unexpected keyword argument 'model_tokens' for Anthropic model

[andywatkins-ancoris]

Describe the bug
Using "model": "anthropic/claude-3-7-sonnet-latest", I got a warning message saying I needed to specify model_tokens.

Then, when I did specify model_tokens, I got an error saying
"Messages.create() got an unexpected keyword argument 'model_tokens'"

To Reproduce
I'm following the tutorial at https://scrapegraph-ai.readthedocs.io/en/latest/getting_started/examples.html
but have swapped out the model.

graph_config = {
    "llm": {
        # "api_key": openai_key,
        "api_key": anthropic_key,
        # "model": "openai/gpt-4o",
        "model": "anthropic/claude-3-7-sonnet-latest",
    },
}

When I ran this, the output started like so:

Max input tokens for model anthropic/claude-3-7-sonnet-latest not found,
                    please specify the model_tokens parameter in the llm section of the graph configuration.
                    Using default token size: 8192

So, I modified my graph_config to include model_tokens, like this, expecting this might solve the problem:

graph_config = {
    "llm": {
        # "api_key": openai_key,
        "api_key": anthropic_key,
        # "model": "openai/gpt-4o",
        "model": "anthropic/claude-3-7-sonnet-latest",
        "model_tokens": 4000,   # ADDED model_tokens TO RESOLVE THE ABOVE ERROR
    },
}

When I ran the program again, I got a Python stacktrace saying

TypeError: Messages.create() got an unexpected keyword argument 'model_tokens'

here's the whole stacktrace (I've redacted personal information from the path)

Error during chain execution: Messages.create() got an unexpected keyword argument 'model_tokens'
Traceback (most recent call last):
  File "/redacted_path/scrapegraph.py", line 51, in <module>
    result = smart_scraper_graph.run()
             ^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/redacted_path/venv/lib/python3.12/site-packages/scrapegraphai/graphs/smart_scraper_graph.py", line 296, in run
    self.final_state, self.execution_info = self.graph.execute(inputs)
                                            ^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/redacted_path/venv/lib/python3.12/site-packages/scrapegraphai/graphs/base_graph.py", line 363, in execute
    state, exec_info = self._execute_standard(initial_state)
                       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/redacted_path/venv/lib/python3.12/site-packages/scrapegraphai/graphs/base_graph.py", line 308, in _execute_standard
    raise e
  File "/redacted_path/venv/lib/python3.12/site-packages/scrapegraphai/graphs/base_graph.py", line 281, in _execute_standard
    result, node_exec_time, cb_data = self._execute_node(
                                      ^^^^^^^^^^^^^^^^^^^
  File "/redacted_path/venv/lib/python3.12/site-packages/scrapegraphai/graphs/base_graph.py", line 205, in _execute_node
    result = current_node.execute(state)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/redacted_path/venv/lib/python3.12/site-packages/scrapegraphai/nodes/generate_answer_node.py", line 193, in execute
    answer = self.invoke_with_timeout(
             ^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/redacted_path/venv/lib/python3.12/site-packages/scrapegraphai/nodes/generate_answer_node.py", line 79, in invoke_with_timeout
    response = chain.invoke(inputs)
               ^^^^^^^^^^^^^^^^^^^^
  File "/redacted_path/venv/lib/python3.12/site-packages/langchain_core/runnables/base.py", line 3046, in invoke
    input_ = context.run(step.invoke, input_, config)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/redacted_path/venv/lib/python3.12/site-packages/langchain_core/language_models/chat_models.py", line 395, in invoke
    self.generate_prompt(
  File "/redacted_path/venv/lib/python3.12/site-packages/langchain_core/language_models/chat_models.py", line 980, in generate_prompt
    return self.generate(prompt_messages, stop=stop, callbacks=callbacks, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/redacted_path/venv/lib/python3.12/site-packages/langchain_core/language_models/chat_models.py", line 799, in generate
    self._generate_with_cache(
  File "/redacted_path/venv/lib/python3.12/site-packages/langchain_core/language_models/chat_models.py", line 1045, in _generate_with_cache
    result = self._generate(
             ^^^^^^^^^^^^^^^
  File "/redacted_path/venv/lib/python3.12/site-packages/langchain_anthropic/chat_models.py", line 1498, in _generate
    data = self._create(payload)
           ^^^^^^^^^^^^^^^^^^^^^
  File "/redacted_path/venv/lib/python3.12/site-packages/langchain_anthropic/chat_models.py", line 1357, in _create
    return self._client.messages.create(**payload)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/redacted_path/venv/lib/python3.12/site-packages/anthropic/_utils/_utils.py", line 283, in wrapper
    return func(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^
TypeError: Messages.create() got an unexpected keyword argument 'model_tokens'

Expected behavior
I expect model_tokens to work, as per the error message I received

Desktop/Laptop environment

• OS: macOS
• Python 3.12.7 (in a venv)
• scrapegraphai==1.61.0

Additional context
For context, I'm a seasoned Python programmer.

However, I'm completely new to Scrapegraph, and to writing AI agents. Maybe this can be easily explained by something simple that I've missed.

I tried searching the documentation (https://docs.scrapegraphai.com/introduction) for key words like model_tokens but could not find anything.

[dosubot]

Hi @andywatkins-ancoris! I'm Dosu and I’m helping the Scrapegraph-ai team.

This error happens because ScrapegraphAI expects 'model_tokens' for some models, but does not filter it out before passing the config to the Anthropic SDK, which does not recognize this parameter. As a result, you get the "unexpected keyword argument 'model_tokens'" TypeError.

The official ScrapegraphAI documentation and examples for Anthropic models do not include or mention 'model_tokens' in the config, and there is currently no user-level workaround to set token limits for Anthropic models without hitting this error. The best approach is to remove 'model_tokens' from your config when using Anthropic models and ignore the warning about missing token limits unless it causes functional issues. This is a known limitation/bug in the current version of ScrapegraphAI and is not yet addressed in the codebase or docs (example).

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}    

[andywatkins-ancoris]

Thank you for your quick feedback.

I asked Gemini Code Assist, which advised that the Anthropic API is expecting "max_tokens" instead.

This might be helpful for the next newbie who runs into this issue.

graph_config = {
    "llm": {
        "api_key": anthropic_key,
        "model": "anthropic/claude-3-sonnet-latest",
        # FIX: The Anthropic API expects 'max_tokens' to limit the output.
        "max_tokens": 4000,
    },
}