Get an OpenAI API Key
For: Danum tenants. Used for RAG embeddings (knowledge base search) plus LLM fallback. Time: 5 minutes.
Why OpenAI (not just Claude)?
Danum uses text-embedding-3-small from OpenAI for RAG (retrieval-augmented generation): finding relevant chunks from your knowledge base before answering with Claude. Anthropic’s embedding model is not yet production-ready for multi-lingual (Indonesian) text, so we still use OpenAI for embeddings.
Steps
1. Create an OpenAI account
- Go to https://platform.openai.com/signup
- Sign up with email or Google.
- Verify your phone number (required for API access).
2. Set up billing
- After signing in, go to https://platform.openai.com/account/billing/overview
- Click Add payment method and enter your card.
- Add credit (minimum $5). OpenAI also charges per request.
- Set a usage limit at https://platform.openai.com/account/billing/limits (we recommend $30/month to start).
3. Generate the API key
- Go to https://platform.openai.com/api-keys
- Click Create new secret key.
- Name it (for example
danum-ai-production). - Copy the key (format
sk-*, about 50 characters). Save it in a password manager if needed.
4. Enter it in the Danum dashboard
- Open https://ai.numintek.com/app/settings/llm-keys
- Find OpenAI API Key.
- Paste and save.
- Effective within 60 seconds.
Cost estimate
text-embedding-3-small: $0.02 per 1M tokens (about Rp 310). For a knowledge base of 1000 documents (roughly 3000 words each = 4M tokens), full re-embed costs about Rp 1,240 one-time. Per-query embeddings (per user question) cost about Rp 0.01 each.
Fallback LLM model gpt-4o-mini: $0.15 / $0.60 per 1M input/output tokens - much cheaper than Claude Sonnet.
Troubleshooting
Q: You exceeded your current quota error.
Top up credit at the OpenAI billing dashboard.
Q: RAG retrieval is not accurate.
Check that documents are ingested (Knowledge > Documents > status = READY). Re-index if you just uploaded many documents.
Q: Required or optional?
- Required if you use the Knowledge Base and RAG features.
- Optional if your AI agent is plain conversation without document retrieval.