Lost in translation - Fix or buy new?
I need some help. My Clawdbot runs on a VPS and connects mainly via Telegram. It access my Mac via Tailscale and SSH.
Right now, the system is unstable. Logs show repeated unclean gateway restarts (“shutdown timed out”) and session “role ordering conflicts” that force the agent session to restart. That leads to inconsistent behavior, apparent forgetting, and difficulty completing long multi-step tasks reliably. Tooling is also incomplete/misconfigured and sometimes simply disappear: some commands/scripts fail (eg. missing rg, Serper script permission errors), and Google access is in transition (a service account with domain-wide delegation exists, but token/library errors have appeared out of the blue).
Net effect: the Clawdbot can seem forgetful, get stuck mid-task, and behave unpredictably until we harden restarts/session handling, clean up dependencies, and tighten guardrails.
To its credit, I have really played around with the install, tools and multiple versions so I might have really screw up things along the way.
What would be the best option now, to get it right once and for all? Is it still fixable from your point of view or the VPS is now doomed with zombie services, broken paths that interferes and instructions all over the place?
I know EVERYTHING is fixable, but sometimes, it’s better to simply let that boat sink and get another one.
Right now, the system is unstable. Logs show repeated unclean gateway restarts (“shutdown timed out”) and session “role ordering conflicts” that force the agent session to restart. That leads to inconsistent behavior, apparent forgetting, and difficulty completing long multi-step tasks reliably. Tooling is also incomplete/misconfigured and sometimes simply disappear: some commands/scripts fail (eg. missing rg, Serper script permission errors), and Google access is in transition (a service account with domain-wide delegation exists, but token/library errors have appeared out of the blue).
Net effect: the Clawdbot can seem forgetful, get stuck mid-task, and behave unpredictably until we harden restarts/session handling, clean up dependencies, and tighten guardrails.
To its credit, I have really played around with the install, tools and multiple versions so I might have really screw up things along the way.
What would be the best option now, to get it right once and for all? Is it still fixable from your point of view or the VPS is now doomed with zombie services, broken paths that interferes and instructions all over the place?
I know EVERYTHING is fixable, but sometimes, it’s better to simply let that boat sink and get another one.
