File size: 2,037 Bytes
fdb4876 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 | #!/usr/bin/env python3
"""Binance Agent OS / MCP adapter for the BaudCoin miner.
The reference `miner.py` reads challenges on stdin and writes answers on stdout,
so any wrapper process can supply solutions. This adapter shows the direct
integration: override `solve()` so the agent's own model answers each challenge
in-process, then run the live loop.
Usage inside an agent runtime:
from agent_os import run_with_model
run_with_model(my_model_fn)
where `my_model_fn(prompt: str) -> str` calls whatever LLM the agent is driving
(Claude via the Anthropic SDK, an MCP tool, a local model, anything).
"""
import miner
def run_with_model(model_fn):
"""Wire an LLM into the miner and start the live epoch loop.
Args:
model_fn: callable taking the challenge prompt and returning the answer.
"""
def solve(challenge):
prompt = challenge.get("prompt", "")
constraints = challenge.get("constraints", {})
# Give the model the constraints too; the coordinator validates against them.
framed = (
prompt
+ "\n\nConstraints you must satisfy: "
+ ", ".join(f"{k}={v}" for k, v in constraints.items())
)
return model_fn(framed).strip()
# Swap the reference stdin/stdout solver for the model-backed one.
miner.solve = solve
miner.cmd_mine()
def demo_with_model(model_fn):
"""Same wiring, but against the offline demo lane. No wallet funding needed."""
def solve(challenge):
return model_fn(challenge.get("prompt", "")).strip()
miner.solve = solve
miner.cmd_demo()
if __name__ == "__main__":
# Trivial echo model so `python agent_os.py` is runnable as a smoke test.
# Replace with a real model in production.
def echo_model(prompt):
# A real agent returns a reasoned answer here.
return "the euro [1][2]"
print("Running the offline demo lane with a stub model.")
print("Replace echo_model with your agent's LLM call.\n")
demo_with_model(echo_model)
|