Skip to main content

Text Generation

Drop-in OpenAI and Anthropic compatible

Chat and text completion via OpenAI-compatible /v1/chat/completions and Anthropic-compatible /v1/messages — point your existing SDK at the gateway and pick any model.

Text generation is the gateway's core surface. Both API formats route through the same multi-provider engine, so you can call a Claude model in OpenAI format or a GPT model in Anthropic format — the gateway translates and bills per token from your credit balance.

Generating text​

POST /v1/chat/completions (OpenAI format) · POST /v1/messages (Anthropic format)

Authentication is the same as every other gateway endpoint — a Bankr API key (bk_...) via X-API-Key or Authorization: Bearer. Generate keys at bankr.bot/api-keys.

curl -X POST https://llm.bankr.bot/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "X-API-Key: bk_YOUR_API_KEY" \
  -d '{
    "model": "claude-opus-4.8",
    "messages": [
      {"role": "system", "content": "You are a helpful assistant."},
      {"role": "user", "content": "Hello!"}
    ]
  }'

Using the OpenAI SDK, only the base URL and key change:

from openai import OpenAI

client = OpenAI(base_url="https://llm.bankr.bot/v1", api_key="bk_YOUR_API_KEY")

completion = client.chat.completions.create(
    model="claude-opus-4.8",
    messages=[{"role": "user", "content": "Hello!"}],
)
print(completion.choices[0].message.content)

Both formats are documented in full — including request/response fields and the Anthropic /v1/messages shape — in the API Reference. For framework integrations (Vercel AI SDK, LangChain, etc.), see Apps.

Streaming​

Set "stream": true to receive server-sent events as the completion is generated; the stream shape matches the provider format you called (OpenAI chat.completion.chunk or Anthropic message events). See Streaming.

Parameters​

The body is forwarded to the provider, so any parameter it accepts works. The most common:

ParameterDescription
modelRequired. Any text model from /v1/models.
messagesRequired. The conversation, in the format you're calling.
temperatureSampling temperature.
max_tokensMax tokens to generate (max_tokens on /v1/messages is required).
streamtrue for server-sent events.

Supported models​

Text generation runs on any model that produces text output — the large majority of the catalog. The source of truth is the /v1/models endpoint: a text model reports "output_modalities": ["text"].

bankr llm models    # or GET /v1/models

See Supported Models for the full catalog, and Private Inference to run select models inside a TEE.

Pricing​

Text generation bills per token from your LLM credit balance — separate input and output rates per model (some also price cached tokens). Exact per-model rates are in /v1/models under pricing and in the Models tab at bankr.bot/llm. See credit management to top up.

Next steps​