ikkinchi

your second brain, a zero-friction CLI for capturing and retrieving thoughts.

Status: In Development

ikkinchi means "second" in Uyghur

Thoughts are stored as plain markdown files in ~/.ikkinchi/memories/.

Search is fuzzy by default. Add --semantic for Ollama-powered semantic search. No folders, no cloud.

Install

cargo install ikkinchi

Setup

ikkinchi uses Ollama for semantic search — no API key, no data leaving your machine. Fuzzy search works without Ollama.

Embeddings are powered by Rig, a Rust library for building LLM-powered applications.

# 1. Install Ollama
brew install ollama        # macOS
# or https://ollama.com/download

# 2. Pull the embedding model
ollama pull nomic-embed-text

# 3. Start Ollama (runs in background)
ollama serve

# 4. Initialize ikkinchi
ikkinchi init

How it works

Capture thoughts instantly:

ikkinchi add "what do fish think about all day?"
ikkinchi add "learned how the borrow checker works" --tag rust --tag til
ikkinchi add "There is a @swc/react-compiler package but no documentation"

Each thought is appended to a daily markdown file (~/.ikkinchi/memories/2026-03-10.md). In the background, Rig calls Ollama to embed the text into a 768-dimension vector, which is persisted in a local SQLite store (~/.ikkinchi/vectors.db).

Search your memories:

# Fuzzy search (default — works without Ollama)
$ ikkinchi search "work"
  1  2026-03-10/19:54:08  There is a @swc/react-compiler package but no documentation
  ...

# Semantic search (requires Ollama running)
$ ikkinchi search --semantic "borrowing in rust"
  1  2026-03-10/19:54:08  learned how the borrow checker works
  ...

Fuzzy search uses SkimMatcherV2 to rank results by text similarity — fast, offline, no setup. Semantic search embeds your query with Ollama and ranks by cosine similarity against stored vectors — understands meaning, not just keywords.

Browse and manage:

ikkinchi tui               # open interactive terminal UI
ikkinchi list              # newest first, with tags shown
ikkinchi list --tag rust   # filter by tag
ikkinchi edit <id> <text>  # update a thought (re-embeds automatically)
ikkinchi delete <id>       # remove a thought
ikkinchi stats             # memory count, date range, vector count, tag count

Manage tags:

ikkinchi tags                          # list all tags with counts
ikkinchi tag add <id> rust til         # add tags to a memory
ikkinchi tag remove <id> til           # remove a tag

Rebuild the vector index:

ikkinchi reindex           # re-embed all memories (useful after import or model change)

Import existing notes:

ikkinchi import notes.md          # one chunk per paragraph
ikkinchi import ~/notes/          # all .md and .txt files in a directory

Storage

Everything lives in ~/.ikkinchi/:

~/.ikkinchi/
  config.toml          # model, embedding settings
  memories/
    2026-03-10.md      # plain markdown, one file per day
    2026-03-11.md
  vectors.db           # SQLite, local embedding vectors (Rig + nomic-embed-text)

You can read, edit, or back up your memories directly — they're just markdown.

Under the hood

• Rig — Rust LLM framework, handles all Ollama communication and embedding calls
• Ollama — runs nomic-embed-text locally, no internet required
• SQLite (via sqlx) — persists embedding vectors as raw binary blobs
• fuzzy-matcher — SkimMatcherV2 for fast offline fuzzy text search
• Plain markdown — memories are human-readable files, not a proprietary database

Commands

ikkinchi init                          Set up ~/.ikkinchi/, create config
ikkinchi add <text> [--tag <tag>]...   Capture a thought, optionally tagged
ikkinchi search <query> [--tag <tag>] [--semantic]  Search memories (fuzzy by default)
ikkinchi tui                           Launch interactive terminal UI
ikkinchi list [--count N] [--tag <tag>] Browse recent memories
ikkinchi edit <id> <text>              Update a memory
ikkinchi delete <id>...                Delete one or more memories
ikkinchi tags                          List all tags with counts
ikkinchi tag add <id> <tag>...         Add tags to a memory
ikkinchi tag remove <id> <tag>...      Remove tags from a memory
ikkinchi import <path>                 Import .md/.txt files
ikkinchi export [--format json|markdown]  Export all memories
ikkinchi reindex                       Rebuild vector index from markdown files
ikkinchi stats                         Show memory, vector, and tag counts

Run ikkinchi --help for details on any command.

Configuration

~/.ikkinchi/config.toml (created by ikkinchi init):

[embedding]
provider = "ollama"
model = "nomic-embed-text"
url = "http://localhost:11434"
ndims = 768

[display]
list_count = 20