Anthropic error: "anthropic-version header is required"

[cdolek]

Whatever I do, I can't use Anthropic's claude with spacy-llm. I tried both spacy.Claude-instant-1.v1 and spacy.Claude-2.v1.

I am getting the following error during nlp = assemble(config_file_path):

...
File ~/miniconda3/envs/_my_env_/lib/python3.9/site-packages/requests/models.py:1021, in Response.raise_for_status(self)
   1020 if http_error_msg:
-> 1021     raise HTTPError(http_error_msg, response=self)

HTTPError: 400 Client Error: Bad Request for url: https://api.anthropic.com/v1/complete
...
...

File ~/miniconda3/envs/_my_env_/lib/python3.9/site-packages/spacy_llm/models/rest/anthropic/model.py:80, in Anthropic.__call__.<locals>._request(json_data)
     78     if error["type"] == "not_found_error":
     79         error_msg += f". Ensure that the selected model ({self._name}) is supported by the API."
---> 80     raise ValueError(error_msg) from ex
     81 response = r.json()
     83 # c.f. https://console.anthropic.com/docs/api/errors

ValueError: Request to Anthropic API failed: {'type': 'invalid_request_error', 'message': 'anthropic-version header is required'}

Here is my config file:

[nlp]
lang = "en"
pipeline = ["llm"]

[components]

[components.llm]
factory = "llm"

[components.llm.model]
@llm_models = "spacy.Claude-instant-1.v1"
name = "claude-instant-1"
max_request_time = 5000
max_tries = 10
interval = 3
# config = {"temperature": 0.0}

[components.llm.task]
@llm_tasks = "spacy.TextCat.v3"
labels = ["MY_LABEL"]
exclusive_classes = True

[rmitsch]

Hi @cdolek, thanks for reporting this! This is strange, we test for this in our CI. We'll look into this and report back here.

[rmitsch]

Hm, I tried this locally and it works for me. Could you provide which spacy-llm version you are using and a minimal, reproducible Python script?

For reference my setup:

from pathlib import Path
from spacy_llm.util import assemble

if __name__ == '__main__':
    nlp = assemble(Path(__file__).parent.parent / "config.cfg")
    doc = nlp("This is a test.")

And your config (max_tokens_to_sample has to be set, we should probably have a default value for that) in config.cfg:

[nlp]
lang = "en"
pipeline = ["llm"]

[components]

[components.llm]
factory = "llm"

[components.llm.model]
@llm_models = "spacy.Claude-instant-1.v1"
name = "claude-instant-1"
max_request_time = 5000
max_tries = 10
interval = 3
config = {"temperature": 0.0, "max_tokens_to_sample": 1024}

[components.llm.task]
@llm_tasks = "spacy.TextCat.v3"
labels = ["MY_LABEL"]
exclusive_classes = True

[cdolek]

Same example and config as yours, gives the error for me.

from spacy_llm.util import assemble

nlp = assemble("./config.cfg")
doc = nlp("This is a test.")

I'm on a jupyter notebook and here is the output for pip list | grep spacy

spacy                         3.7.1
spacy-alignments              0.9.0
spacy-legacy                  3.0.12
spacy-llm                     0.6.0
spacy-loggers                 1.0.4
spacy-transformers            1.2.5

I'm sure that my Anthropic setup works, since the code below works:

from anthropic import Anthropic, HUMAN_PROMPT, AI_PROMPT

anthropic = Anthropic()
completion = anthropic.completions.create(
    model="claude-instant-1",
    max_tokens_to_sample=300,
    prompt=f"{HUMAN_PROMPT} How many toes do dogs have?{AI_PROMPT}",
)
print(completion.completion)

Another note, I'm using the same code and config structure for openai and palm, and they work as expected.

[cdolek]

Tried your code (not in notebook) and got the following error again with Python 3.9.16:

Traceback (most recent call last):
  File "/Users/_my_username_/miniconda3/envs/_my_env_name_/lib/python3.9/site-packages/spacy_llm/models/rest/anthropic/model.py", line 72, in _request
    r.raise_for_status()
  File "/Users/_my_username_/miniconda3/envs/_my_env_name_/lib/python3.9/site-packages/requests/models.py", line 1021, in raise_for_status
    raise HTTPError(http_error_msg, response=self)
requests.exceptions.HTTPError: 400 Client Error: Bad Request for url: https://api.anthropic.com/v1/complete

