Find const declarations in JS"),
("interface ", "*.ts", "Find interface definitions"),
("#pragma", "*.c", "Find pragma directives in C"),
("SELECT ", "*.sql", "Find SQL queries"),
("TODO", "*.md", "Find TODOs in documentation"),
]
# ============================================================================
# Scenario Templates
# ============================================================================
SCENARIOS = {
"FileRead": [
{
"user": "Read the file at src/main.py and show me what's in it",
"tool_call": {"name": "FileRead", "arguments": {"path": "src/main.py"}},
"result": "Successfully read file: src/main.py\n```python\ndef main():\n print('Hello from Stack 2.9!')\n\nif __name__ == '__main__':\n main()\n```",
"response": "Here's the contents of `src/main.py`:\n\n```python\ndef main():\n print('Hello from Stack 2.9!')\n\nif __name__ == '__main__':\n main()\n```\n\nThis is a simple Python script that prints a greeting when executed directly."
},
{
"user": "Can you show me the README.md file?",
"tool_call": {"name": "FileRead", "arguments": {"path": "README.md"}},
"result": "Successfully read file: README.md\n```markdown\n# My Project\n\nA sample project for Stack 2.9.\n\n## Installation\n\npip install -r requirements.txt\n```",
"response": "Here's the README.md:\n\n```markdown\n# My Project\n\nA sample project for Stack 2.9.\n\n## Installation\n\npip install -r requirements.txt\n```"
},
{
"user": "What's in the config.yaml file?",
"tool_call": {"name": "FileRead", "arguments": {"path": "config.yaml"}},
"result": "Successfully read file: config.yaml\n```yaml\ndatabase:\n host: localhost\n port: 5432\n name: myapp\n\napp:\n port: 3000\n debug: true\n```",
"response": "The config.yaml contains:\n\n```yaml\ndatabase:\n host: localhost\n port: 5432\n name: myapp\n\napp:\n port: 3000\n debug: true\n```\n\nIt has database connection settings and app configuration."
},
{
"user": "Check the package.json to see the dependencies",
"tool_call": {"name": "FileRead", "arguments": {"path": "package.json"}},
"result": "Successfully read file: package.json\n```json\n{\n \"name\": \"my-app\",\n \"version\": \"1.0.0\",\n \"dependencies\": {\n \"express\": \"^4.18.0\",\n \"react\": \"^18.2.0\"\n }\n}\n```",
"response": "The package.json shows these dependencies:\n\n```json\n{\n \"name\": \"my-app\",\n \"version\": \"1.0.0\",\n \"dependencies\": {\n \"express\": \"^4.18.0\",\n \"react\": \"^18.2.0\"\n }\n}\n```"
},
],
"FileWrite": [
{
"user": "Create a new file called hello.py with a hello world function",
"tool_call": {"name": "FileWrite", "arguments": {"path": "hello.py", "content": "def hello():\n print('Hello, World!')\n\nif __name__ == '__main__':\n hello()\n"}},
"result": "Successfully wrote file: hello.py",
"response": "Created `hello.py` with a hello world function:\n\n```python\ndef hello():\n print('Hello, World!')\n\nif __name__ == '__main__':\n hello()\n```\n\nYou can run it with `python hello.py`."
