Most chatbots forget everything the moment the window scrolls, so users repeat themselves constantly. This layer gives your Telegram bot durable memory by embedding each message and storing it in Supabase with pgvector.
Most chatbots forget everything the moment the window scrolls, so users repeat themselves constantly. This layer gives your Telegram bot durable memory by embedding each message and storing it in Supabase with pgvector.
When someone asks a new question, it pulls the most relevant past messages and feeds them back as context. The bot recalls what you told it last week, and the memory keeps growing with every chat instead of resetting.
How it runs
Used at step 01 to kick off the pipeline.
Write
Used at step 01 to kick off the pipeline.
WebFetch
Used at step 01 to kick off the pipeline.
WebSearch
Used at step 01 to kick off the pipeline.
On a new question, retrieve the most relevant past messages by similarity search.
pending
Feed the retrieved memories plus the recent turns into the model as context.
pending
Save the new exchange so the memory grows with every conversation.
pending
Sample output
json
// Sample output
// (generated when the pipeline finishes)
Answer using the retrieved memories below; if a memory conflicts with the new message, trust the newest one.
Unlock the rest
The full agent definition, install snippet, and starter task are gated for community members.
Members get the full `.md` agent file, the npm / pnpm install one-liners, a starter prompt that we've tuned against real runs, and the open-source repo when this automation ships there. One email, magic link, done.