Graphify-Labs/graphify · error · ValueError

Azure OpenAI backend requires AZURE_OPENAI_ENDPOINT to be se

Error message

Azure OpenAI backend requires AZURE_OPENAI_ENDPOINT to be set.

What it means

ValueError raised when backend='azure' is configured but the AZURE_OPENAI_ENDPOINT environment variable is empty/whitespace. graphify constructs the Azure client from this env var (stripped) plus the API key; without an endpoint there is nothing to call, so it fails fast before any request.

Source

Thrown at graphify/llm.py:2685

                read_timeout=_resolve_api_timeout(),
                connect_timeout=10,
                retries={"max_attempts": _resolve_max_retries() + 1, "mode": "adaptive"},
            ),
        )
        resp = client.converse(
            modelId=mdl,
            messages=[{"role": "user", "content": [{"text": prompt}]}],
            inferenceConfig=_bedrock_inference_config(max_tokens, mdl),
        )
        bu = resp.get("usage") or {}
        if bu:
            _rec(bu.get("inputTokens", 0), bu.get("outputTokens", 0))
        return _bedrock_response_text(resp, default="")

    if backend == "azure":
        endpoint = os.environ.get("AZURE_OPENAI_ENDPOINT", "").strip()
        if not endpoint:
            raise ValueError(
                "Azure OpenAI backend requires AZURE_OPENAI_ENDPOINT to be set."
            )
        azure_client = _azure_client(key, endpoint)
        azure_kwargs: dict = {
            "model": mdl,
            "messages": [{"role": "user", "content": prompt}],
            "max_completion_tokens": max_tokens,
        }
        azure_temp = _resolve_temperature(cfg.get("temperature", 0), mdl)
        if azure_temp is not None:
            azure_kwargs["temperature"] = azure_temp
        resp = azure_client.chat.completions.create(**azure_kwargs)
        if not resp.choices or resp.choices[0].message is None:
            raise ValueError("Azure OpenAI returned empty or filtered response")
        au = getattr(resp, "usage", None)
        if au is not None:
            _rec(getattr(au, "prompt_tokens", 0), getattr(au, "completion_tokens", 0))
        return resp.choices[0].message.content or ""

View on GitHub (pinned to 7fe58b0b0f)

Solutions

Set the endpoint: export AZURE_OPENAI_ENDPOINT='https://<your-resource>.openai.azure.com' (find it on the Azure OpenAI resource Overview page).
If using a .env file, confirm it is actually loaded by the process that runs graphify (printenv AZURE_OPENAI_ENDPOINT to verify).
If you did not mean Azure, switch the backend to openai/claude/etc.

Example fix

# before
export AZURE_OPENAI_API_KEY=...
# AZURE_OPENAI_ENDPOINT missing
$ graphify label --backend azure   # ValueError

# after
export AZURE_OPENAI_ENDPOINT="https://my-resource.openai.azure.com"  # your resource URL
$ graphify label --backend azure

Defensive patterns

Strategy: validation

Validate before calling

import os
endpoint = os.environ.get("AZURE_OPENAI_ENDPOINT", "").strip()
if not endpoint:
    raise SystemExit(
        "backend 'azure' requires AZURE_OPENAI_ENDPOINT"
        " (https://<resource>.openai.azure.com)"
    )

Try / catch

try:
    result = call_llm(prompt, backend="azure")
except ValueError as exc:
    if "AZURE_OPENAI_ENDPOINT" in str(exc):
        raise SystemExit("Set AZURE_OPENAI_ENDPOINT to your Azure OpenAI resource URL") from exc
    raise

Prevention

Treat Azure as a three-part config: endpoint env var, key env var, deployed model name - check all three together.
Fail fast at startup on missing env config instead of discovering it mid-run.
Prefer a single sourced env file over scattered exports.

When it happens

Trigger: Selecting backend 'azure' with AZURE_OPENAI_ENDPOINT unset or containing only whitespace (llm.py:2684-2687). The key comes from _azure_client(key, endpoint); the endpoint must be the Azure OpenAI resource endpoint (https://<resource>.openai.azure.com).

Common situations: Copy-pasting config from a colleague without the env var; .env file not loaded in CI; setting AZURE_OPENAI_API_KEY but forgetting the endpoint var; shell quoting mistakes that leave the var empty.

Related errors

AI-assisted analysis of Graphify-Labs/graphify@7fe58b0b0f (2026-08-14). Data as JSON: /api/errors/0a477ede58ed0f35. Report an issue: GitHub.