},
{
"user": "Write a simple React component to src/components/Button.jsx",
"tool_call": {"name": "FileWrite", "arguments": {"path": "src/components/Button.jsx", "content": "import React from 'react';\n\nexport default function Button({ onClick, children }) {\n return (\n <button onClick={onClick} className=\"btn\">\n {children}\n </button>\n );\n}\n"}},
"result": "Successfully wrote file: src/components/Button.jsx",
"response": "Created `src/components/Button.jsx`:\n\n```jsx\nimport React from 'react';\n\nexport default function Button({ onClick, children }) {\n return (\n <button onClick={onClick} className=\"btn\">\n {children}\n </button>\n );\n}\n```"
},
{
"user": "Create a config.json file with some settings",
"tool_call": {"name": "FileWrite", "arguments": {"path": "config.json", "content": "{\n \"app\": {\n \"name\": \"MyApp\",\n \"version\": \"1.0.0\"\n },\n \"server\": {\n \"port\": 8080,\n \"host\": \"localhost\"\n }\n}\n"}},
"result": "Successfully wrote file: config.json",
"response": "Created `config.json`:\n\n```json\n{\n \"app\": {\n \"name\": \"MyApp\",\n \"version\": \"1.0.0\"\n },\n \"server\": {\n \"port\": 8080,\n \"host\": \"localhost\"\n }\n}\n```"
},
],
"Bash": [
{
"user": "Run npm install to install dependencies",
"tool_call": {"name": "Bash", "arguments": {"command": "npm install"}},
"result": "added 245 packages in 3.5s",
"response": "Dependencies installed successfully! Added 245 packages."
},
{
"user": "Check git status",
"tool_call": {"name": "Bash", "arguments": {"command": "git status"}},
"result": "On branch main\nYour branch is up to date with 'origin/main'.\n\nChanges not staged for commit:\n modified: README.md\n modified: src/main.py\n\nno changes added to commit (but untracked files present)",
"response": "Git status shows:\n- Modified: README.md, src/main.py (not staged)\n- Untracked files present"
},
{
"user": "Run the tests with pytest",
"tool_call": {"name": "Bash", "arguments": {"command": "pytest -v"}},
"result": "tests/test_main.py::test_hello PASSED\ntests/test_utils.py::test_add PASSED\n\n2 passed in 0.5s",
"response": "All tests passed!\n- `tests/test_main.py::test_hello` ✓\n- `tests/test_utils.py::test_add` ✓"
},
{
"user": "List all files in the current directory",
"tool_call": {"name": "Bash", "arguments": {"command": "ls -la"}},
"result": "total 64\ndrwxr-xr-x 12 user staff 384 Apr 6 00:58 .\ndrwxr-xr-x 9 user staff 288 Apr 6 00:50 ..\n-rw-r--r-- 1 user staff 220 Apr 6 00:50 README.md\ndrwxr-xr-x 3 user staff 96 Apr 6 00:51 src\ndrwxr-xr-x 3 user staff 96 Apr 6 00:51 tests\n...",
"response": "Files in current directory:\n- README.md\n- src/\n- tests/\n- And more (permissions, dates shown)"
},
{
"user": "Build the project with make",
"tool_call": {"name": "Bash", "arguments": {"command": "make build"}},
"result": "Building...\ngcc -o main main.c\nobuild successful!",
"response": "Build completed successfully!"
},
],
"WebSearch": [
{
"user": "How do I use async/await in Python?",
"tool_call": {"name": "WebSearch", "arguments": {"query": "python async await best practices", "count": 5}},
"result": "Search results for 'python async await best practices':\n1. Real Python - Async IO in Python: A Complete Walkthrough\n2. Python Docs - asyncio - Asynchronous I/O\n3. Towards Data Science - Mastering Python Async/Await\n4. Stack Overflow - Common async/await mistakes\n5. Medium - Async/Await Patterns in Python",
"response": "Here are some resources on Python async/await:\n\n1. **Real Python** - Async IO in Python: A Complete Walkthrough\n2. **Python Docs** - asyncio - Asynchronous I/O\n3. **Towards Data Science** - Mastering Python Async/Await\n\nKey points:\n- Use `async def` to define async functions\n- Use `await` to call other async functions\n- `asyncio.run()` to execute the main coroutine"
},
{
"user": "What's the difference between Docker and Kubernetes?",
"tool_call": {"name": "WebSearch", "arguments": {"query": "Docker vs Kubernetes differences", "count": 5}},
"result": "Search results for 'Docker vs Kubernetes differences':\n1. Docker Blog - Containers 101\n2. Kubernetes.io - What is Kubernetes\n3. Red Hat - Docker vs Kubernetes explained\n4. DevOps.com - Container orchestration comparison\n5. Container Journal - When to use each",
"response": "**Docker** is a container platform - it packages applications into containers.\n\n**Kubernetes** is a container orchestrator - it manages multiple containers across machines.\n\nThink of Docker as individual shipping containers, and Kubernetes as the port management system that coordinates all of them."
