Integration Docs

Connect your IDE or agent to Lightning Zeus.

Use one OpenAI-compatible endpoint for multi-model inference, model checks, coding agents, and custom tooling.

Canonical Base URL

https://lightningzeus.com/v1

Replace YOUR_API_KEY with an active key and YOUR_MODEL_ID with an ID returned by the model check below.

Model guide · DeepSeek

Start building with DeepSeek V4.1 Flash Abliterated

Connect with an active Lightning Zeus API key using OpenAI-compatible Chat Completions or Anthropic-compatible Messages. Availability depends on your account and the upstream service.

01 · Create an account

02 · Activate your key

New keys require administrator activation. Ask your administrator about access and limits.

03 · Send a request

Check your model list, then use either API format below.

Model ID

deepseek-v4.1-flash-abliterated

OpenAI base URL: https://lightningzeus.com/v1 · Anthropic base URL: https://lightningzeus.com

1. Set your key and check access

Use a private terminal. Enter your active key when prompted; keep it out of source code and public repositories. Commands below use this shell variable.

Terminal · model list

read -s -p "Lightning Zeus API key: " LZ_API_KEY; echo
export LZ_API_KEY
curl https://lightningzeus.com/v1/models \
  -H "Authorization: Bearer $LZ_API_KEY"

Look for the exact model ID in data. If your key was just created, request activation before testing.

2. OpenAI-compatible request

Send a chat completion. The answer is in choices[0].message.content.

Terminal · Chat Completions

curl https://lightningzeus.com/v1/chat/completions \
  -H "Authorization: Bearer $LZ_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-v4.1-flash-abliterated",
    "messages": [{"role": "user", "content": "Reply with exactly: READY"}],
    "max_tokens": 40
  }'

3. Anthropic-compatible request

The Messages endpoint accepts x-api-key and an anthropic-version header. Specify max_tokens and the exact model ID.

Terminal · Messages

curl https://lightningzeus.com/v1/messages \
  -H "x-api-key: $LZ_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "content-type: application/json" \
  -d '{
    "model": "deepseek-v4.1-flash-abliterated",
    "max_tokens": 40,
    "messages": [{"role": "user", "content": "Reply with exactly: READY"}]
  }'

Read text from items in the response’s content array whose type is text. Other block types, such as thinking, may also be present; do not assume the first block contains the answer.

Use with a compatible application

Choose the application’s custom provider setting, enter your active API key, the appropriate base URL above and the exact model ID. Apps differ in which endpoint and features they support; test your particular workflow. The tabs below provide additional examples.

Troubleshooting and key care

401 or 403: Check the key, activation, account status and model access with your administrator.
Model unavailable: Check the authenticated model list and spelling; availability also depends on the upstream service.
429: Your account or key may have reached a rate or usage limit. Reduce request frequency and ask your administrator about your allowance.
5xx or timeouts: Retry with a delay; if the issue persists, contact your administrator with the time, endpoint and status code. Never include your API key.
Key exposed? Ask your administrator to disable it and issue a replacement. Run unset LZ_API_KEY when finished.

Basic text requests have been tested on both endpoints. Streaming, tools, images and individual client integrations should be validated for your use case; compatibility can vary.

Check available models

The model list is authenticated. Run this with your active API key.

DeepSeek V4.1 Flash Abliterated uses the model ID deepseek-v4.1-flash-abliterated. Check the returned list for models available to your account.

bash

curl "https://lightningzeus.com/v1/models" \
  -H "Authorization: Bearer YOUR_API_KEY"

Use one of the returned model IDs in chat completions, Claude-compatible tooling, or an IDE custom provider.

Basic chat request

Use the OpenAI Chat Completions shape with the Lightning Zeus base URL.

bash

curl "https://lightningzeus.com/v1/chat/completions" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "YOUR_MODEL_ID",
    "messages": [
      {"role": "user", "content": "Plan a cost-efficient model routing strategy."}
    ]
  }'

OpenAI Python SDK

Point the client at Lightning Zeus and keep your existing OpenAI-compatible code.

python

from openai import OpenAI

client = OpenAI(
    api_key="YOUR_API_KEY",
    base_url="https://lightningzeus.com/v1",
)

response = client.chat.completions.create(
    model="YOUR_MODEL_ID",
    messages=[{"role": "user", "content": "Summarize the value of multi-model inference."}],
)

print(response.choices[0].message.content)

Claude Code

Use the Anthropic-compatible environment variables when your workflow expects Claude-style APIs.

json

{
  "env": {
    "ANTHROPIC_BASE_URL": "https://lightningzeus.com/",
    "ANTHROPIC_AUTH_TOKEN": "YOUR_API_KEY",
    "ANTHROPIC_MODEL": "YOUR_MODEL_ID",
    "ANTHROPIC_DEFAULT_SONNET_MODEL": "YOUR_MODEL_ID",
    "ANTHROPIC_SMALL_FAST_MODEL": "YOUR_MODEL_ID"
  }
}

Place this in your Claude Code settings file and choose model IDs that exist in your authenticated model list.

Codex / OpenAI-compatible tools

For tools that support a custom OpenAI base URL, use the Lightning Zeus endpoint and your API key.

env

OPENAI_API_KEY=YOUR_API_KEY
OPENAI_BASE_URL=https://lightningzeus.com/v1
OPENAI_MODEL=YOUR_MODEL_ID

If a Codex surface only supports ChatGPT subscription OAuth and not custom API endpoints, use this gateway through SDK, CLI, or IDE tools that expose a custom provider setting.

OpenClaw

Configure Lightning Zeus as an OpenAI-compatible custom provider.

yaml

providers:
  lightningzeus:
    type: openai-compatible
    baseUrl: "https://lightningzeus.com/v1"
    apiKey: "YOUR_API_KEY"

agents:
  defaults:
    models:
      - provider: lightningzeus
        model: YOUR_MODEL_ID

Run the model check first, then replace model IDs with the exact IDs returned for your account.

Hermes IDE

Use a custom OpenAI-compatible provider for each Hermes session that should route through Lightning Zeus.

json

{
  "providers": {
    "lightningzeus": {
      "kind": "openai-compatible",
      "baseUrl": "https://lightningzeus.com/v1",
      "apiKey": "YOUR_API_KEY",
      "models": ["YOUR_MODEL_ID"]
    }
  },
  "defaultProvider": "lightningzeus",
  "defaultModel": "YOUR_MODEL_ID"
}

Use separate Hermes sessions for different models when you want review, implementation, and verification to run independently.

One base URL

Point tools at https://lightningzeus.com/v1.

One API key

Use an active key from your customer dashboard.

Many model families

Check `/models` and choose the right model per workflow.