Thank you for contributing to LangChain!
- [x] **PR title**: langchain: add URL parameter to ChatDeepInfra class
- [x] **PR message**: add URL parameter to ChatDeepInfra class
- **Description:** This PR introduces a url parameter to the
ChatDeepInfra class in LangChain, allowing users to specify a custom
URL. Previously, the URL for the DeepInfra API was hardcoded to
"https://stage.api.deepinfra.com/v1/openai/chat/completions", which
caused issues when the staging endpoint was not functional. The _url
method was updated to return the value from the url parameter, enabling
greater flexibility and addressing the problem. out!
---------
Co-authored-by: Erick Friis <erick@langchain.dev>
- **Description:** This PR fixes the response parsing logic for
`ChatDeepInfra`, more specifially `_convert_delta_to_message_chunk()`,
which is invoked when streaming via `ChatDeepInfra`.
- **Issue:** Streaming from DeepInfra via `ChatDeepInfra` is currently
broken because the response parsing logic doesn't handle that
`tool_calls` can be `None`. (There is no GitHub issue for this problem
yet.)
- **Dependencies:** –
- **Twitter handle:** –
Keeping this here as a reminder:
> If no one reviews your PR within a few days, please @-mention one of
baskaryan, efriis, eyurtsev, ccurme, vbarda, hwchase17.
### Description
adds an init method to ChatDeepInfra to set the model_name attribute
accordings to the argument
### Issue
currently, the model_name specified by the user during initialization of
the ChatDeepInfra class is never set. Therefore, it always chooses the
default model (meta-llama/Llama-2-70b-chat-hf, however probably since
this is deprecated it always uses meta-llama/Llama-3-70b-Instruct). We
stumbled across this issue and fixed it as proposed in this pull
request. Feel free to change the fix according to your coding guidelines
and style, this is just a proposal and we want to draw attention to this
problem.
### Dependencies
no additional dependencies required
Feel free to contact me or @timo282 and @finitearth if you have any
questions.
---------
Co-authored-by: Bagatur <baskaryan@gmail.com>
Co-authored-by: Bagatur <22008038+baskaryan@users.noreply.github.com>
Description: DeepInfra 500 errors have useful information in the text
field that isn't being exposed to the user. I updated the error message
to fix this.
As an example, this code
```
from langchain_community.chat_models import ChatDeepInfra
from langchain_core.messages import HumanMessage
model = "meta-llama/Meta-Llama-3-70B-Instruct"
deepinfra_api_token = "..."
model = ChatDeepInfra(model=model, deepinfra_api_token=deepinfra_api_token)
messages = [HumanMessage("All work and no play makes Jack a dull boy\n" * 9000)]
response = model.invoke(messages)
```
Currently gives this error:
```
langchain_community.chat_models.deepinfra.ChatDeepInfraException: DeepInfra Server: Error 500
```
This change would give the following error:
```
langchain_community.chat_models.deepinfra.ChatDeepInfraException: DeepInfra Server error status 500: {"error":{"message":"Requested input length 99009 exceeds maximum input length 8192"}}
```
- Description: callback on_llm_new_token before yield chunk for
_stream/_astream for some chat models, make all chat models in a
consistent behaviour.
- Issue: N/A
- Dependencies: N/A
- Description: Add missing chunk parameter for _stream/_astream for some
chat models, make all chat models in a consistent behaviour.
- Issue: N/A
- Dependencies: N/A
