Title: langchain-pinecone: improve test structure and async handling
Description: This PR improves the test infrastructure for the
langchain-pinecone package by:
1. Implementing LangChain's standard test patterns for embeddings
2. Adding comprehensive configuration testing
3. Improving async test coverage
4. Fixing integration test issues with namespaces and async markers
The changes make the tests more robust, maintainable, and aligned with
LangChain's testing standards while ensuring proper async behavior in
the embeddings implementation.
Key improvements:
- Added standard EmbeddingsTests implementation
- Split custom configuration tests into a separate test class
- Added proper async test coverage with pytest-asyncio
- Fixed namespace handling in vector store integration tests
- Improved test organization and documentation
Dependencies: None (uses existing test dependencies)
Tests and Documentation:
- ✅ Added standard test implementation following LangChain's patterns
- ✅ Added comprehensive unit tests for configuration and async behavior
- ✅ All tests passing locally
- No documentation changes needed (internal test improvements only)
Twitter handle: N/A
---------
Co-authored-by: Erick Friis <erick@langchain.dev>
- **Refactoring PDF loaders step 1**: "community: Refactoring PDF
loaders to standardize approaches"
- **Description:** Declare CloudBlobLoader in __init__.py. file_path is
Union[str, PurePath] anywhere
- **Twitter handle:** pprados
This is one part of a larger Pull Request (PR) that is too large to be
submitted all at once.
This specific part focuses to prepare the update of all parsers.
For more details, see [PR
28970](https://github.com/langchain-ai/langchain/pull/28970).
@eyurtsev it's the start of a PR series.
Fixes#29010
This PR updates the example for FewShotChatMessagePromptTemplate by
modifying the human input prompt to include a more descriptive and
user-friendly question format ('What is {input}?') instead of just
'{input}'. This change enhances clarity and usability in the
documentation example.
Co-authored-by: Erick Friis <erick@langchain.dev>
- [x] **PR title**: "docs: add langchain-pull-md Markdown loader"
- [x] **PR message**:
- **Description:** This PR introduces the `langchain-pull-md` package to
the LangChain community. It includes a new document loader that utilizes
the pull.md service to convert URLs into Markdown format, particularly
useful for handling web pages rendered with JavaScript frameworks like
React, Angular, or Vue.js. This loader helps in efficient and reliable
Markdown conversion directly from URLs without local rendering, reducing
server load.
- **Issue:** NA
- **Dependencies:** requests >=2.25.1
- **Twitter handle:** https://x.com/eugeneevstafev?s=21
- [x] **Add tests and docs**:
1. Added unit tests to verify URL checking and conversion
functionalities.
2. Created a comprehensive example notebook detailing the usage of the
new loader.
- [x] **Lint and test**:
- Completed local testing using `make format`, `make lint`, and `make
test` commands as per the LangChain contribution guidelines.
**Related Links:**
- [Package Repository](https://github.com/chigwell/langchain-pull-md)
- [PyPI Package](https://pypi.org/project/langchain-pull-md/)
---------
Co-authored-by: Erick Friis <erick@langchain.dev>
Add option to return content and artifacts, to also be able to access
the full info of the retrieved documents.
They are returned as a list of dicts in the `artifacts` property if
parameter `response_format` is set to `"content_and_artifact"`.
Defaults to `"content"` to keep current behavior.
---------
Co-authored-by: Erick Friis <erick@langchain.dev>
(Inspired by https://github.com/langchain-ai/langchain/issues/26918)
We rely on some deprecated public functions in the hot path for tool
binding (`convert_pydantic_to_openai_function`,
`convert_python_function_to_openai_function`, and
`format_tool_to_openai_function`). My understanding is that what is
deprecated is not the functionality they implement, but use of them in
the public API -- we expect to continue to rely on them.
Here we update these functions to be private and not deprecated. We keep
the public, deprecated functions as simple wrappers that can be safely
deleted.
The `@deprecated` wrapper adds considerable latency due to its use of
the `inspect` module. This update speeds up `bind_tools` by a factor of
~100x:
Before:
After:
---------
Co-authored-by: Erick Friis <erick@langchain.dev>
Thank you for contributing to LangChain!
- [x] **PR title**: "docs: fix typo"
- [x] **PR message**: ***Delete this entire checklist*** and replace
with
- **Description:** a minor fix of typo
- **Issue:** NA
- **Dependencies:** NA
- **Twitter handle:** NA
- [x] **Add tests and docs**: If you're adding a new integration, please
include
1. ~~a test for the integration, preferably unit tests that do not rely
on network access,~~
2. ~~an example notebook showing its use. It lives in
`docs/docs/integrations` directory.~~
- [x] **Lint and test**: Run `make format`, `make lint` and `make test`
from the root of the package(s) you've modified. See contribution
guidelines for more: https://python.langchain.com/docs/contributing/
Add a retriever to interact with Dappier APIs with an example notebook.
The retriever can be invoked with:
```python
from langchain_dappier import DappierRetriever
retriever = DappierRetriever(
data_model_id="dm_01jagy9nqaeer9hxx8z1sk1jx6",
k=5
)
retriever.invoke("latest tech news")
```
To retrieve 5 documents related to latest news in the tech sector. The
included notebook also includes deeper details about controlling filters
such as selecting a data model, number of documents to return, site
domain reference, minimum articles from the reference domain, and search
algorithm, as well as including the retriever in a chain.
The integration package can be found over here -
https://github.com/DappierAI/langchain-dappier
This commit updates the documentation and package registry for the
FalkorDB Chat Message History integration.
**Changes:**
- Added a comprehensive example notebook
falkordb_chat_message_history.ipynb demonstrating how to use FalkorDB
for session-based chat message storage.
