Local Extractive Summarizer

Relevant source files

The following files were used as context for generating this wiki page:

• bot.py
• README.md
• requirements.txt
• .env.example
• Dockerfile

The Local Extractive Summarizer is a dependency-light component that generates short summaries of Telegram chat messages. It extracts the first few sentences from provided text or historical chat logs without requiring external AI API keys.

This module operates within the bot.py script and supports both direct command-line input and historical context analysis. It captures incoming text messages from Telegram groups or private chats and maintains a rolling window of recent messages for summary generation.

Sources: README.md:14-16, bot.py:1-5

System Architecture and Logic

The summarizer uses a rule-based approach to process text. It splits input into sentences based on punctuation markers and rejoins a configurable number of starting sentences to form the summary.

Message Processing Flow

The system processes messages in two ways:

1. Collection: It stores incoming non-command text in a deque for each chat.
2. Extraction: It triggers the summarization logic when a user sends the /summary command.

The diagram shows the decision logic for handling incoming updates and routing them to the summarizer.
Sources: bot.py:108-132

Summarization Algorithm

The summarize_text function performs the following steps:

1. Cleans whitespace from the input string.
2. Replaces "!" and "?" with "." to standardize sentence boundaries.
3. Splits the text by "." into a list of strings.
4. Filters out empty strings and joins the first SUMMARY_SENTENCES together.

Sources: bot.py:79-96

Configuration Parameters

The behavior of the summarizer is controlled by environment variables loaded via python-dotenv.

Variable	Default	Description
MAX_STORED_MESSAGES	100	Maximum number of recent messages retained per chat in the chat_messages buffer.
SUMMARY_SENTENCES	5	The number of sentences extracted from the beginning of the text to create the summary.
LOG_LEVEL	INFO	The verbosity of the system logs.

Sources: bot.py:24-26, README.md:43-47

Data Structures

The system uses Python's collections module to manage chat history in-memory.

chat_messages: dict[int, deque[str]] = defaultdict(lambda: deque(maxlen=MAX_STORED_MESSAGES))

Sources: bot.py:34-35

• defaultdict: Automatically creates a new entry when a unique chat_id is encountered.
• deque: A thread-safe double-ended queue with a maxlen attribute that automatically discards the oldest messages when the limit is reached.

Command Interface

The summarizer interacts with users through specific command patterns.

Command	Action
/summary <text>	Processes the explicitly provided <text> immediately.
/summary	Concatenates all messages stored in chat_messages[chat_id] and summarizes them.

Sources: bot.py:122-129, README.md:10-12

Component Sequence

The following sequence diagram illustrates the interaction between the Telegram API, the bot, and the local summarization logic.

The diagram illustrates the loop of receiving updates and the conditional execution of summarization.
Sources: bot.py:79-142

Summary

The Local Extractive Summarizer provides a lightweight, privacy-focused method for condensing chat history. By relying on basic string manipulation rather than external LLM services, it ensures high performance and zero operational costs for simple text extraction tasks.

Sources: README.md:14-16, bot.py:79-96