ScrapeGraphAI/Scrapegraph-ai · error · Exception

Error instancing model: {e}

Error message

Error instancing model: {e}

What it means

A catch-all Exception raised at the end of _create_llm's try block: any exception thrown while constructing the provider-specific chat model (bad API key, unknown kwargs, network/auth errors from the LangChain class, etc.) is re-raised as 'Error instancing model: {e}'. The original message is embedded but the original type/traceback context is flattened, so inspect the inner text.

Source

Thrown at scrapegraphai/graphs/abstract_graph.py:291

                elif model_provider == "xai":
                    return XAI(**llm_params)

                elif model_provider == "togetherai":
                    try:
                        from langchain_together import ChatTogether
                    except ImportError:
                        raise ImportError(
                            """The langchain_together module is not installed.
                                          Please install it using `pip install langchain-together`."""
                        )
                    return ChatTogether(**llm_params)

                elif model_provider == "nvidia":
                    return Nvidia(**llm_params)

        except Exception as e:
            raise Exception(f"Error instancing model: {e}")

    def get_state(self, key=None) -> dict:
        """ ""
        Get the final state of the graph.

        Args:
            key (str, optional): The key of the final state to retrieve.

        Returns:
            dict: The final state of the graph.
        """

        if key is not None:
            return self.final_state[key]
        return self.final_state

    def append_node(self, node):
        """

View on GitHub (pinned to 532dfffbf6)

Solutions

1. Read the inner '{e}' text — it usually names the real cause (auth, kwarg, deployment).
2. Validate the llm config keys against the LangChain class signature for your provider and remove stray keys.
3. Test the model in isolation: instantiate the LangChain class directly with the same params to reproduce.
4. Check API key validity/quotas in the provider console.

Example fix

# before
config = {'llm': {'model_provider': 'openai', 'model': 'gpt-4o', 'api_keys': key}}  # typo: api_keys

# after
config = {'llm': {'model_provider': 'openai', 'model': 'gpt-4o', 'api_key': key}}

cfg = config['llm']
allowed = {'model_provider', 'model', 'api_key', 'temperature', 'max_tokens'}
extra = set(cfg) - allowed
assert not extra, f'unexpected llm config keys that get splatted into the constructor: {extra}'

try:
    graph = SmartScraperGraph(prompt=p, config=config)
except Exception as e:
    msg = str(e)
    if 'Error instancing model' not in msg:
        raise
    logger.error('LLM construction failed: %s', msg)
    raise SystemExit('Check api_key / model name / constructor kwargs') from e

When it happens

Trigger: Any constructor failure of OpenAI(), ChatTogether(), Nvidia(), AzureChatOpenAI(), etc. with **llm_params: missing/invalid api_key, unexpected kwarg passed through from config, wrong endpoint/deployment name, or an auth connectivity failure at init.

Common situations: Typos in config keys that get splatted into the model constructor (e.g. 'temperature' misspelled or an extra key the class rejects); expired API keys; azure deployment_name mismatches; version changes in LangChain constructor signatures.

Related errors

• model_tokens not specified
• Provider {llm_params["model_provider"]} is not supported.
• Provider {llm_params["model_provider"]} is not supported.
• Model not supported
• The langchain_together module is not installed.

AI-assisted analysis of ScrapeGraphAI/Scrapegraph-ai@532dfffbf6 (2026-08-28). Data as JSON: /api/errors/e41b270f9d6ead02. Report an issue: GitHub.