v2.9.0 — First public release

Oblivion v2.9.0 — First public release 🚀

Oblivion is now a real distributable Python package. Install with pip, talk to it with voice, use it from Claude Desktop via MCP.

✨ Highlights

• 🎙️ Voice in & out — press Ctrl+T, talk to Meera, hear her reply (Whisper + Edge TTS / ElevenLabs)
• 🤖 9 LLM backends — Gemini, Groq, Claude, GPT-4o, DeepSeek, Ollama, Cerebras
• ⚡ Auto-fallback — silent model swap when one rate-limits
• 🧠 Hybrid code search — symbol lookup + full-text + semantic embeddings
• 🔌 MCP server — Claude Desktop can use Oblivion's 10 read-only tools
• 📚 Knowledge packs — React, Django, FastAPI, Tailwind, Docker, security, more
• 📜 Session history — Ctrl+G to resume any past conversation
• 🌃 Cyberpunk TUI — Textual-based with animated boot sequence

📦 Install

```bash
pip install oblivion-agent
```

With voice support:
```bash
pip install "oblivion-agent[voice]"
```

🚀 First Run

```bash
oblivion
```

Setup wizard walks you through picking an LLM provider (Gemini recommended — free + 1500 req/day).

🔌 Claude Desktop Integration

Add to claude_desktop_config.json:
```json
{
"mcpServers": {
"oblivion": {
"command": "oblivion",
"args": ["mcp"],
"env": { "WORKSPACE_DIR": "/your/project" }
}
}
}
```

📋 What's New Since v2.0

• Pip-installable package structure (~/.oblivion config dir)
• Auto-fallback chain for LLMs
• MCP server with 10 read-only code tools
• Session history picker (Ctrl+G)
• Subcommand CLI (oblivion mcp, oblivion init)
• Dynamic /model autocomplete (all 9 models)
• Voice transcription bias for proper name recognition

Built with ❤️ by R. Rohit.