structai 0.1.2

A utility package for AI development

StructAI

StructAI is a comprehensive utility package for AI development, offering a robust set of tools for file operations, LLM interactions, parallel processing, and general programming tasks.

Installation

Recommended for most users. Installs the latest stable release from PyPI.

pip install structai

For development. Installs StructAI in editable mode from source, enabling live code changes.

git clone https://github.com/black-yt/structai.git
cd structai
pip install -e .

API Reference & Usage

load_file(path)

Automatically reads a file based on its extension.

Supported formats: .json, .jsonl, .csv, .txt, .md, .pkl, .parquet, .xlsx, .py, .npy, .pt, .png, .jpg, .jpeg.

from structai import load_file

# Load a JSON file
data = load_file("config.json")

# Load a CSV file as a pandas DataFrame
df = load_file("data.csv")

# Load an image
image = load_file("photo.jpg")

save_file(data, path)

Automatically saves data to a file based on the extension.

from structai import save_file

data = {"key": "value"}

# Save as JSON
save_file(data, "output.json")

# Save as Pickle
save_file(data, "backup.pkl")

print_once(msg)

Prints a message to stdout only the first time it is called. Useful for logging inside loops.

from structai import print_once

for i in range(10):
    print_once("Starting processing...") # Prints only once
    # process(i)

make_print_once()

Returns a new function that prints a message only once. This allows for creating local "print once" scopes.

from structai import make_print_once

logger1 = make_print_once()
logger2 = make_print_once()

logger1("Hello") # Prints "Hello"
logger1("Hello") # Does nothing

logger2("World") # Prints "World"

LLMAgent

A powerful wrapper class for interacting with OpenAI-compatible LLM APIs. It handles retries, timeouts, and structured output validation.

Initialization:

from structai import LLMAgent

agent = LLMAgent(
    api_key="sk-...",              # Optional if LLM_API_KEY env var is set
    api_base="https://...",        # Optional if LLM_BASE_URL env var is set
    model_version='gpt-4.1-mini',  # Default model
    system_prompt='You are a helpful assistant.',
    temperature=0,
    time_limit=300,                # Timeout in seconds
    max_try=1                      # Number of retries
)

Basic Usage (__call__ or safe_api):

response = agent("What is the capital of France?")
print(response)
# Output: "Paris"

Structured Output Validation:

You can enforce the output format (List, Dict, or specific types) using return_example.

# Enforce a list of integers
numbers = agent(
    "Generate 3 random numbers", 
    return_example=[1], 
    list_len=3
)
# Output: [10, 42, 7]

# Enforce a dictionary with specific keys
profile = agent(
    "Create a user profile for Alice", 
    return_example={"name": "str", "age": 1, "city": "str"}
)
# Output: {'name': 'Alice', 'age': 25, 'city': 'New York'}

Multimodal Input:

# Pass image paths for vision models
description = agent(
    "Describe this image", 
    image_paths=["image.jpg"]
)

sanitize_text(text)

Sanitizes text by keeping only ASCII English characters, digits, and common punctuation. Removes control characters and ANSI codes.

from structai import sanitize_text

clean = sanitize_text("Hello \x1b[31mWorld\x1b[0m!")
print(clean) # "Hello World!"

str2dict(s)

Robustly converts a string representation of a dictionary to a Python dict. It handles common formatting errors and uses json_repair as a fallback.

from structai import str2dict

d = str2dict("{'a': 1, 'b': 2}")
print(d['a']) # 1

str2list(s)

Robustly converts a string representation of a list to a Python list.

from structai import str2list

l = str2list("[1, 2, 3]")
print(len(l)) # 3

add_no_proxy_if_private(url)

Checks if the hostname in the URL is a private IP address. If so, it adds it to the no_proxy environment variable to bypass proxies.

from structai import add_no_proxy_if_private

add_no_proxy_if_private("http://192.168.1.100:8080/v1")

read_image(image_path)

Reads an image from a path and returns a PIL Image object.

from structai import read_image

img = read_image("photo.jpg")

encode_image(image_obj)

Encodes a PIL Image object into a base64 string.

from structai import encode_image

b64_str = encode_image(img)

messages_to_responses_input(messages)

Converts standard Chat Completions messages format (list of dicts) to the input format required by the Responses API.

from structai import messages_to_responses_input

messages = [{"role": "user", "content": "Hello"}]
system_prompt, input_blocks = messages_to_responses_input(messages)

extract_text_outputs(result)

Extracts the text content from an LLM API response object (supports both Chat Completions and Responses API formats).

from structai import extract_text_outputs

# Assuming 'response' is the object returned by the OpenAI client
texts = extract_text_outputs(response)
print(texts[0])

multi_thread(inp_list, function, max_workers=40, use_tqdm=True)

Executes a function concurrently for each item in inp_list using a thread pool.

from structai import multi_thread
import time

def square(x):
    return x * x

inputs = [{"x": i} for i in range(10)]
results = multi_thread(inputs, square, max_workers=4)
print(results) # [0, 1, 4, 9, ...]

multi_process(inp_list, function, max_workers=40, use_tqdm=True)

Executes a function concurrently for each item in inp_list using a process pool. Ideal for CPU-bound tasks.

from structai import multi_process

def heavy_computation(n):
    return sum(range(n))

inputs = [{"n": 1000000} for _ in range(5)]
results = multi_process(inputs, heavy_computation)

run_server(host="0.0.0.0", port=8001)

Starts a FastAPI server that acts as a proxy to an OpenAI-compatible LLM provider.

from structai import run_server

if __name__ == "__main__":
    run_server()

timeout_limit(timeout=None)

A decorator that enforces a maximum execution time on a function. Raises TimeoutError if the limit is exceeded.

from structai import timeout_limit
import time

@timeout_limit(timeout=2.0)
def task():
    time.sleep(5)

# This will raise TimeoutError
task()

run_with_timeout(func, args=(), kwargs=None, timeout=None)

Runs a function with a specified timeout without using a decorator.

from structai import run_with_timeout

def task(x):
    return x * 2

result = run_with_timeout(task, args=(10,), timeout=1.0)

parse_think_answer(text)

Parses a string containing Chain-of-Thought tags (<think>...</think> and <answer>...</answer>) and returns the content of both.

from structai import parse_think_answer

raw_text = "<think>Step 1...</think><answer>42</answer>"
think, answer = parse_think_answer(raw_text)
print(f"Reasoning: {think}")
print(f"Result: {answer}")

extract_within_tags(content, start_tag='<answer>', end_tag='</answer>', default_return=None)

Extracts the substring found between two specific tags.

from structai import extract_within_tags

text = "Result: <json>{...}</json>"
json_str = extract_within_tags(text, "<json>", "</json>")

get_all_file_paths(directory, suffix='')

Recursively retrieves all file paths in a directory that match a given suffix.

from structai import get_all_file_paths

# Get all Python files in the current directory
py_files = get_all_file_paths(".", suffix=".py")
print(py_files)

remove_tag(s, tags=["<think>", "</think>", "<answer>", "</answer>"], r="\n")

Removes specified tags from a string, replacing them with a separator (default newline).

from structai import remove_tag

clean_text = remove_tag("<think>...</think> Answer")

License

MIT License