Summary of "Мой AI-агент теперь помнит, что делал вчера"

What the System Demonstrates

The video shows an AI-agent workflow that remembers and structures daily project development using Obsidian notes and storage. This ensures that each new session starts with the right context—such as improvements, errors, and next steps—without repeating past mistakes.

What the System Looks Like (UI / Structure)

A Docs folder containing project components.
A Daily folder containing daily notes (a log of what the agent did each day).
In the UI graph (shown in red), hovering reveals which components the agent worked on that day.

When you open a daily note, it contains a structured entry including:

Improvements
Errors
Enough context to understand development over time

A components list (right side) includes:

Component descriptions
Finalized dates
Short notes about what the agent did when working on each component

A calendar view helps you jump to a specific date and see what the agent did that day.

Why This Is Needed (Problem Analysis)

When an agent starts a new day/session, it may have no working memory of prior actions.

Daily notes act as the persistent “source of truth” by:

preventing the agent from repeating the same fixes/mistakes (“same rake”)
helping the user track progress and know what to do next

Setup Guide (Step-by-Step)

Create an Obsidian storage for the project:
- Name it however you want (typically matching the project name).
- Remove the default/welcome note.
Create core folders inside the storage:
- Daily/
- Docs/
Create a rules / codex note so the agent knows how to write daily notes:
- The exact name varies by tool, for example:
  - Obsidian “code” note / “codex” note (for code workflows)
  - Alternatives like “Cloud MD / AGN agent…” depending on whether you’re using Cloud Code/OpenCloud/etc.
Open the terminal and launch the agent configured to the current storage folder.
Provide a prompt/query instructing the agent, such as:
- develop an HTML + CSS calculator in a specific style (e.g., Iron Man/Marvel)
- specify what:
  - Docs is for (components)
  - Daily is for (errors/improvements log)
  - the agent should fill/update the .m / .md rules file so it can consistently follow daily-note rules each session
The agent can:
- proceed using the folders you already created, or
- create folders on request (the video notes that pre-creating reduces friction)

Obsidian Plugin + Daily Notes Formatting

Install and enable the Obsidian plugin “Calendar” (highest downloads).
In Obsidian settings, configure Daily Notes to store entries in the Daily folder.
To match the calendar layout shown in the demo:
- reposition the plugin/tabs (drag/move the calendar display up/down)
- links are often moved to the bottom border

Daily Note Template Generated by the Agent

During setup, the agent writes:

rules for working with daily notes
a template / recommended note structure
categories such as:
- purpose of the session
- what was done
- mistakes/problems
- solutions
- next steps

The template is editable to fit personal workflows—for example, adding a “training” category for a sports project.

Demonstrated Result

After the prompt, the agent reports “Done.”
It updates the rules file (MD code) and:
- creates the first daily note
- records:
  - the current session
  - changes made
  - next steps
The video concludes that the agent’s behavior can be customized by editing the rules and adding categories.

Main Speakers / Sources

Speaker / author (primary): the presenter demonstrating the Obsidian + AI agent workflow (speaks throughout; no named second person).
Main software sources referenced:
- Obsidian (storage, settings, Daily Notes)
- Calendar plugin for Obsidian
- an AI agent launched via terminal that uses a rules/codex note