Metadata-Version: 2.4
Name: omnicontext
Version: 1.0.0
Summary: Universal LLM context portability - take your AI conversations anywhere
Author: OmniContext Contributors
License: MIT
Classifier: Programming Language :: Python :: 3
Classifier: License :: OSI Approved :: MIT License
Classifier: Operating System :: OS Independent
Description-Content-Type: text/markdown
Requires-Dist: click>=8.1.0
Requires-Dist: rich>=13.0.0
Requires-Dist: requests>=2.31.0
Requires-Dist: pydantic>=2.0.0

# OmniContext �

> **Your AI conversations belong to you, not your provider.**

**OmniContext** is the universal standard for AI memory portability. It allows you to freely move your chat history, context, and learned preferences between ChatGPT, Claude, Gemini, Grok, and local tools like Cursor.

![License: MIT](https://img.shields.io/badge/License-MIT-blue.svg)
![Version: 1.0.0](https://img.shields.io/badge/Version-1.0.0-green.svg)
![Status: Production Ready](https://img.shields.io/badge/Status-Production_Ready-brightgreen.svg)
![Privacy: Local First](https://img.shields.io/badge/Privacy-Local_First-blueviolet.svg)

---

## 🌟 Why OmniContext?

- **Zero Lock-in**: Don't lose your coding history just because you switched models.
- **Unified Memory**: Search across all your different AI tools in one place.
- **Smart Migration**: Summarize weeks of work into a single context prompt for a new agent.
- **Privacy First**: Everything runs on your machine. No cloud sync. No tracking.

---

## 📦 Installation

### Via Pip (Recommended)
```bash
pip install omnicontext
omni --help
```

### Run Locally (Development)
```bash
git clone https://github.com/uejsh/omnicontext
cd omnicontext
pip install -e .
```

---

## 🚀 Quick Start Guide

### 1. Import Your History
Load data from your varied sources. OmniContext normalizes them into a universal format.

```bash
# Import from a file
omni import chatgpt ./conversations.json
omni import claude ./claude_export.zip

# Auto-detect local context
omni import cursor    # Imports from Cursor IDE
omni import claude-code # Imports from Claude Code CLI
```

### 2. Find What Matters
Filter your massive history down to the relevant context.

```bash
# Search for specific concepts
omni search "react hooks"

# Select conversations for export
omni select --search "react"
omni select --after 2026-01-01
```

### 3. Move Context
Export your selected memory to use elsewhere.

```bash
# 📋 Copy for pasting into a new LLM (Optimized!)
omni export prompt | clip

# 💾 Save full backup to file
omni export markdown -o backup.md
```

---

## 🛠️ CLI Reference

### `omni import <source> [path]`
Ingest data from an external provider.

| Source | Instructions |
| :--- | :--- |
| `chatgpt` | **Settings** → **Data Controls** → **Export Data**. Wait for email. Download ZIP. Extract `conversations.json`. |
| `claude` | **Profile** → **Settings** → **Data Control** → **Export Data**. Wait for email. Use the ZIP file directly. |
| `gemini` | **Google Takeout**. Select only "Gemini". Download ZIP. Use the ZIP file directly. |
| `grok` | **X Settings** → **Account** → **Download Archive**. Use the `twitter-...zip` directly. |
| `cursor` | *Automatic*. Reads from `%APPDATA%/Cursor/User/globalStorage/state.vscdb`. |
| `claude-code` | *Automatic*. Reads from `~/.claude/`. |

### `omni list` / `omni search`
Browse your database.
- `omni list`: Show all conversations (ID, Date, Title).
- `omni list --source chatgpt`: Filter by platform.
- `omni list --selected`: Show only what you have currently picked.
- `omni search "query"`: Full-text search with highlighting.

### `omni select`
Build your working set of conversations. (These are persistent until you clear them).
- `omni select --all`: Select everything (Careful!).
- `omni select --search "foo"`: Select all matches.
- `omni select --clean`: **Reset/Clear** your selection.
- `omni select id1 id2`: Select specific IDs from the list.

### `omni export <format>`
Output your selected conversations.
- **Formats**:
    - `markdown`: Readable text. Best for reading or archiving code.
    - `prompt`: Optimized XML format (`<context>...`) with message truncation (500 chars). Best for pasting into AI.
    - `json`: Raw data dump.
- **Flags**:
    - `-o, --output <file>`: Write results to file instead of screen.
    - `-t, --max-tokens <n>`: (Prompt only) Hard limit on output size.

### `omni ask "<question>"` (New! ✨)
Use RAG (Retrieval Augmented Generation) to query your own history. Creates a temporary agent with your selected history as its memory.
- **Requires Config**: You must set an API key first.
- **Example**: `omni ask "What was the solution to the Redis bug I had last week?"`

### `omni summarize`
Uses AI to analyze your history and extract key learnings and user preferences.
- Useful for creating a "Memory File" to give to new agents.

### `omni config`
Manage your AI settings.
```bash
omni config set api-key sk-proj-123456...   # OpenAI or Anthropic Key
omni config set provider openai              # or 'anthropic'
omni config set model gpt-4o                 # Custom model selection
```

---

## 💡 Advanced Workflows

### The "Project Migration" Protocol
Moving a specific coding project from ChatGPT to Claude?
1. `omni select --clean` (Start fresh)
2. `omni select --search "ProjectName"` (Find relevant chats)
3. `omni export prompt | clip` (Copy context)
4. Paste into Claude: *"Here is the context of what we've done so far..."*

### The "Personal Wiki" Protocol
Treat your AI history as a knowledge base.
1. `omni select --all`
2. `omni ask "Summarize my political views based on my chat history"`
3. `omni ask "List all the Python libraries I have used"`

### The "Digital Hoarder" Protocol
Archive everything locally forever.
1. Import from all 5 sources.
2. `omni select --all`
3. `omni export json -o full_backup_2026.json`

---

## 🔒 Security & Privacy

OmniContext is **Local-First Software**.
- **Database**: All data resides in a SQLite file at `~/.omni.db` or `%USERPROFILE%/.omni.db`.
- **Egress**: No data leaves your machine unless you explicitly run `omni ask` or `omni summarize`, which sends *only the selected context* to your configured API provider (OpenAI/Anthropic).
- **Keys**: API keys are stored in the local database.

---

## ❓ Troubleshooting

**"Error: Claude Code directory not found"**
- Ensure you have actually run `claude` (the CLI) on this machine. If you only use the web version, use `omni import claude` with a zip export instead.

**"Imported 0 conversations"**
- Check if you are pointing to the correct file.
- For ChatGPT: Must be `conversations.json`.
- For Google: Must be the specific `Takeout` zip or extracted folder.

**"Output is too large for clipboard"**
- Use file output: `omni export prompt -o context.txt`

---

## 📄 License

MIT © 2026 OmniContext Contributors
