README
¶
chat — Basic RAG chat (CLI)
Interactive RAG chat with in-memory storage. Loads a set of documents, then accepts questions via stdin and answers with inline citations [1], [2].
The store lives in memory — data is not persisted between runs.
Quick start
EMBEDDER_API_KEY=sk-... \
LLM_API_KEY=sk-... \
go run ./examples/chat/
Environment variables
| Variable | Default | Description |
|---|---|---|
EMBEDDER_API_KEY |
— | Required. API key for the embedder |
EMBEDDER_BASE_URL |
https://api.openai.com |
Embedder API base URL |
EMBEDDER_MODEL |
text-embedding-ada-002 |
Embedding model |
LLM_API_KEY |
— | Required. API key for the LLM |
LLM_BASE_URL |
https://api.openai.com |
LLM API base URL |
LLM_MODEL |
gpt-4o-mini |
Language model |
Local mode (Ollama)
# Start Ollama and pull the required models:
ollama pull nomic-embed-text
ollama pull llama3.2
EMBEDDER_BASE_URL=http://localhost:11434 \
EMBEDDER_API_KEY=ollama \
EMBEDDER_MODEL=nomic-embed-text \
LLM_BASE_URL=http://localhost:11434 \
LLM_API_KEY=ollama \
LLM_MODEL=llama3.2 \
go run ./examples/chat/
Example session
Indexing knowledge base...
Indexed 8 documents.
RAG chat ready. Enter your question (Ctrl+C to exit):
────────────────────────────────────────────────────────────
> How do I add a Zigbee device?
To add a Zigbee device, open the SmartHome application,
select "Add Device" -> "Zigbee" and put the device
into pairing mode [1]. The hub will detect it within 30 seconds [1].
Sources:
[1] smarthome-zigbee (score=0.921)
────────────────────────────────────────────────────────────
Knowledge base
The example uses a built-in knowledge base about the SmartHome Hub product (8 documents). To use your own documents, replace the knowledgeBase slice in main.go.
Documentation
¶
Overview ¶
Базовый RAG-чат (CLI) — пример использования draftRAG.
Загружает набор документов в in-memory store, затем запускает интерактивный цикл вопрос-ответ с inline-цитатами.
Переменные окружения:
EMBEDDER_BASE_URL — базовый URL embedder API (по умолчанию: https://api.openai.com) EMBEDDER_API_KEY — ключ API для embedder (обязательно) EMBEDDER_MODEL — модель embeddings (по умолчанию: text-embedding-ada-002) LLM_BASE_URL — базовый URL LLM API (по умолчанию: https://api.openai.com) LLM_API_KEY — ключ API для LLM (обязательно) LLM_MODEL — модель LLM (по умолчанию: gpt-4o-mini)
Ollama (локальный режим):
Установите EMBEDDER_BASE_URL=http://localhost:11434 и EMBEDDER_MODEL=nomic-embed-text, LLM_BASE_URL=http://localhost:11434 и LLM_MODEL=llama3.2.
Запуск:
EMBEDDER_API_KEY=sk-... LLM_API_KEY=sk-... go run ./examples/chat/
Source Files
Click to show internal directories.
Click to hide internal directories.