[[CoreEngineResponsibilities]] competitors: https://joinsidekick.com/ list of automations 3500 n8n 1. https://agentchef.ai/ 2. https://n8n.io/workflows/categories/marketing/ 3. https://github.com/Zie619/n8n-workflows sandbox is warmed up parallely Create Chat , initializes session object and get session id and sandbox url and then generate llm response. (but dont like that it is awaiting to resume or createSandbox before hitting the llm. this can happen before execution) 1. llm_chunk 2. llm_response_complete 3. puts it into the session. (this needs to become a session setter and getter) 4. console_output 1. stdout 2. stderr 5. execution result 6. add it to memory 7. process_result 1. exit or 2. gen plan for loop - goes to step 1 Possible new [[Reactive Streaming Architecture]] Use Cases, how does siva do this Someone receives an email asking to be interviewed. they forward it to siva, who on their behalf, sends the candidate a hackerrank link, and schedules a time, and once the interview is complete, evalues it, and emails the user back with a result. First the trigger with the prompt is via email, a session is started. and it figures out a tasklist , which is persisted and registers triggers(for that email thread or from that user, and to follow up) I guess. The tasklist is obtained from the llm, which says do 1. send email now asking for slots (send email) 2. follow up if no reply is received. (cron job if no movement on state) 3. When slots are received, check in my calendar and book a timeslot, to the calendar slot add the hackerrank (receive email) 4. queries hackerrank periodically to check status of result (cron job to check) 5. when ready leaves a note.