FoundationAgents/MetaGPT · error · RuntimeError

Environment has not been reset yet

Error message

Environment has not been reset yet

What it means

MinecraftExtEnv._step raises RuntimeError('Environment has not been reset yet') when the has_reset flag is False. The Mineflayer bot and the bridge server are only brought up by reset(), so stepping before the first reset has no bot to execute code against; the flag makes this an explicit failure instead of a cryptic HTTP error.

Source

Thrown at metagpt/environment/minecraft/minecraft_ext_env.py:165

            "position": options.get("position", None),
        }

        self.unpause()
        self.mineflayer.stop()
        time.sleep(1)  # wait for mineflayer to exit

        returned_data = self.check_process()
        self.has_reset = True
        self.connected = True
        # All the reset in step will be soft
        self.reset_options["reset"] = "soft"
        self.pause()
        return json.loads(returned_data)

    @mark_as_writeable
    def _step(self, code: str, programs: str = "") -> dict:
        if not self.has_reset:
            raise RuntimeError("Environment has not been reset yet")
        self.check_process()
        self.unpause()
        data = {
            "code": code,
            "programs": programs,
        }
        res = requests.post(f"{self.server}/step", json=data, timeout=self.request_timeout)
        if res.status_code != 200:
            raise RuntimeError("Failed to step Minecraft server")
        returned_data = res.json()
        self.pause()
        return json.loads(returned_data)

    @mark_as_writeable
    def pause(self) -> bool:
        if self.mineflayer.is_running and not self.server_paused:
            res = requests.post(f"{self.server}/pause")
            if res.status_code == 200:

View on GitHub (pinned to 11cdf466d0)

Solutions

1. Call await env.reset(...) once after construction and before the first step; reset sets has_reset=True and starts the bot.
2. If reset() itself failed, fix that first — check the mineflayer process is running (check_process) and the server URL/port are correct.
3. In multi-episode loops, call reset at the start of each episode since has_reset persists per instance state.

Example fix

# before
env = MinecraftExtEnv(server="http://127.0.0.1", mc_port=3000, replay_dir="...")
obs = await env.step("bot.chat('hi')")  # RuntimeError: Environment has not been reset yet

# after
env = MinecraftExtEnv(server="http://127.0.0.1", mc_port=3000, replay_dir="...")
await env.reset()
obs = await env.step("bot.chat('hi')")

if not getattr(env, "has_reset", False):
    await env.reset()  # mandatory before the first step
obs = await env.step(code)

try:
    obs = await env.step(code)
except RuntimeError as e:
    if "not been reset" in str(e):
        await env.reset()
        obs = await env.step(code)  # retry once after establishing the bot
    else:
        raise

When it happens

Trigger: Creating MinecraftExtEnv(server=..., mc_port=...) and calling await env.step(...) / env.write_to_api('step', ...) before calling await env.reset(...); also in a fresh process resuming a session without re-resetting.

Common situations: Adapting gym-style examples that step immediately after construction; reset() failed earlier (server startup issues) and the exception was swallowed so the code continued to step; multiple episodes where reset happens in a try block that was skipped.

Related errors

• Failed to step Minecraft server
• device-id: {device_id} not found
• create device path: {folder_path} failed
• api_name: {api_name} not found
• {rw_api} not exists

AI-assisted analysis of FoundationAgents/MetaGPT@11cdf466d0 (2026-08-14). Data as JSON: /api/errors/d6e6c822660136d8. Report an issue: GitHub.