Flashback

Flashback is a command-line knowledge store. It captures text, URLs, and commands, extracts structured metadata, and makes everything searchable.

It is designed for developers who want a fast, local, scriptable memory system.

---

Features

• CLI-first workflow (Cobra)
• Metadata extraction for URLs (OpenGraph, Twitter, JSON-LD)
• AI enrichment using Google Gemini (strict JSON schema)
• Local-first storage backed by Turso/libSQL
• TUI viewer built with Bubbletea
• Fast search with filters and tags
• Separate metadata table for incremental enrichment

---

Usage

flashback help

TUI:

flashback

Add entries:

flashback add kubectl rollout restart deployment web
flashback add https://blog.bytebytego.com/p/understanding-load-balancers

Search:

flashback search load balancer
flashback search kubernetes

View entries:

flashback list
flashback show <id>

---

How it works

Flashback processes inputs through a simple pipeline:

1. Detect type (text, URL, code)
2. Scrape metadata (OpenGraph, JSON-LD, fallbacks)
3. Run Gemini enrichment (tags, summary, normalization)
4. Store in SQLite/Turso with structured metadata

All AI output is enforced via JSON schema to ensure deterministic results.

---

Data model

Each record is stored in the flashbacks table. Metadata is stored separately in the metadata table as key/value pairs. This allows multiple enrichment passes and avoids schema churn.

Example metadata fields:

tldr
description
tags (JSON array)
image (url)

---

Configuration

Location: ~/.config/flashback/config.yaml

---

Install

Using Homebrew

brew install yagnikpt/tap/flashback

From source:

git clone https://github.com/yagnikpt/flashback
cd flashback
make install

Direct Go Install

go install github.com/yagnikpt/flashback@latest

---

Roadmap

• More site-specific extractors (GitHub, YouTube, Reddit, Medium)
• Configurable metadata schema
• Plugin system for custom enrichers