Connecting to DeepSeek

# What is DeepSeek

DeepSeek V3 is a mixture-of-experts (MoE) language model with 671 billion parameters, 37 billion of which are activated per token. It performs especially well at writing and analyzing computer code.

This is a noteworthy alternative to GPT-4o, 4o-mini, Claude 3.5 Sonnet and Haiku, offering better pricing.

Here's why it makes sense to try DeepSeek-V3 in HARPA AI:

1. DeepSeek has great analytical capabilities, comparable to GPT-4o, while the cost is 10 times less than GPT-4o, or only about 50% more than basic GPT-4o-mini.

2. It performs almost as fast as 4o-mini, but is notably smarter.

# How do I get a DeepSeek V3 API Key?

1. Go to the DeepSeek Sign Up page to get started.

2. Sign Up or Login via Google Account.

3. Top Up your Account. During testing, we found that summarizing 4x 15-minute YouTube videos costs about 1 cent, so $2 should be enough to check the functionality.

4. Go to the API Keys tab, name and generate your key.

# Adding DeepSeek connection to HARPA

5. Open Harpa and click the Connections Switch. Scroll down to the API section. Create a new API key connection (you can click copy on an existing connection).

6. Copy your DeepSeek API Key to HARPA.

• Set chat model to: deepseek-chat
• API Proxy URL: https://api.deepseek.com
• If you didn't copy the key in step 4 - simply create a new one and delete the old one.

7. Pick this DeepSeek connection and try asking something in the chat to make sure the connection works.

# DeepSeek API Issues & Solutions

# Error "API key is missing or is incorrect."

You likely need to check your chat model and API Proxy URL settings:

• Chat model: deepseek-chat
• API Proxy URL: https://api.deepseek.com

Until you confirm the connection works, remove all additional API settings - context window, system prompt, temperature, etc.

You can also create a new API key in case your old one stopped working or extra characters were added when copying it.

Find the latest information here: https://api-docs.deepseek.com/

# Error: "Request failed. Please check if API is up and running and Key is correct."

Based on the error message shown, your balance is insufficient.

Please top up your DeepSeek balance to fix this issue.

# DeepSeek API Token Consumption

To check your DeepSeek token usage, go to the usage tab in your DeepSeek account. You will find a list of requests, token consumption, and your API costs there.

# Links for further reading

• HARPA AI YouTube Channel
• Connecting to CloudGPT
• Connecting to ChatGPT
• Connecting to OpenAI API
• Connecting to Claude AI
• Connecting to Google Gemini
• Connecting to OpenRouter