},
{
"user": "How do I use React hooks properly?",
"tool_call": {"name": "WebSearch", "arguments": {"query": "React hooks useEffect dependency array", "count": 5}},
"result": "Search results for 'React hooks useEffect dependency array':\n1. React Docs - Synchronizing with Effects\n2. Kent C. Dodds - A Complete Guide to useEffect\n3. Robin Wieruch - React useEffect Hook Tutorial\n4. Stack Overflow - Common useEffect mistakes\n5. LogRocket - useEffect dependency array deep dive",
"response": "**React Hooks Guide:**\n\n1. **useState** - For state management\n2. **useEffect** - For side effects, takes dependency array\n3. **useContext** - For global state\n\nThe dependency array controls when the effect runs. Empty `[]` = once on mount. Include variables to re-run when they change."
},
],
"Grep": [
{
"user": "Find all TODO comments in the Python files",
"tool_call": {"name": "Grep", "arguments": {"pattern": "TODO", "path": ".", "recursive": True, "file_pattern": "*.py"}},
"result": "Found 3 matches:\nsrc/main.py:14: # TODO: Add error handling\nsrc/utils.py:22: # TODO: Implement caching\ntests/test_main.py:8: # TODO: Add more tests",
"response": "Found 3 TODO comments in Python files:\n- `src/main.py:14` - Add error handling\n- `src/utils.py:22` - Implement caching\n- `tests/test_main.py:8` - Add more tests"
},
{
"user": "Search for function definitions in the JavaScript files",
"tool_call": {"name": "Grep", "arguments": {"pattern": "function ", "path": "src", "recursive": True, "file_pattern": "*.js"}},
"result": "Found 5 matches:\nsrc/index.js:3: function setup() {\nsrc/handlers.js:12: function handleClick(event) {\nsrc/handlers.js:25: function handleSubmit(form) {\nsrc/utils.js:8: function formatDate(date) {\nsrc/utils.js:15: function parseJSON(str) {",
"response": "Found 5 function definitions in src/:\n- `src/index.js:3` - setup()\n- `src/handlers.js:12` - handleClick(event)\n- `src/handlers.js:15` - handleSubmit(form)\n- `src/utils.js:8` - formatDate(date)\n- `src/utils.js:15` - parseJSON(str)"
},
{
"user": "Find all imports in the Python codebase",
"tool_call": {"name": "Grep", "arguments": {"pattern": "import ", "path": ".", "recursive": True, "file_pattern": "*.py"}},
"result": "Found 12 matches:\nsrc/main.py:1: import os\nsrc/main.py:2: import sys\nsrc/main.py:3: import json\nsrc/utils.py:1: from typing import Dict\nsrc/api.py:5: import requests\n...",
"response": "Found 12 import statements across Python files. Main imports:\n- `os`, `sys`, `json` in src/main.py\n- `typing.Dict` in src/utils.py\n- `requests` in src/api.py"
},
{
"user": "Look for class definitions in TypeScript files",
"tool_call": {"name": "Grep", "arguments": {"pattern": "class ", "path": "src", "recursive": True, "file_pattern": "*.ts"}},
"result": "Found 2 matches:\nsrc/models/User.ts:3: class User {\nsrc/models/Config.ts:8: class Config {",
"response": "Found 2 class definitions:\n- `src/models/User.ts:3` - User class\n- `src/models/Config.ts:8` - Config class"
},
],
}
# ============================================================================
# Data Augmentation Helpers
# ============================================================================
def vary_file_paths(template: Dict) -> Dict:
"""Vary file paths in a scenario to create diversity."""
