fix(chat_models): add bind_tools impl for MoonshotChat
#397
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
mdrxy
changed the title
bind_tools impl for MoonshotChat
…streaming output is enabled. (langchain-ai#111) In the latest Alibaba Cloud Qwen models, such as Qwen-Plus-latest and Qwen3-235B-A22B, these models must use incremental_output for streaming output in reasoning mode. However, there will be issues when using parallel_tool_calls at the same time. The sample code is as follows: ```python import asyncio from langchain_community.chat_models import ChatTongyi from langchain_core.tools import tool @tool def get_weather(city: str): """Get the weather of a city""" return f"{city} is sunny" model = ChatTongyi( model="qwen-plus-latest", streaming=True, model_kwargs={"incremental_output": True, "enable_thinking": True}, ) # pyright:ignore model = model.bind_tools([get_weather]) print( model.invoke( "Check the weather in San Francisco and New York", parallel_tool_calls=True ) ) ```  The following output will be generated, showing that there is a problem with tool_calls. ```base tool_calls=[{'name': 'get_weatherget_weather', 'args': {'city': 'San Francisco'}, 'id': 'call_4835d51c1d444b0886ade5call_5a2c75ffc0ca4beb9a9eef', 'type': 'tool_call'}] ``` It should originally contain two tools' JSON schemas, but there is only one, and even that one is problematic. The specific reason is that the same field in the two JSON schemas has been merged. This is clearly an issue. The reason is that this source code uses the array's subscript as the index, but in the case of incremental streaming output, this index is almost always 0. That is, this code does not adapt well to incremental streaming output. However, the documentation of BaiLian Qwen already states that it will return the corresponding index, https://help.aliyun.com/zh/model-studio/deep-thinking  Therefore, the following modifications were made, resulting in this PR. ```python tool_calls.append( { "name": value["function"].get("name"), "args": value["function"].get("arguments"), "id": value.get("id"), # Tongyi does not respond with index, # use index in the list instead "index": value.get("index", index), } ) ``` After correction, it was found to be successfully parsed.  --------- Co-authored-by: Mason Daugherty <[email protected]> Co-authored-by: Mason Daugherty <[email protected]>
…#270) Co-authored-by: Mason Daugherty <[email protected]>
…angchain-ai#145) This PR enables injecting a [`WebClient`](https://tools.slack.dev/python-slack-sdk/api-docs/slack_sdk/web/client.html) instance into all tools within the Slack toolkit. This simplifies usage for agents that need to work with [user tokens](https://api.slack.com/concepts/token-types#user). --------- Co-authored-by: Mason Daugherty <[email protected]>
…langchain-ai#104) ## Description This PR enhances the `PlaywrightURLLoader` by adding configurable timeout and page load strategy options, making it more flexible for handling dynamic web pages. This addresses issue langchain-ai#103. ### Changes - Added `timeout` parameter (default: 30000ms) to control page navigation timeout - Added `wait_until` parameter to control when navigation is considered complete - Supported `wait_until` options: - `"load"` (default): wait for the "load" event - `"domcontentloaded"`: wait for the "DOMContentLoaded" event - `"networkidle"`: wait until there are no network connections for at least 500ms - `"commit"`: wait for the first network request to be sent ### Why The current implementation has a hardcoded 30-second timeout, which can be insufficient for heavy dynamic pages. This change allows users to: - Set longer timeouts for complex pages - Choose appropriate page load strategies based on their needs - Better handle dynamic content loading ### Real-World Examples This PR solves timeout issues with various types of websites: 1. Weather websites: ```python loader = PlaywrightURLLoader( urls=["https://weather.com/en-IN/weather/tenday/l/Chennai+Tamil+Nadu?canonicalCityId=251b7b4afedf19f747b425e048038eb1"], timeout=60000, # 60 second timeout wait_until="domcontentloaded" ) ``` 2. Dynamic news sites: ```python loader = PlaywrightURLLoader( urls=["https://www.reuters.com/markets/"], timeout=45000, wait_until="networkidle" ) ``` 3. E-commerce sites: ```python loader = PlaywrightURLLoader( urls=["https://www.amazon.com/dp/B08N5KWB9H"], timeout=90000, # 90 second timeout for complex product pages wait_until="load" ) ``` ### Testing - Added new test cases for both sync and async methods - Maintained backward compatibility - All existing tests pass - Tested with various real-world websites ### Related Issues Closes langchain-ai#103 --------- Co-authored-by: Parth Pathak <[email protected]> Co-authored-by: Mason Daugherty <[email protected]> Co-authored-by: Mason Daugherty <[email protected]>
…in-ai#46) In the current implementation of the method _get_child_links_recursive, the requests.get call doesn't accept verify as a parameter. This does not allow users to disable SSL certificate verification when needed. Please consider exposing the verify parameter as a configurable argument to the method, defaulting to True for safety, but allowing users to override it when necessary. https://github.com/langchain-ai/langchain-community/blob/bc87773064735e649cfd798185502e156d5e948a/libs/community/langchain_community/document_loaders/recursive_url_loader.py#L376-L377 --------- Co-authored-by: Mason Daugherty <[email protected]>
Fixes inconsistent behaviour across `add_embeddings` and `aadd_embeddings` for Azure Search. Fixes langchain-ai#417
…ai#416) # feat(vectorstores): add routing support for hybrid search ## Summary This PR adds routing parameter support for hybrid search (`HYBRID_SEARCH`) in the `similarity_search_with_score` method of the `OpenSearchVectorSearch` class. ## Changes ### Modified Files - `libs/community/langchain_community/vectorstores/opensearch_vector_search.py` ### Change Details Modified the hybrid search processing within the `similarity_search_with_score` method to add the `routing` parameter to the request parameters when `self.routing` is set. **Before:** ```python response = self.client.transport.perform_request( method="GET", url=path, body=payload ) ``` **After:** ```python request_args: Dict[str, Any] = { "method": "GET", "url": path, "body": payload, } if self.routing: request_args["params"] = {"routing": self.routing} response = self.client.transport.perform_request(**request_args) ``` ## Background OpenSearch's routing feature is used to route requests to specific shards, which helps optimize performance and improve data locality. In the existing code, routing was already supported for `APPROXIMATE_SEARCH`, `SCRIPT_SCORING_SEARCH`, and `PAINLESS_SCRIPTING_SEARCH` search types (lines 1320-1322), but it was not supported for hybrid search (`HYBRID_SEARCH`). This change ensures that routing functionality is consistently available across all search types. ## Compatibility - **Backward Compatibility**: No impact on existing APIs. When the `routing` parameter is not set, the behavior remains unchanged. - **Breaking Changes**: None
Author
|
@mdrxy have fixed lint problem |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
9 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
FIX THE BUG #396