Your first conversation¶
Run the node¶
From source:
export TELEGRAM_BOT_TOKEN=...
./bin/ganglion run
In a container, the image already runs ganglion run --db /data/ganglion.db.
The node prints log lines as it starts. Keep it running. Only one process may poll a bot token at a time, so stop any other program that uses the same bot.
Say hello¶
Open a private chat with your bot in Telegram and write a message. Only the admin you named in telegram.admin gets answers. Everyone else gets no reply.
There is no /start or /help command. Just write.
The introduction¶
On a fresh node the agent has no persona yet. It starts a short introduction: it asks your name, how you want it to talk, and in which language. Then it proposes a persona and waits for you to approve it with a button. The persona is saved as plain Markdown files in the workspace/persona/ folder, which you can edit later.
Ask for something simple¶
Try a question or a small task:
What is the weather forecast for Lisbon tomorrow?
The bot shows a reaction emoji while it works, then answers. A web search needs no button. See Tools and safety for what needs a press.
Send /status to see how the node is doing. See Commands.
Ask for a small automation¶
Describe a routine task in your own words:
Every 30 minutes read the temperature from https://example.com/api/temp
and tell me if it drops below 5 degrees.
The agent picks one of two templates (a reader for the outside service, a reflex that reacts to the values), writes the code, builds it in the background and shows you a live card.
When the build is done, you get an install window. It lists, in plain words:
- the network hosts the module may reach,
- the secrets it needs,
- the cells it creates and reads,
- when it runs,
- how many messages a day it may send you.
Press Approve to install or Deny to drop it. The question folds to a line such as "Approved, 21:09" once you decide. If you do nothing it expires.
After the install, /modules lists the module and /cells shows its values. More in Modules.