import copy
result = copy.deepcopy(template)
# Vary the file paths
if "path" in result["tool_call"]["arguments"]:
original_path = result["tool_call"]["arguments"]["path"]
for old_path in FILE_PATHS:
if old_path in result["user"]:
result["user"] = result["user"].replace(old_path, random.choice(FILE_PATHS))
result["tool_call"]["arguments"]["path"] = random.choice(FILE_PATHS)
break
return result
def vary_content(template: Dict) -> Dict:
"""Vary content in a scenario."""
import copy
result = copy.deepcopy(template)
if "content" in result["tool_call"]["arguments"]:
# Vary code snippets
lang = random.choice(["python", "javascript"])
result["tool_call"]["arguments"]["content"] = random.choice(CODE_SNIPPETS[lang])
return result
def vary_bash_command(template: Dict) -> Dict:
"""Vary bash commands."""
import copy
result = copy.deepcopy(template)
if "command" in result["tool_call"]["arguments"]:
original = result["tool_call"]["arguments"]["command"].split()[0] if result["tool_call"]["arguments"]["command"] else ""
if "npm" in original:
commands = ["npm install", "npm run build", "npm test", "npm start"]
elif "git" in original:
commands = ["git status", "git log --oneline -5", "git diff", "git branch -a"]
elif "pytest" in original:
commands = ["pytest -v", "pytest tests/", "pytest -xvs", "pytest --cov"]
elif "ls" in original:
commands = ["ls -la", "ls -1", "ls -lah"]
elif "make" in original:
commands = ["make build", "make clean", "make test", "make install"]
else:
commands = ["echo 'hello'", "pwd", "whoami", "date"]
result["tool_call"]["arguments"]["command"] = random.choice(commands)
return result
def vary_search_query(template: Dict) -> Dict:
"""Vary web search queries."""
import copy
result = copy.deepcopy(template)
if "query" in result["tool_call"]["arguments"]:
result["tool_call"]["arguments"]["query"] = random.choice(WEB_SEARCH_QUERIES)
return result
def vary_grep_pattern(template: Dict) -> Dict:
"""Vary grep patterns."""
import copy
result = copy.deepcopy(template)
pattern, file_pattern, _ = random.choice(GREP_PATTERNS)
result["tool_call"]["arguments"]["pattern"] = pattern
result["tool_call"]["arguments"]["file_pattern"] = file_pattern
return result
# ============================================================================
# Main Generation Functions
# ============================================================================
def create_tool_call_message(tool_call: Dict, tool_call_id: str) -> Dict:
"""Create a tool_calls message in Qwen format."""
return {
"role": "assistant",
"content": None,
"tool_calls": [
{
"id": tool_call_id,
"type": "function",
"function": {
"name": tool_call["name"],
"arguments": json.dumps(tool_call["arguments"])
}
}
]
}
def create_tool_message(role: str, tool_call_id: str, tool_name: str, result: str) -> Dict:
"""Create a tool message (result of tool execution)."""
return {
"role": role, # typically "tool"
"content": result,
"tool_call_id": tool_call_id,
"name": tool_name
}
def generate_example(scenario: Dict, system_prompt: str = None) -> Dict:
"""Generate a single training example in Qwen2.5-Coder format."""
if system_prompt is None:
system_prompt = "You are a helpful AI assistant that can use tools to help users solve problems. When you need to perform actions like reading files, running commands, searching the web, or searching code, use the appropriate tool."
tool_call_id = f"call_${random.randint(1000, 9999)}"
messages = [
{"role": "system", "content": system_prompt},
{"role": "user", "content": scenario["user"]},
create_tool_call_message(scenario["tool_call"], tool_call_id),
create_tool_message("tool", tool_call_id, scenario["tool_call"]["name"], scenario["result"]),
{"role": "assistant", "content": scenario["response"]}
]
return {
"messages": messages,
"tools": TOOL_DEFINITIONS
}
def augment_scenario(scenario: Dict, tool_name: str) -> Dict:
"""Apply random augmentations to a scenario."""
