LLM Unify

A local-first, privacy-focused tool for managing conversations across multiple LLM platforms

What Is This?

LLM Unify consolidates your AI conversation history from multiple platforms into a single, searchable, offline database. Your chats with ChatGPT, Claude, Gemini, and Copilot - all accessible through one interface, stored locally on your machine, under your control.

# Import your ChatGPT export
llm-unify import chatgpt ~/Downloads/conversations.json

# Search across all providers
llm-unify search "async rust tokio"

# Browse in terminal UI
llm-unify tui

Why?

The Problem: Your AI conversations are scattered across platforms, locked in proprietary formats, subject to deletion, and not easily searchable.

The Solution: A unified local archive that:

• Preserves your conversations permanently on your machine

• Unifies multiple providers into one searchable database

• Protects your privacy - no data ever leaves your computer

• Empowers you to search, export, and analyze your AI interactions

Features

Implemented (v0.1.0)

Feature	Description
Multi-Platform Import	ChatGPT, Claude, Gemini, and GitHub Copilot - all four major providers
Full-Text Search	SQLite FTS5-powered search with snippet highlighting
Terminal UI	Ratatui-based browser with vim bindings (j/k, / search)
14 CLI Commands	import, list, show, search, delete, export, stats, validate, backup, restore, init, schema, tui, version
Local-First Storage	SQLite database with WAL mode and zero network dependencies
Type-Safe Rust	Idiomatic Rust with zero unsafe blocks
Schema Versioning	Automatic migrations with full version history
Backup Integrity	SHA-256 checksums with atomic operations and validation
Connection Pooling	Efficient SQLite access with up to 5 concurrent connections

Coming Soon (v0.2.0)

• Database encryption (SQLCipher)

• Export file encryption (age/GPG)

• Enhanced TUI with message viewing

• 80%+ test coverage

See ROADMAP.adoc for the full development plan.

Installation

From Source (Recommended)

# Prerequisites: Rust 1.75+, SQLite 3.35+
git clone https://github.com/Hyperpolymath/llm-unify.git
cd llm-unify
cargo install --path crates/llm-unify-cli

From crates.io (Coming Soon)

cargo install llm-unify-cli

Quick Start

1. Initialize Database

llm-unify init
# Creates ~/.local/share/llm-unify/conversations.db

2. Import Conversations

# ChatGPT (export from Settings → Data Controls → Export)
llm-unify import chatgpt ./conversations.json

# Claude (export from Anthropic Console → Settings → Export)
llm-unify import claude ./claude-export.json

# Gemini (use browser extensions like Gem Chat Exporter)
llm-unify import gemini ./gemini-export.json

# GitHub Copilot (VS Code: Chat: Export Chat... command)
llm-unify import copilot ./copilot-chat.json

3. Search Your History

# Search all conversations
llm-unify search "machine learning"

# Limit results
llm-unify search "rust async" --limit 5

# List all conversations
llm-unify list

# Filter by provider
llm-unify list --provider chatgpt

4. Launch Terminal UI

llm-unify tui

Table 1. TUI Keybindings

Key	Action
j / ↓	Move down
k / ↑	Move up
/	Search mode
Enter	Select conversation
q	Quit

5. Backup Your Data

# Create backup
llm-unify backup ~/llm-unify-backup.db

# Restore from backup
llm-unify restore ~/llm-unify-backup.db

Architecture

LLM Unify is structured as a Rust workspace with 6 focused crates:

llm-unify/
├── crates/
│   ├── llm-unify-core/     # Domain models, traits, errors (179 LOC)
│   ├── llm-unify-storage/  # SQLite persistence layer (322 LOC)
│   ├── llm-unify-parser/   # Provider import parsers (246 LOC)
│   ├── llm-unify-search/   # Full-text search engine (130 LOC)
│   ├── llm-unify-cli/      # Command-line interface (261 LOC)
│   └── llm-unify-tui/      # Terminal UI (216 LOC)
├── docs/                    # Additional documentation
├── .well-known/            # RFC 9116 compliant metadata
└── ...

