python-echo 0.1.0

Echo is a lightweight and intuitive Python library for building and managing LLM-based chat conversations.

Echo

Echo is a lightweight and intuitive Python library for building and managing LLM-based chat conversations
It provides structured tools for managing chat, including FIFO queue management,
conversation topic evaluation, pertinence filtering, archiving, and more

Features

• Capacity: Limit the number of messages stored in the conversation (FIFO)
  
• Topic evaluation: Evaluate the topic of the conversation and update it dynamically
  
• Pertinence filtering: Filter out irrelevant messages based on the current topic
  
• Archive messages: Archive the original message, keep a summarized version in the chat
  

Algorithms

Topic Evaluation

• Init Message Topic (IMT) - The conversation topic is inferred solely from the content of the first message
  
• Nth Message Topic (NMT) - The conversation topic is inferred from the content of the first N messages
  
• Anchored Topic Evaluation (ATE) - The conversation topic is initially anchored using the first N messages
  Afterward, the topic is dynamically re-evaluated with each new message,
  allowing the model to adapt to topic shifts while maintaining stability around the original anchor
  A new message may or may not affect the topic,
  depending on whether it introduces a significant shift in content relative to the anchor
  

gemini.py

import google.generativeai as genai
from synapse import Synapse

class GeminiClient(Synapse):
    def __init__(self, model_name: str = "gemini-2.5-flash-lite"):
        genai.configure(api_key='<API Key>')
        self.model = genai.GenerativeModel(model_name)

    def prompt(self, text: str) -> str:
        """ Sends a single string to the model and returns the text response """
        try:
            response = self.model.generate_content(text)
            return response.text
        except Exception as e:
            return f"An error occurred: {e}"

main.py

# main.py

client = GeminiClient()

rec = Record(
    synapse=client,
    topic_evaluation_algorithm='SATE2'
)

rec.add(
    Message(
        text="Hello World",
        author='John Doe'
    )
)

Examples

Archive Example

[Message]
"""
The capital city of England is London.
London is a large, historic, and highly influential city located in the southeast of England, along the River Thames.
It has been a major settlement for nearly 2,000 years, originally founded by the Romans as Londinium.
Key characteristics:
- Political center: Home to the UK government, Parliament, and the Prime Minister’s residence at 10 Downing Street.
- Cultural hub: Famous landmarks include Big Ben, Buckingham Palace, Tower Bridge, and the British Museum.
- Economic powerhouse: One of the world’s leading financial centers, especially around the City of London and Canary Wharf.
- Diverse and global: Known for its multicultural population, languages, food, and arts.
- Transport and infrastructure: Extensive public transport, including the London Underground, one of the oldest metro systems in the world.
- London combines deep historical roots with modern global influence, making it one of the most important cities in the world.
"""

[Distilled Message]
"Description of London, England. (Archived 9e63c59908b84572bb6356af3db6a015)"

Pertinence Filter and Topic Evaluation Examples

1. Topic Evaluation Algorithm: Non-Strict Init Message Topic (NSIMT)

[Messages]
"The sky is blue."
"The sun is yellow."
"My name is John Doe."
"I am 21 years old."

[Evaluated Topic]
"Observations of the world"

[Included Messages]
"The sky is blue."
"The sun is yellow."

[Excluded Messages]
"My name is John Doe."
"I am 21 years old."

2. Topic Evaluation Algorithm: Strict Init Message Topic (SIMT)

[Messages]
"The sky is blue."
"The sun is yellow."
"My name is John Doe."
"I am 21 years old."

[Evaluated Topic]
"The Sky's Color"

[Included Messages]
"The sky is blue."
"The sun is yellow."

[Excluded Messages]
"My name is John Doe."
"I am 21 years old."

3. Topic Evaluation Algorithm: Non-Strict Nth Message Topic - First 2 Messages (NSNMT2)

[Messages]
"The sky is blue."
"The sun is yellow."
"My name is John Doe."
"I am 21 years old."

[Evaluated Topic]
"Observations about the environment"

[Included Messages]
"The sky is blue."
"The sun is yellow."

[Excluded Messages]
"My name is John Doe."
"I am 21 years old."

4. Topic Evaluation Algorithm: Strict Nth Message Topic - First 2 Messages (SNMT2)

[Messages]
"The sky is blue."
"The sun is yellow."
"My name is John Doe."
"I am 21 years old."

[Evaluated Topic]
"Sky and Sun Colors"

[Included Messages]
"The sky is blue."
"The sun is yellow."

[Excluded Messages]
"My name is John Doe."
"I am 21 years old."

5. Topic Evaluation Algorithm: Non-Strict Anchored Topic Evaluation - First 2 Messages (NSATE2)

[Messages]
"The sky is blue."
"The sun is yellow."
"My name is John Doe."
"I am 21 years old."

[Evaluated Topic]
"Observations about the natural world"

6. Topic Evaluation Algorithm: Strict Anchored Topic Evaluation - First 2 Messages (SATE2)

[Messages]
"The sky is blue."
"The sun is yellow."
"My name is John Doe."
"I am 21 years old."

[Evaluated Topic]
Sky and Sun Colors