Skip to Content
Model GuideIntegrationsLangChain

LangChain

Use the langchain-openai integration for standard OpenAI-compatible chat calls. LangChain’s provider adapter does not guarantee preservation of nonstandard upstream response fields.

1. Install the integration

python -m pip install -U langchain-openai

Set FENNO_API_KEY and an available Chat Completions FENNO_MODEL. For production, pin the package version you validate rather than upgrading implicitly on each deployment.

2. Make a minimal request

import os
from langchain_openai import ChatOpenAI
 
llm = ChatOpenAI(
    model=os.environ["FENNO_MODEL"],
    base_url="https://api.fenno.ai/v1",
    api_key=os.environ["FENNO_API_KEY"],
    timeout=60,
    max_retries=0,
)
response = llm.invoke("Reply with OK only.")
print(response.content)

Passing both base_url and api_key explicitly avoids accidentally sending the request to the default OpenAI service or reading another provider’s key.

3. Stream text

Using the same llm:

for chunk in llm.stream("Explain what an API gateway does."):
    if isinstance(chunk.content, str):
        print(chunk.content, end="", flush=True)

Some message content can be structured blocks rather than a string. Applications using tools should handle tool-call data as well, not only render text.

Tools, retrieval and other features

  • bind_tools prepares tool definitions; your application or agent executor still controls tool execution and permissions.
  • with_structured_output can select different mechanisms depending on method and version. Choose a mechanism your model supports and verify its request fields.
  • An embedding model is a separate configuration. When adding an embedding adapter, explicitly set its model, Fenno endpoint and credentials too.
  • Chat Completions compatibility does not promise all Responses features or nonstandard reasoning fields are preserved.
  • Tracing is separate from model routing. Review any enabled tracing or callbacks before sending private prompts to an external observability service.

Compare Fenno usage against a minimal direct SDK request when debugging. Check the adapter’s actual endpoint and retry settings before adding another application retry loop.

See the official ChatOpenAI guide , tool calling, and embeddings.

OpenAI SDKLlamaIndex