import random
augmented = scenario.copy()
if tool_name == "FileRead":
augmented = vary_file_paths(augmented)
elif tool_name == "FileWrite":
augmented = vary_file_paths(augmented)
augmented = vary_content(augmented)
elif tool_name == "Bash":
augmented = vary_bash_command(augmented)
elif tool_name == "WebSearch":
augmented = vary_search_query(augmented)
elif tool_name == "Grep":
augmented = vary_grep_pattern(augmented)
return augmented
def generate_dataset(num_examples: int = 1000, output_path: str = None) -> List[Dict]:
"""Generate the complete dataset."""
examples = []
tools = list(SCENARIOS.keys())
# Track counts for balance
examples_per_tool = num_examples // len(tools)
remainder = num_examples % len(tools)
for i, tool_name in enumerate(tools):
# Determine how many examples for this tool
count = examples_per_tool + (1 if i < remainder else 0)
base_scenarios = SCENARIOS[tool_name]
for j in range(count):
# Use base scenario and vary
base = base_scenarios[j % len(base_scenarios)]
# Apply augmentations for variety
if j >= len(base_scenarios):
scenario = augment_scenario(base, tool_name)
else:
scenario = base
example = generate_example(scenario)
examples.append(example)
# Shuffle for better training
random.shuffle(examples)
return examples
def save_jsonl(examples: List[Dict], output_path: str):
"""Save examples to JSONL format."""
output_file = Path(output_path)
output_file.parent.mkdir(parents=True, exist_ok=True)
with open(output_file, 'w', encoding='utf-8') as f:
for example in examples:
f.write(json.dumps(example, ensure_ascii=False) + '\n')
def save_json(examples: List[Dict], output_path: str):
"""Save examples to JSON format."""
output_file = Path(output_path)
output_file.parent.mkdir(parents=True, exist_ok=True)
with open(output_file, 'w', encoding='utf-8') as f:
json.dump(examples, f, ensure_ascii=False, indent=2)
def main():
parser = argparse.ArgumentParser(description="Generate synthetic tool-calling training data")
parser.add_argument("--num-examples", type=int, default=1000, help="Number of examples to generate")
parser.add_argument("--output-dir", type=str, default="training-data", help="Output directory")
parser.add_argument("--output-format", choices=["jsonl", "json", "both"], default="jsonl", help="Output format")
parser.add_argument("--seed", type=int, default=42, help="Random seed")
args = parser.parse_args()
# Set seed for reproducibility
random.seed(args.seed)
print(f"🎯 Generating {args.num_examples} tool-calling training examples...")
print(f" Output directory: {args.output_dir}")
print(f" Format: {args.output_format}")
print()
# Generate dataset
examples = generate_dataset(args.num_examples)
output_dir = Path(args.output_dir)
# Save based on format
if args.output_format in ["jsonl", "both"]:
jsonl_path = output_dir / "tool_examples.jsonl"
save_jsonl(examples, str(jsonl_path))
print(f"✅ Saved JSONL: {jsonl_path}")
if args.output_format in ["json", "both"]:
json_path = output_dir / "tool_examples.json"
save_json(examples, str(json_path))
print(f"✅ Saved JSON: {json_path}")
# Statistics
print(f"\n📊 Statistics:")
print(f" Total examples: {len(examples)}")
# Count by tool
tool_counts = {}
for ex in examples:
for msg in ex["messages"]:
if msg.get("tool_calls"):
tool_name = msg["tool_calls"][0]["function"]["name"]
tool_counts[tool_name] = tool_counts.get(tool_name, 0) + 1
print(f" Examples by tool:")
for tool, count in sorted(tool_counts.items(), key=lambda x: x[1], reverse=True):
print(f" - {tool}: {count}")
# Show sample
print(f"\n📝 Sample example (first in dataset):")
sample = examples[0]
print(f" Tools defined: {len(sample['tools'])}")
print(f" Messages: {len(sample['messages'])}")
print(f" First user message: {sample['messages'][1]['content'][:60]}...")
print(f"\n✨ Generation complete!")
if __name__ == "__main__":
main()
|