write-passive-swarm-client

Introduction:

Your task is to connect to an external agent and enable dynamic model tuning under the AgentJet reinforcement learning framework, this is very simple.

User --> Application Interface (WEB, TUI, GUI) --> Application Backend --> Fake vLLM (fake_vllm_endpoint.py, you need to write this) --> In fake vLLM, duplicate each request multiple times (on_user_submit_new_requests) --> Calculate relative reward (on_compute_relative_reward) --> Submit reward (swarm_client.end_episode) --> Select the item with highest reward --> Return to user via original path

用户 --> Application界面(WEB, TUI, GUI) --> Application后端 --> 假vLLM（fake_vllm_endpoint.py, 你要写这个） --> 假vLLM中, 将每个请求复制多份（on_user_submit_new_requests） --> 计算相对奖励（on_compute_relative_reward） --> 提交奖励 (swarm_client.end_episode) --> 选取奖励最高的一项 --> 原路返回给用户

First, give the agent system a name based on the user's requirements, always place your code at ``tutorial/opencode_build_***, for example opencode_build_openclaw_agent`.

Next, create the directory: tutorial/opencode_build_openclaw_agent

Then, create the Agent source files: