Documentation

DeepSeek Configuration

This guide needs an API key

Create an account and generate one in about a minute, then use it wherever the steps below expect a key.

No card required to sign up · API usage requires a top-up

DeepSeek models are available through Code By AI's OpenAI-compatible Chat Completions endpoint. Point any OpenAI SDK or compatible client at the site base URL, use a Code By AI API key, and send an explicit DeepSeek model ID.

Models and prices

  • deepseek-v4.1-flash
  • deepseek-v4-pro

Neither model is a channel default, so send the exact ID in every request. See the DeepSeek pricing page for current input, cache-hit, and output rates. The off-peak (valley) rate applies around the clock with no peak surcharge.

Configure the environment variables

Both variables below are read by OpenAI-compatible clients. The key is created on the API Keys page.

macOS or Linux

Add these lines to ~/.zshrc (zsh) or ~/.bashrc (bash), then run source on the file:

1
export OPENAI_BASE_URL="https://www.codebyai.net/api/agent"
2
export OPENAI_API_KEY="sk-proj-xxxx"

Windows

Set the same variables for your user account in Command Prompt:

1
setx OPENAI_BASE_URL "https://www.codebyai.net/api/agent"
2
setx OPENAI_API_KEY "sk-proj-xxxx"

Open a new terminal after setx so the client inherits the values.

Send a Chat Completions request

The client must use OpenAI-compatible Chat Completions. Both /api/agent/v1/chat/completions and /api/agent/chat/completions are routed:

1
curl --fail-with-body "https://www.codebyai.net/api/agent/v1/chat/completions" \
2
  -H "Authorization: Bearer $OPENAI_API_KEY" \
3
  -H "content-type: application/json" \
4
  -d '{
5
    "model": "deepseek-v4.1-flash",
6
    "messages": [{ "role": "user", "content": "Hello" }]
7
  }'

Replace the model ID with deepseek-v4-pro to use the larger model. The Responses endpoint does not serve DeepSeek, and omitting model does not select a DeepSeek model.

Use an OpenAI SDK

The official SDK reads OPENAI_BASE_URL and OPENAI_API_KEY from the environment configured above. Pass the model ID on every call:

1
from openai import OpenAI
2
 
3
client = OpenAI()
4
response = client.chat.completions.create(
5
    model="deepseek-v4.1-flash",
6
    messages=[{"role": "user", "content": "Hello"}],
7
)
8
print(response.choices[0].message.content)

Troubleshooting

  • 403 with Invalid key means the key is invalid, revoked, or expired. Check OPENAI_API_KEY and make sure the Authorization header sends it as a Bearer token.
  • 400 Unknown model means the ID is not in the catalog. Use deepseek-v4.1-flash or deepseek-v4-pro exactly.
  • 400 on /v1/responses is expected: DeepSeek is served only on Chat Completions.
  • Balance and per-model usage are listed in the dashboard; a request rejected before the model answers is not charged.

Get started

Four steps to your first request

  1. 01

    Create a free account

    Sign up with Google, GitHub or email and set a password. No card is required.

  2. 02

    Generate an API key

    Create a key in one click from the dashboard. It lets your CLI authenticate with Code By AI.

  3. 03

    Top up your balance

    Add credit before your first request; usage is charged per token from your balance.

  4. 04

    Point your CLI at Code By AI

    Export one base-URL variable, or add a model provider block to config.toml. The setup guides list the exact lines for each agent.

Use it nowTop up to send a requestWhat a working month costs

No card required to sign up · API usage requires a top-up