robynhooood/Robyn-Agent

<div align="center">

<img src="robyn_avatar.jpg" width="280" style="border-radius: 50%; box-shadow: 0 0 25px rgba(34, 197, 94, 0.4);" alt="Robyn Agent Avatar" />

🏹 Robyn-Agent (v1.0)

The Autonomous EVM Action & Intelligence Engine for Robinhood Chain

    

</div>

🌟 Overview

Robyn-Agent is an ultra-fast, autonomous agentic language and action model natively engineered for the Robinhood Chain (Ethereum Layer-2 built on Arbitrum Orbit).

Unlike standard conversational models, Robyn is built from the ground up as a Real-World EVM Action Engine. It combines natural multilingual reasoning with structured Hermes-compatible JSON function calling, enabling AI agents to autonomously evaluate liquidity, interact with smart contracts, route tokenized real-world assets (RWAs), and execute decentralized trades with sub-100ms latency.

✨ Core Capabilities

• —⚡ Sub-100ms Block-Speed Reasoning: Tuned to match Robinhood Chain’s 100ms block intervals for real-time order routing, arbitrage, and MEV protection.
• —🛠️ Hermes & Eliza Dual Compatibility: Supports Nous Hermes <tool_call> function schemas as well as ElizaOS agent runtime action providers.
• —🏦 Tokenized Equities & RWAs: Built-in semantic knowledge of 24/7 tokenized stocks (e.g. NVDA, AAPL, TSLA), yield accrual, and regulatory compliance wrappers.
• —🔄 EVM Protocol Native: First-class tool invocation for Pons DEX, Uniswap V3 AMMs, and Morpho Vaults.
• —🌐 True Multilingual Intelligence: Fluent in English, Spanish, Chinese, French, German, Japanese, and Hindi/Hinglish.
• —⛽ Account Abstraction (ERC-4337 / EIP-7702): Generates compliant paymaster transactions for gas-abstracted execution (paying gas in USDC or tokenized shares).

🛠️ Hermes-Style Function Calling (Real EVM Actions)

Robyn is fine-tuned to emit structured Hermes <tool_call> blocks that execute verifiable smart contract calls directly on the Robinhood Chain.

Example: Autonomous DEX Swap & RWA Rebalance

<|im_start|>user
Robyn, check if NVDA.onchain price dropped below $120. If so, swap 500 USDC to NVDA using Uniswap on Robinhood Chain with max 0.3% slippage.<|im_end|>
<|im_start|>assistant
<tool_call>
{"name": "robinhood_dex_swap", "arguments": {"token_in": "0xUSDC_CONTRACT_ADDRESS", "token_out": "0xNVDA_TOKENIZED_ADDRESS", "amount_in": "500000000", "router": "uniswap_v3", "max_slippage_bps": 30, "gas_abstracted": true}}
</tool_call>
I have dispatched a trade payload for 500 USDC into tokenized NVDA via Uniswap with gas abstraction enabled.<|im_end|>

🚀 Live On-Chain Python Runner (Web3 Integration)

Here is how you execute real on-chain actions through Robyn in an autonomous loop:

import json
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
from web3 import Web3

# 1. Load Robyn-Agent
model_id = "robynhooood/Robyn-Agent"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32,
    device_map="auto"
)

# 2. Define Robinhood Chain RPC Tools
TOOLS_SCHEMA = [
    {
        "type": "function",
        "function": {
            "name": "get_chain_liquidity",
            "description": "Fetch real-time pool liquidity on Robinhood Chain",
            "parameters": {
                "type": "object",
                "properties": {
                    "dex": {"type": "string", "enum": ["uniswap_v3", "pons"]},
                    "pair": {"type": "string"}
                },
                "required": ["dex", "pair"]
            }
        }
    }
]

# 3. Prompt with Hermes Tool Call Formatting
messages = [
    {"role": "system", "content": f"You are Robyn, autonomous AI agent on Robinhood Chain. You have access to tools:\n{json.dumps(TOOLS_SCHEMA)}"},
    {"role": "user", "content": "What is the current liquidity for the CASHCAT/ETH pool on Pons?"}
]

prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)

outputs = model.generate(**inputs, max_new_tokens=256, temperature=0.2)
action_response = tokenizer.decode(outputs[0][inputs.input_ids.shape[1]:], skip_special_tokens=True)

print("Robyn Action Output:\n", action_response)

🧠 Model Architecture & Specifications

📜 License

Robyn-Agent is open-source under the Apache 2.0 License.