DevMind V1 - Walkthrough

DevMind is a local, privacy-first developer memory engine. It ingests your activity from Terminal, Browser, and VSCode, embeds it using a local model, and stores it in a vector database for semantic search.

Features Implemented

Data Sources

• Terminal History: Parses ~/.zsh_history (Zsh).
• Browser History: Reads Chrome history from ~/.config/google-chrome/Default/History.
• VSCode Activity: Reads recent files/folders from ~/.config/Code/User/globalStorage/state.vscdb.

Vector Pipeline

• Embeddings: Uses fastembed-rs with all-MiniLM-L6-v2 (runs locally, no API keys).
• Vector Store: Uses LanceDB (embedded, serverless) to store and search vectors.

CLI Interface

• dev-mind ingest: Runs the ingestion pipeline.
• dev-mind query "<question>": Searches your developer memory.

Usage

1. Ingest Data

cargo run -- ingest

This will:

1. Connect to data sources.
2. Extract recent activity.
3. Generate embeddings.
4. Upsert into LanceDB (~/.local/share/dev-mind/lancedb).

2. Query

cargo run -- query "docker commands"

Returns top 5 relevant records.

Architecture

• Language: Rust
• Database: LanceDB
• Model: all-MiniLM-L6-v2 (ONNX)
• Extensibility: Uses DataSource trait for easy addition of new sources (Slack, Claude, etc.).

Verification

• Build successful (cargo check).
• Ingestion run successful.
• Query run successful.