A Home Assistant custom integration for Groq Cloud with memory, web search, and link ingestion.
--- ## Features - Multiple instances (different models, personalities, and settings) - Persistent memory (device/user/global scopes) - Tool calling via Home Assistant LLM API - Optional web search via SearxNG - Auto-fetch and ingest URLs from user messages - Rich generation controls (temperature, top_p, penalties, max tokens, response format) ## Install (HACS) 1. Add this repo as a **Custom Repository** in HACS (type: Integration): `https://github.com/sudoxreboot/groqd` 2. Install `groqd`. 3. Restart Home Assistant. 4. Add the integration from **Settings → Devices & Services**. ## Quick Start - Create a new groqd entry. - Set your Groq API key. - Pick a model (default: `meta-llama/llama-4-maverick-17b-128e-instruct`). - Set a personality prompt if you want tone/style changes. ## Options Open **groqd → Options** to configure: - Memory scope: `device`, `user`, `global`, or `conversation` - Context window size - Max tokens, temperature, top_p, penalties - Response format (`text` / `json_object`) - SearxNG (enable + URL + language + safe search) - Auto-fetch URLs in user messages - Time format (12h / 24h) ## Recommended Prompt Additions To make tone warmer and lyrical: """ Be warm, playful, and a bit lyrical. If the user quotes lyrics, continue one line then respond empathetically. """ ## Notes - TTS/STT are handled by Home Assistant and other integrations. - Web search is optional and only used if enabled. - Auto-fetch pulls page text and injects it into the prompt for better grounding. ## Support Open issues or requests at: https://github.com/sudoxreboot/groqd/issues