Core Types

// Provider enumeration
pub enum Provider {
    ChatGpt,
    Claude,
    Gemini,
    Copilot,
    Other(String),
}

// Conversation structure
pub struct Conversation {
    pub id: String,
    pub title: String,
    pub provider: Provider,
    pub messages: Vec<Message>,
    pub created_at: DateTime<Utc>,
    pub updated_at: DateTime<Utc>,
}

// Message structure
pub struct Message {
    pub id: String,
    pub role: MessageRole,  // User, Assistant, System
    pub content: String,
    pub timestamp: DateTime<Utc>,
    pub metadata: Metadata,
}

Database Schema

-- Conversations table
CREATE TABLE conversations (
    id TEXT PRIMARY KEY,
    title TEXT NOT NULL,
    provider TEXT NOT NULL,
    created_at TEXT NOT NULL,
    updated_at TEXT NOT NULL
);

-- Messages table with FTS5 index
CREATE TABLE messages (
    id TEXT PRIMARY KEY,
    conversation_id TEXT NOT NULL,
    role TEXT NOT NULL,
    content TEXT NOT NULL,
    timestamp TEXT NOT NULL,
    metadata TEXT,
    FOREIGN KEY (conversation_id) REFERENCES conversations(id)
);

-- Full-text search
CREATE VIRTUAL TABLE messages_fts USING fts5(
    content,
    content=messages,
    content_rowid=rowid
);

Commands Reference

Command	Description
init	Initialize the database
import <provider> <file>	Import conversations from a provider export
list [--provider <name>]	List all conversations, optionally filtered
show <id>	Display a specific conversation
search <query> [--limit <N>]	Full-text search across messages
delete <id>	Remove a conversation
export <id> [--output <path>]	Export conversation to JSON
stats	Show database statistics
validate	Check database integrity (SQLite + data consistency)
backup <output>	Create database backup with SHA-256 checksum
restore <input>	Restore from backup with integrity verification
schema	Display schema version and migration history
tui	Launch terminal UI
version	Show version information (schema, backup, export formats)

Security

Guarantees

• Zero unsafe code - Memory safety verified by Rust compiler

• Parameterized queries - SQL injection prevention via SQLx

• Backup integrity - SHA-256 checksums with pre-restore validation

• Atomic operations - WAL mode prevents corruption during writes

• No telemetry - Zero data collection or phone-home behavior

• Local-only - All data stays on your machine

• Open source - Fully auditable codebase

Known Limitations (v0.1.x)

• Database not encrypted at rest (use disk encryption)

• Export files are plaintext (encryption coming v0.2)

• Relies on filesystem permissions for access control

See SECURITY.md for vulnerability reporting.

Philosophy

We Believe In

• User autonomy over platform lock-in

• Privacy by default, not by policy

• Open source as a public good

• Local-first data ownership

• Accessibility for all users

We Reject

• Surveillance capitalism

• Dark patterns and manipulation

• Gatekeeping and elitism

• Vendor lock-in

Compliance & Standards

Standard	Status
RSR (Rhodium Standard Repository)	🥈 Silver (51/55 points)
TPCF (Trusted Perimeter Classification)	Perimeter 3 (Community Sandbox)
RFC 9116	Compliant (security.txt)
REUSE	SPDX license headers
Keep a Changelog	v1.0.0 format
Semantic Versioning	v2.0.0 spec

Development

Prerequisites

• Rust 1.75+ (latest stable recommended)

• SQLite 3.35+

• Just command runner

• Git 2.30+

Building

# Clone repository
git clone https://github.com/Hyperpolymath/llm-unify.git
cd llm-unify

# Build all crates
just build

# Run tests
just test

# Run lints
just lint

# Full CI simulation
just ci

# See all recipes
just --list

Contributing

We welcome contributions! See CONTRIBUTING.md for:

• Development workflow

• Coding standards

• PR process

• Path to maintainership

Documentation

Document	Purpose
ROADMAP.adoc	Development roadmap and feature backlog
CHANGELOG.md	Release history
CONTRIBUTING.md	How to contribute
SECURITY.md	Security policy and reporting
CODE_OF_CONDUCT.md	Community standards
MAINTAINERS.md	Governance model
RSR-COMPLIANCE.md	Rhodium Standard compliance report
docs/CITATIONS.adoc	Citation formats for academic use

License

This project is dual-licensed:

• AGPL-3.0-or-later for the codebase - Strong copyleft ensuring user freedom

• Palimpsest Protocol for data portability - Your data, your rights

See LICENSE and LICENSE-PALIMPSEST for details.

Community

• TPCF Perimeter: 3 (Community Sandbox)

• Code of Conduct: Contributor Covenant 2.1 + CCCP extensions

• Governance: Consensus-seeking with voting fallback

Acknowledgments

See .well-known/humans.txt for credits and attribution.

---

No AI training on user data. See .well-known/ai.txt for our AI policy.