langchain-veida

A LangChain tool that generates images with no API key and no account.

pip install langchain-veida

from langchain_veida import VeidaImageGenerator

tool = VeidaImageGenerator()
url = tool.invoke({"prompt": "a paper boat on a calm lake at sunrise, soft light", "aspect_ratio": "16:9"})
# https://cdn.veida.ai/uploads/....webp

Hand it to an agent like any other tool:

from langchain.agents import create_agent

agent = create_agent("openai:gpt-5-mini", tools=[VeidaImageGenerator()])

The tool calls the anonymous tier of Veida, so it works on a machine with nothing configured. Measured against the running service: one image per client id (the library mints a new id each call), about seven images a day per machine, 1K output with a watermark. A run takes 40 to 120 seconds.

When the free allowance is spent or the content filter refuses a prompt, the tool returns that as text instead of raising, so the agent can wait or reword rather than crash.

Signing in on veida.ai removes the watermark and opens larger models such as GPT Image 2 for pictures that must contain readable text. Prompt examples live in the prompt library.

Built on the zero-dependency veida client; full docs at veida.readthedocs.io. MIT.