VectifyAI/PageIndex · warning · FutureWarning

CHATGPT_API_KEY is deprecated — set OPENAI_API_KEY instead.

Error message

CHATGPT_API_KEY is deprecated — set OPENAI_API_KEY instead.

What it means

Module-level FutureWarning: if OPENAI_API_KEY is unset but the legacy CHATGPT_API_KEY is set, the latter is copied into OPENAI_API_KEY so requests still authenticate. It exists for backward compatibility with older ChatGPT-key deployments.

Source

Thrown at pageindex/utils.py:64

        pass  # best-effort: a failed repair leaves litellm's own error


def _mute_litellm_bridge_usage_warning() -> None:
    """litellm's chat→Responses bridge (e.g. OpenAI gpt-5.4+ with function
    tools) logs a chat-shaped usage dict inside a ResponseAPIUsage field
    (litellm_logging._get_assembled_streaming_response, 1.97–1.98), and
    pydantic reports it on every streamed turn. Hide exactly that message;
    every other warning still surfaces."""
    import warnings
    warnings.filterwarnings(
        "ignore",
        message=r"Pydantic serializer warnings:\s+"
                r"(PydanticSerializationUnexpectedValue\()?Expected `ResponseAPIUsage`")

# Backward compatibility: support CHATGPT_API_KEY as alias for OPENAI_API_KEY
if not os.getenv("OPENAI_API_KEY") and os.getenv("CHATGPT_API_KEY"):
    import warnings
    warnings.warn("CHATGPT_API_KEY is deprecated — set OPENAI_API_KEY "
                  "instead.", FutureWarning)
    os.environ["OPENAI_API_KEY"] = os.getenv("CHATGPT_API_KEY")

def count_tokens(text, model=None):
    if not text:
        return 0
    import litellm
    return litellm.token_counter(model=model, text=text)


def _strip_prefix(s, prefix):
    if s.startswith(prefix):
        return s[len(prefix):]
    return s


def run_off_loop(func, *args):
    """Run func now, or on a worker thread when this thread already runs an

View on GitHub (pinned to afb5e11976)

Solutions

1. Rename the env var: export OPENAI_API_KEY=<same value> and remove CHATGPT_API_KEY from .env/CI secrets
2. Reload the shell/session so the old var is gone
3. Filter FutureWarning only temporarily if you cannot change the environment yet

Example fix

# before (.env)
CHATGPT_API_KEY=sk-...
# after (.env)
OPENAI_API_KEY=sk-...

import os
if not os.getenv('OPENAI_API_KEY') and os.getenv('CHATGPT_API_KEY'):
    os.environ['OPENAI_API_KEY'] = os.environ.pop('CHATGPT_API_KEY')
# import pageindex afterwards

import warnings
with warnings.catch_warnings():
    warnings.simplefilter('ignore', FutureWarning)
    import pageindex  # legacy env still works, warning silenced

When it happens

Trigger: Importing pageindex.utils in an environment that only defines CHATGPT_API_KEY (no OPENAI_API_KEY) — commonly a leftover from older scripts or CI secrets named the old way.

Common situations: CI/CD secrets or .env files created before the rename; shared containers where the old var leaks in; team conventions stuck on the legacy name.

Related errors

• The OpenAI backend is not configured: {exc}
• The Anthropic backend is not configured: set the ANTHROPIC_A
• api_key is an empty string. Pass a real PageIndex API key fo
• optimize_expand is deprecated: pass optimize='full', 'merge'

AI-assisted analysis of VectifyAI/PageIndex@afb5e11976 (2026-08-27). Data as JSON: /api/errors/ac852056c10a30e2. Report an issue: GitHub.