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
  File "/Users/_my_username_/_src/spacy-llm-test/main.py", line 5, in <module>
    nlp = assemble(Path(__file__).parent / "config.cfg")
  File "/Users/_my_username_/miniconda3/envs/_my_env_name_/lib/python3.9/site-packages/spacy_llm/util.py", line 48, in assemble
    return assemble_from_config(config)
  File "/Users/_my_username_/miniconda3/envs/_my_env_name_/lib/python3.9/site-packages/spacy_llm/util.py", line 28, in assemble_from_config
    nlp = load_model_from_config(config, auto_fill=True)
  File "/Users/_my_username_/miniconda3/envs/_my_env_name_/lib/python3.9/site-packages/spacy/util.py", line 587, in load_model_from_config
    nlp = lang_cls.from_config(
  File "/Users/_my_username_/miniconda3/envs/_my_env_name_/lib/python3.9/site-packages/spacy/language.py", line 1864, in from_config
    nlp.add_pipe(
  File "/Users/_my_username_/miniconda3/envs/_my_env_name_/lib/python3.9/site-packages/spacy/language.py", line 821, in add_pipe
    pipe_component = self.create_pipe(
  File "/Users/_my_username_/miniconda3/envs/_my_env_name_/lib/python3.9/site-packages/spacy/language.py", line 709, in create_pipe
    resolved = registry.resolve(cfg, validate=validate)
  File "/Users/_my_username_/miniconda3/envs/_my_env_name_/lib/python3.9/site-packages/confection/__init__.py", line 756, in resolve
    resolved, _ = cls._make(
  File "/Users/_my_username_/miniconda3/envs/_my_env_name_/lib/python3.9/site-packages/confection/__init__.py", line 805, in _make
    filled, _, resolved = cls._fill(
  File "/Users/_my_username_/miniconda3/envs/_my_env_name_/lib/python3.9/site-packages/confection/__init__.py", line 860, in _fill
    filled[key], validation[v_key], final[key] = cls._fill(
  File "/Users/_my_username_/miniconda3/envs/_my_env_name_/lib/python3.9/site-packages/confection/__init__.py", line 877, in _fill
    getter_result = getter(*args, **kwargs)
  File "/Users/_my_username_/miniconda3/envs/_my_env_name_/lib/python3.9/site-packages/spacy_llm/models/rest/anthropic/registry.py", line 103, in anthropic_claude_instant_1
    return Anthropic(
  File "/Users/_my_username_/miniconda3/envs/_my_env_name_/lib/python3.9/site-packages/spacy_llm/models/rest/base.py", line 65, in __init__
    self._verify_auth()
  File "/Users/_my_username_/miniconda3/envs/_my_env_name_/lib/python3.9/site-packages/spacy_llm/models/rest/anthropic/model.py", line 51, in _verify_auth
    raise err
  File "/Users/_my_username_/miniconda3/envs/_my_env_name_/lib/python3.9/site-packages/spacy_llm/models/rest/anthropic/model.py", line 43, in _verify_auth
    self(["test"])
  File "/Users/_my_username_/miniconda3/envs/_my_env_name_/lib/python3.9/site-packages/spacy_llm/models/rest/anthropic/model.py", line 96, in __call__
    responses = [
  File "/Users/_my_username_/miniconda3/envs/_my_env_name_/lib/python3.9/site-packages/spacy_llm/models/rest/anthropic/model.py", line 97, in <listcomp>
    _request({"prompt": f"{SystemPrompt.HUMAN} {prompt}{SystemPrompt.ASST}"})
  File "/Users/_my_username_/miniconda3/envs/_my_env_name_/lib/python3.9/site-packages/spacy_llm/models/rest/anthropic/model.py", line 80, in _request
    raise ValueError(error_msg) from ex
ValueError: Request to Anthropic API failed: {'type': 'invalid_request_error', 'message': 'anthropic-version header is required'}
[1]    58865 bus error  python main.py

Code:

from pathlib import Path
from spacy_llm.util import assemble

if __name__ == "__main__":
    nlp = assemble(Path(__file__).parent / "config.cfg")
    doc = nlp("This is a test.")

Config:

[nlp]
lang = "en"
pipeline = ["llm"]

[components]

[components.llm]
factory = "llm"

[components.llm.model]
@llm_models = "spacy.Claude-instant-1.v1"
name = "claude-instant-1"
max_request_time = 5000
max_tries = 10
interval = 3
config = {"temperature": 0.0, "max_tokens_to_sample": 1024}

[components.llm.task]
@llm_tasks = "spacy.TextCat.v3"
labels = ["INSULT", "PRAISE"]
exclusive_classes = True

[rmitsch]

Thanks for the debug info!

It's strange that I can't reproduce it. Anyway, the issues seems to be that the version header isn't included in our requests - see here. That should be an easy fix (although by not being able to replicate it I can't guarantee it'll work before releasing). I hope I get around to doing this later today, otherwise a patch release 0.6.1 should be available on Monday.

[cdolek]

Great news, thank you!

[rmitsch]

@cdolek v0.6.1 is out. Give it a try and let me know if that fixes things?

[cdolek]

@rmitsch Just tried it, same error unfortunately...

Would using anthropic-version instead of anthropic_version in spacy_llm/models/rest/anthropic/model.py work?

...
"anthropic_version": self._config.get("anthropic_version", "2023-06-01"),
...

https://github.com/search?q=repo%3Aanthropics%2Fanthropic-sdk-python+anthropic-version&type=code

[rmitsch]

spacy-llm 0.6.2 is out and uses anthropic-version instead of anthropic_version. Give it a try?

[cdolek]

@rmitsch thank you so much, it worked!