Skip to content

Latest commit

 

History

History
219 lines (171 loc) · 10.1 KB

README.md

File metadata and controls

219 lines (171 loc) · 10.1 KB

funAGI

fundamental AGI point of departure agi.py bdi.py logic.py and SocraticReasoning.py using api.py with memory.py
debug environment includes logging for deeper understanding of the reasoning process and troubleshooting
This "working" version is an exercise in the display of SocraticReasoning technique. This repository has been archived and research work is continuining at

https://github.com/pythaiml/funAGI

FundamentalAGI Project Setup

codebase to create a simplified and fun AGI for fundamentalAGI basic reasoning. funAGI.py uses SocraticReasoning.py with logic.py verbose logging to reason to truth as export of belief into ./memory/truth. This simplified version will omits reasoning.py and SimpleMind.py to directly leverage the core logic and memory management of the basic reasoning AGI. For a deep dive into SocraticReasoning.py visit
https://rage.pythai.net/understanding-socraticreasoning-py/

project structure

fundamentalAGI.py: Main script to run the AGI.
agi.py: Core AGI logic.
api.py: API key management.
bdi.py: BDI (Belief-Desire-Intention) model.
chatter.py: Interface for different chat models.
logic.py: Logic table management and evaluation.
memory.py: Memory management for storing dialogues and truth values.
SocraticReasoning.py: Socratic reasoning engine for the AGI.

agi.py AGI workflow

API Key Management (api.py):

Loads API keys from api_keys.json or environment variables.
Manages adding, removing, and listing API keys.

Core AGI Logic (agi.py):

Initializes the AGI with API manager and chatter models.
Learns from data by creating propositions.
Makes decisions by leveraging SocraticReasoning.

Memory Management (memory.py):

Handles creating necessary directories.
Stores and loads conversation memories.
Stores valid truths and dialogue entries.

Socratic Reasoning (SocraticReasoning.py):

Manages premises and draws logical conclusions.
Logs actions and results.
Validates conclusions using LogicTables.

Logic Table Management (logic.py):

Manages logic variables, expressions, and truth tables.
Evaluates expressions and validates truths.
Logs all actions and saves beliefs.

Chat Models (chatter.py):

Interfaces for different chat models like GPT4o, GroqModel, and OllamaModel.

Key Components and Flow

API Key Management:
    APIManager class handles loading, saving, and managing API keys.
    In EasyAGI.__init__, the user is prompted to manage API keys if necessary.

Learning from Data:
    AGI.learn_from_data method processes input data into two propositions.
    These propositions are added as premises to the SocraticReasoning engine.

Making Decisions:
    AGI.make_decisions method invokes the Socratic reasoning process to draw a conclusion.
    The conclusion (decision) is returned.

Main Loop:
    EasyAGI.main_loop handles user input and processes it using the AGI system.
    Decisions are communicated back to the user and stored in short-term memory.

Belief-Desire-Intention Model (bdi.py):

Manages beliefs, desires, and intentions.
Processes BDI components for decision making.


funAGI.py is the UI. fundatmentalAGI funAGI has been published to showcase, audit and debug the easyAGI SocraticReasoning.py and logic.py

#########################################################################

Process Flow from User Input to Response

Step 1: User Input

Action: The user provides an input via the command line.
Input Example: "Explain the workflow to building AGI"

Step 2: Perceive Environment

Method: EasyAGI.perceive_environment
Action: The system captures the user's input.
Output: The user's input string is returned for processing.

Step 3: Learning from Data

Method: AGI.learn_from_data
Action: The user's input string is processed to extract propositions.
Process:
    The input string is split into two propositions based on a delimiter (e.g., ";").
    If the input contains only one proposition, the second proposition is set as an empty string.
    The propositions are added as premises in the Socratic reasoning engine.
Output: Two propositions are returned for further processing.

Adding Premises to Socratic Reasoning:
    Validates and adds each proposition as a premise.
    Logs each addition.

skipped ::::: Processing THOT:
    Iterates through reasoning methods.
    Generates intermediate conclusions.
    Aggregates combined results.

skipped ::::: Logging THOT Data:
    Logs premises, combined results, and final decision.
    Appends to THOT log file (thots.json).

Drawing a Conclusion:
    Generates logical conclusion using language model.
    Validates conclusion.
    Logs conclusion.

Communicating Response:
    Prints final decision to console.
    Logs communicated response.

Storing Conversation Memory:
    Creates DialogEntry with input and decision.
    Saves entry to STM.

This workflow provides a comprehensive view of how user input is processed, reasoned with, and responded to in fundamentalAGI funAGI.py system. Each component plays a critical role in ensuring the AGI provides accurate and logical responses based on the input provided. funAGI is the simplified agi --> SocraticReasoning --> logic --> response dialogue ################################################################

+-------------------+       +-------------------+       +---------------------+
|    User Input     | ----> | Perceive          | ----> | Learning from Data  |
| (Command Line)    |       | Environment       |       | (Extract Propositions)|
+-------------------+       +-------------------+       +---------------------+
                                      |
                                      v
                             +-------------------+
                             | Add Premises to   |
                             | Socratic Reasoning|
                             +-------------------+
                                      |
                                      v
                             +-------------------+
                             |  Process THOT     |
                             |  (Various Reasoning|
                             |   Techniques)      |
                             +-------------------+
                                      |
                                      v
                             +-------------------+
                             |  Log THOT Data    |
                             +-------------------+
                                      |
                                      v
                             +-------------------+
                             |  Draw Conclusion  |
                             |  (GPT-4 Model)    |
                             +-------------------+
                                      |
                                      v
                             +-------------------+
                             |Communicate Response|
                             |  to User           |
                             +-------------------+
                                      |
                                      v
                             +-------------------+
                             | Store Conversation |
                             | Memory (STM)       |
                             +-------------------+

install

(builds an openai API and groq API ready terminal interaction from openai or groq API key
with response from SocraticReasoning.py and logic.py with logging to local folders from memory.py)
logs SocraticReasoning
./mindx/socraticlogs.txt
logs errors
./mindx/errors/logs.txt
logs logic.py truth tables errors to
./mindx/truth/logs.txt
shows truthtable creation in
./mindx/truth/2024-06-23T22:44:49.670605_belief.txt

to do: include statement of belief in belief log; fix SocraticReasoning from ./mindx/socraticlogs.txt

memory of each input / response is stored in
./memory/stm/1719207889.json
as timestamped instruction / response

{"instruction":"agi","response":"a logical conclusion based on the premise of autonomous general intelligence (agi) is that it has the ability to independently perform a wide range of tasks at a level equal to or beyond that of a human being. agi would be capable of understanding, learning, and applying knowledge across a wide range of domains and tasks, and would be able to adapt to new situations and environments. it would be able to understand and respond to natural language, recognize and interpret visual and auditory information, and make decisions and solve problems based on its understanding. additionally, agi would be capable of self-improvement and continue to learn and get better at tasks over time. however, it is important to note that the development of agi also raises ethical and societal concerns that need to be addressed."}

this is the bare bones working version of funAGI with error logging and first errors to fix from first working deployment. fundamental funAGI.py point of departure version 1. This version is being left as is. Refactoring and upgrade work is happening at https://github.com/pythaiml/funAGI
SocraticReasoning.py overview

git clone https://github.com/autoGLM/funAGI/
cd funAGI
python3 -m venv fun
source fun/bin/activate
pip install -r requirements.txt
python3 funAGI.py

This repo is archived as the first "working" instance of fundamental AGI
work continues at several repos where no single repo is the "working" version:
each component has different "working" parts
lmagi
ezAGI