[Bug]: openai adapter does not work

CyprienRicque commented 7 months ago

Current Behavior

The example in the readme produces the error APIRemovedInV1

Steps To Reproduce

!pip install openai gptcache

# example in the readme
from gptcache import cache
from gptcache.adapter import openai

cache.init()
cache.set_openai_key()

question = 'what is github'
answer = openai.ChatCompletion.create(
      model='gpt-3.5-turbo',
      messages=[
        {
            'role': 'user',
            'content': question
        }
      ],
    )
print(answer)

try at: https://colab.research.google.com/drive/1TjA2plt9ZXLHIQVvZ763Nj6fzshYGSoN?usp=sharing

Environment

Google colab
python 3.10.12
openai 1.3.5
gptcache 0.1.42

Anything else?

SimFG commented 7 months ago

yes, as you can see from the issue, it is due to the incompatibility of the openai upgrade interface. Before being compatible with openai 1.x, you can try to use the get and put methods of gptcache to implement cache

CyprienRicque commented 7 months ago

Ok thank you for the direction! I'll implement it

judahkshitij commented 7 months ago

yes, as you can see from the issue, it is due to the incompatibility of the openai upgrade interface. Before being compatible with openai 1.x, you can try to use the get and put methods of gptcache to implement cache

@SimFG @CyprienRicque How can we in a Python program create and then interact with a cache without involving any LLM model? Asking because I want to benchmark some cache settings in a Python code without worrying about setting up an LLM model to interact with. I tried calling cache.put("Hi", "Hi back") but got the error AttributeError: 'Cache' object has no attribute 'put'. Is there a way to use cache just using get and put in Python code (i.e., after creating and initializing cache with some settings like distance thresahold, etc. in the Python code rather than starting cache as a server) without involving any LLM? Any help on this is appreciated.

judahkshitij commented 7 months ago

@SimFG Also, after I stop a running gptcache server to change some settings in the config yaml file (such as distance/similarity threshold) and restart the server using the cmd "gptcache_server -s 127.0.0.1 -p 8000 -f gptcache_server_config.yaml", I get the following error:

start to install package: ruamel-yaml successfully installed package: ruamel-yaml Traceback (most recent call last): File "/mnt/nfshome/judah.kshitij/.conda/envs/alpaca-lora_env/bin/gptcache_server", line 8, in sys.exit(main()) File "/mnt/nfshome/judah.kshitij/.conda/envs/alpaca-lora_env/lib/python3.8/site-packages/gptcache_server/server.py", line 178, in main init_conf = init_similar_cache_from_config(config_dir=args.cache_config_file) File "/mnt/nfshome/judah.kshitij/.conda/envs/alpaca-lora_env/lib/python3.8/site-packages/gptcache/adapter/api.py", line 221, in init_similar_cache_from_config data_manager = manager_factory(**storage_config) File "/mnt/nfshome/judah.kshitij/.conda/envs/alpaca-lora_env/lib/python3.8/site-packages/gptcache/manager/factory.py", line 125, in manager_factory return get_data_manager(s, v, o, e) File "/mnt/nfshome/judah.kshitij/.conda/envs/alpaca-lora_env/lib/python3.8/site-packages/gptcache/manager/factory.py", line 200, in get_data_manager return SSDataManager(cache_base, vector_base, object_base, eviction_base, max_size, clean_size, eviction) File "/mnt/nfshome/judah.kshitij/.conda/envs/alpaca-lora_env/lib/python3.8/site-packages/gptcache/manager/data_manager.py", line 247, in init self.eviction_base.put(self.s.get_ids(deleted=False)) File "/mnt/nfshome/judah.kshitij/.conda/envs/alpaca-lora_env/lib/python3.8/site-packages/gptcache/manager/eviction/memory_cache.py", line 59, in put self._cache[obj] = True File "/mnt/nfshome/judah.kshitij/.conda/envs/alpaca-lora_env/lib/python3.8/site-packages/cachetools/init.py", line 217, in setitem cache_setitem(self, key, value) File "/mnt/nfshome/judah.kshitij/.conda/envs/alpaca-lora_env/lib/python3.8/site-packages/cachetools/init.py", line 79, in setitem self.popitem() File "/mnt/nfshome/judah.kshitij/.conda/envs/alpaca-lora_env/lib/python3.8/site-packages/gptcache/manager/eviction/memory_cache.py", line 15, in wrapper wrapper_func(keys) TypeError: 'NoneType' object is not callable

Only way the server starts is when I also change the cache dir of the cache in the yaml config file. How can I fix this issue? I am not sure why just changing the simi threshold in yaml config and restarting the same server would give above error. Any insights for resolving above issue is appreciated.

judahkshitij commented 6 months ago

yes, as you can see from the issue, it is due to the incompatibility of the openai upgrade interface. Before being compatible with openai 1.x, you can try to use the get and put methods of gptcache to implement cache

@SimFG Are you suggesting here that until openai adapter becomes compatible with openai 1.x, we use cache by starting the gptcache server and access it using get and put methods? Any help on providing more details on this is appreciated.

SimFG commented 6 months ago

@judahkshitij a example case: https://github.com/zilliztech/GPTCache/blob/main/examples/adapter/api.py about the gptcache server, you can see: https://github.com/zilliztech/GPTCache/blob/main/examples/README.md#How-to-use-GPTCache-server

zilliztech / GPTCache