feat(llm_judge): add BlockRun as LLM provider for x402 payments

[1bcMax]

Summary

Add BlockRun AI Gateway as a new LLM provider option alongside OpenAI, Anthropic, and local providers.

What is BlockRun?

BlockRun provides:

• Access to 30+ AI models (GPT-4o, Claude, Gemini, DeepSeek, Llama, etc.)
• Pay-per-use via USDC micropayments on Base chain
• No API keys needed - your wallet private key is your identity
• Automatic x402 payment flow via blockrun-llm SDK

Usage

from agentarc import LLMJudge

# Set BLOCKRUN_WALLET_KEY env var or pass private_key directly
judge = LLMJudge(
    provider="blockrun",
    model="gpt-4o",  # or claude-sonnet, gemini-pro, etc.
    private_key="0x..."  # or set BLOCKRUN_WALLET_KEY env var
)

analysis = judge.analyze(
    transaction=tx,
    parsed_tx=parsed,
    simulation_result=simulation
)

Installation

pip install agentarc[blockrun]

Available Models

Provider	Models
OpenAI	gpt-4o, gpt-4o-mini, o1, o1-mini
Anthropic	claude-sonnet, claude-haiku
Google	gemini-pro, gemini-flash
DeepSeek	deepseek, deepseek-reasoner
Meta	llama

Changes

• Added blockrun provider option in LLMJudge.__init__
• Added private_key parameter for BlockRun authentication
• Added _analyze_with_blockrun() method
• Added BLOCKRUN_MODELS dict for model alias resolution
• Added blockrun optional dependency in pyproject.toml

Test plan

• Install with pip install agentarc[blockrun]
• Set BLOCKRUN_WALLET_KEY environment variable
• Initialize LLMJudge(provider="blockrun", model="gpt-4o")
• Run analyze() on a test transaction
• Verify x402 payment flow works correctly

Learn more: https://blockrun.ai