Genesis 5.5

A lightweight JSON conversational AI model. Works offline. No GPU required.

Overview

Genesis 5.5 maps user inputs to handcrafted responses. It does not use neural networks or inference engines. Just load the JSON and start chatting.

Available Variants

File Description Size
Genesis-5.5.json Standard model โ€” direct response lookup ~55 MB
Genesis-5.5-Thinking.json Model with <|think|> reasoning blocks in every response ~115 MB
  • Model: Genesis 5.5
  • Format: JSON (key to response)
  • Entries: 525,786 per variant
  • Language: English
  • License: MIT

Features

  • Instant responses. No computation. Just a lookup.
  • Fully offline. No internet or API calls needed.
  • Zero dependencies. Just a JSON parser and string matching.
  • Massive coverage. 525K+ entries spanning greetings, farewells, emotional support, philosophy, science topics, capabilities, and more.
  • Reasoning simulation (Thinking variant). Each response includes a <|think|> internal monologue. Clients can show this as chain-of-thought.

Categories

The model covers 525,786 conversation patterns across 20+ categories.

Category Approx. Count Examples
Emotional Expression ~290,000 i feel sad, im happy, i am tired, i feel lonely
Identity & Nature ~35,000 are you a bot, who are you, are you sentient, what is your name
Philosophy & Beliefs ~10,000 meaning of life, do you believe in god, what is consciousness, aliens
Greetings ~8,000 hello, hi, hey, good morning, howdy, aloha, ayy
Check-ins ~4,000 how are you, hows it going, what's up
Farewells ~1,200 bye, goodbye, see you, catch you later, peace out, adios
Capabilities ~1,000 can you help me, can you code, can you tell a joke, can you translate
Trivia & Facts ~350 did you know, tell me a fact, fun fact
Recommendations ~130 can you recommend a book, can you recommend a movie
Affirmations & Thanks ~70 thanks, you're awesome, i love you, best bot
Discourse & Reactions ~60 ok, cool, nice, wow, got it, agreed, fair enough
Creator Info ~50 who made you, what is xpdevs, what is doorsos
Jokes & Riddles ~30 tell me a joke, crack me up, riddle
Advice & Tips ~15 advice for today, benefits of, best way to
Preferences ~12 coffee or tea, cats or dogs, beach or mountains
Identity ~9 who are you, introduce yourself, what can you do
Binary & Tech ~8 binary for hello, 01001000
Other (topic-based) ~175,000 creative writing, cooking tips, science explainers, history, space, nature, technology, health, etc.

Reasoning Tags (Thinking Variant)

Every response in the Thinking variant includes a <|think|> tag with simulated reasoning.

{
  "who are you": "<|think|>User is asking about my identity. I should be clear and friendly โ€” state my name, who built me, and what I do.</|think|>I'm Genesis 5.5 โ€” a conversational model built by XPDevs. I'm here to chat, help, and keep you company!"
}

Clients can parse these tags. The thinking content can be shown before the clean response.

Usage (Python)

import json, re

with open("5.5/Genesis-5.5.json") as f:
    model = json.load(f)

def respond(input_text, model):
    clean = input_text.lower().strip()
    if clean in model:
        raw = model[clean]
        clean = re.sub(r"<\|think\|>.*?</\|think\|>", "", raw).strip()
        return clean
    return "I'm not sure how to respond to that. Could you rephrase?"

print(respond("hello", model))

File Structure

5.5/
  Genesis-5.5.json             # Standard model data (~55 MB)
  Genesis-5.5-Thinking.json    # Model with reasoning blocks (~115 MB)
  README.md                    # This file

The JSON is a flat object.

{
  "ver": "Genesis 5.5",
  "hello": "Hey, good to see you! What's good? ๐Ÿ˜„",
  "bye": "Goodbye! Talk to you later โ€” Genesis signing off. โœŒ๏ธ",
  ...
}

Limitations

  • No generative capabilities. Responses are pre-written. The model cannot answer questions outside its dataset.
  • No memory. Each response is stateless. Conversation history is not considered.
  • English only. Designed for English conversation patterns.

License

MIT. Free to use, modify, and distribute.


Built by XPDevs.

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support