gi-text-summarizer 1.0.0

Lightweight text summarizer with token budgets and smart chunking

gi_text_summarizer

A lightweight, production-ready Python library for summarizing text with any OpenAI-compatible LLM.

Features

✅ Automatic text chunking for long documents ✅ Token budget guardrails (summaries always shorter than input) ✅ Support for multiple summary lengths, tones, output formats ✅ Works with OpenAI, Azure OpenAI, or custom LLM endpoints ✅ Built-in token estimation (no external dependencies) ✅ Production-grade error handling and logging

Installation

pip install gi-text-summarizer

Quick Start

from gi_text_summarizer import TextSummarizer

summarizer = TextSummarizer(api_key="sk-...")

result = summarizer.summarize(
    text="Your long document here...",
    summary_type="medium",    # "short" | "medium" | "detailed"
    tone="neutral",           # "neutral" | "formal" | "casual"
    focus_area="general",
    output_format="text",     # "text" | "bullets" | "json"
)

print(result.summary)
print(result.compression)  # e.g., "68% shorter"
print(result)              # Pretty-printed with all details

API Reference

TextSummarizer

TextSummarizer(
    api_key=None,              # OpenAI/Azure API key (or env vars)
    azure_endpoint=None,       # Azure endpoint
    deployment_name=None,      # Azure deployment
    model="gpt-4o-mini",       # Model name
    provider="auto",           # "openai" | "azure" | "custom"
)

summarize()

result = summarizer.summarize(
    text: str,
    summary_type="medium",     # "short" (20%) | "medium" (40%) | "detailed" (60%)
    tone="neutral",            # "neutral" | "formal" | "casual"
    focus_area="general",      # Any string
    output_format="text",      # "text" | "bullets" | "json"
    chunk_strategy="character",# "character" | "sentence"
    chunk_size=3000,           # Characters per chunk
) -> SummaryResult

SummaryResult

result.summary          # str: Generated summary
result.input_tokens     # int: Tokens in original
result.output_tokens    # int: Tokens in summary
result.compression      # str: "X% shorter"
result.num_chunks       # int: Number of chunks

Token Budget Guardrails

Automatically calculates token caps to ensure summaries are shorter:

Type	Ratio	Use Case
short	20%	Core idea (3-4 sentences)
medium	40%	Executive summary
detailed	60%	Comprehensive summary

Chunking & Multi-Document Summarization

For longer texts:

1. Text is split into chunks
2. Each chunk is summarized
3. Summaries are combined and re-summarized

result = summarizer.summarize(
    text=long_document,
    chunk_strategy="sentence",  # Better quality
    chunk_size=3000,
)

Environment Variables

# OpenAI
export OPENAI_API_KEY="sk-..."

# Azure OpenAI
export AZURE_OPENAI_API_KEY="..."
export AZURE_OPENAI_ENDPOINT="https://your-resource.openai.azure.com"
export AZURE_OPENAI_DEPLOYMENT="gpt-4o-mini"

Standalone Token Counter

from gi_text_summarizer import count_tokens

tokens = count_tokens("Your text here...")

Examples

Executive Summary (Formal)

result = summarizer.summarize(
    text=financial_report,
    summary_type="medium",
    tone="formal",
    focus_area="financial",
    output_format="bullets",
)

Technical Highlight (JSON)

result = summarizer.summarize(
    text=documentation,
    summary_type="detailed",
    tone="neutral",
    focus_area="technical",
    output_format="json",
)

Ultra-Concise (20% of input)

result = summarizer.summarize(
    text=research_paper,
    summary_type="short",
)

Publishing to PyPI

pip install build twine
python -m build
twine upload dist/*

License

MIT License.