- Added a provider notebook for FalkorDB
- Updated libs/packages.yml to register FalkorDB as an integration
package, following LangChain's new guidelines for community
integrations.
**Notes:**
- This update aligns with LangChain's process for registering new
integrations via documentation updates and package registry
modifications.
- No functional or core package changes were made in this commit.
---------
Co-authored-by: Chester Curme <chester.curme@gmail.com>
- In this PR, I have updated the AzureML Endpoint with the latest
endpoint.
- **Description:** I have changed the existing `/chat/completions` to
`/models/chat/completions` in
libs/community/langchain_community/llms/azureml_endpoint.py
- **Issue:** #25702
---------
Co-authored-by: = <=>
**Description:**
This PR updates the codebase to reflect the deprecation of the AgentType
feature. It includes the following changes:
Documentation Update:
Added a deprecation notice to the AgentType class comment.
Provided a reference to the official LangChain migration guide for
transitioning to LangGraph agents.
Reference Link: https://python.langchain.com/docs/how_to/migrate_agent/
**Twitter handle:** @hrrrriiiishhhhh
---------
Co-authored-by: Chester Curme <chester.curme@gmail.com>
## Description
To integrate ModelScope inference API endpoints for both Embeddings,
LLMs and ChatModels, install the package
`langchain-modelscope-integration` (as discussed in issue #28928 ). This
is necessary because the package name `langchain-modelscope` was already
registered by another party.
ModelScope is a premier platform designed to connect model checkpoints
with model applications. It provides the necessary infrastructure to
share open models and promote model-centric development. For more
information, visit GitHub page:
[ModelScope](https://github.com/modelscope).
- **Description:**
This PR addresses an issue with the `stop_sequences` field in the
`ChatGroq` class. Currently, the field is defined as:
```python
stop: Optional[Union[List[str], str]] = Field(None, alias="stop_sequences")
```
This causes the language server (LSP) to raise an error indicating that
the `stop_sequences` parameter must be implemented. The issue occurs
because `Field(None, alias="stop_sequences")` is different compared to
`Field(default=None, alias="stop_sequences")`.
To resolve the issue, the field is updated to:
```python
stop: Optional[Union[List[str], str]] = Field(default=None, alias="stop_sequences")
```
While this issue does not affect runtime behavior, it ensures
compatibility with LSPs and improves the development experience.
- **Issue:** N/A
- **Dependencies:** None
In the previous commit, the cached model key for this model was omitted.
When using the "gpt-4o-2024-11-20" model, the token count in the
callback appeared as 0, and the cost was recorded as 0.
We add model and cost information so that the token count and cost can
be displayed for the respective model.
- The message before modification is as follows.
```
Tokens Used: 0
Prompt Tokens: 0
Prompt Tokens Cached: 0
Completion Tokens: 0
Reasoning Tokens: 0
Successful Requests: 0
Total Cost (USD): $0.0
```
- The message after modification is as follows.
```
Tokens Used: 3783
Prompt Tokens: 3625
Prompt Tokens Cached: 2560
Completion Tokens: 158
Reasoning Tokens: 0
Successful Requests: 1
Total Cost (USD): $0.010642500000000001
```
Hi Erick. Coming back from a previous attempt, we now made a separate
package for the CrateDB adapter, called `langchain-cratedb`, as advised.
Other than registering the package within `libs/packages.yml`, this
patch includes a minimal amount of documentation to accompany the advent
of this new package. Let us know about any mistakes we made, or changes
you would like to see. Thanks, Andreas.
## About
- **Description:** Register a new database adapter package,
`langchain-cratedb`, providing traditional vector store, document
loader, and chat message history features for a start.
- **Addressed to:** @efriis, @eyurtsev
- **References:** GH-27710
- **Preview:** [Providers » More »
CrateDB](https://langchain-git-fork-crate-workbench-register-la-4bf945-langchain.vercel.app/docs/integrations/providers/cratedb/)
## Status
- **PyPI:** https://pypi.org/project/langchain-cratedb/
- **GitHub:** https://github.com/crate/langchain-cratedb
- **Documentation (CrateDB):**
https://cratedb.com/docs/guide/integrate/langchain/
- **Documentation (LangChain):** _This PR._
## Backlog?
Is this applicable for this kind of patch?
> - [ ] **Add tests and docs**: If you're adding a new integration,
please include
> 1. a test for the integration, preferably unit tests that do not rely
on network access,
> 2. an example notebook showing its use. It lives in
`docs/docs/integrations` directory.
## Q&A
1. Notebooks that use the LangChain CrateDB adapter are currently at
[CrateDB LangChain
Examples](https://github.com/crate/cratedb-examples/tree/main/topic/machine-learning/llm-langchain),
and the documentation refers to them. Because they are derived from very
old blueprints coming from LangChain 0.0.x times, we guess they need a
refresh before adding them to `docs/docs/integrations`. Is it applicable
to merge this minimal package registration + documentation patch, which
already includes valid code snippets in `cratedb.mdx`, and add
corresponding notebooks on behalf of a subsequent patch later?
2. How would it work getting into the tabular list of _Integration
Packages_ enumerated on the [documentation entrypoint page about
Providers](https://python.langchain.com/docs/integrations/providers/)?
/cc Please also review, @ckurze, @wierdvanderhaar, @kneth,
@simonprickett, if you can find the time. Thanks!
- **Description:** `embed_documents` and `embed_query` was throwing off
the error as stated in the issue. The issue was that `Llama` client is
returning the embeddings in a nested list which is not being accounted
for in the current implementation and therefore the stated error is
being raised.
- **Issue:** #28813
---------
Co-authored-by: Chester Curme <chester.curme@gmail.com>
community: optimize DataFrame document loader
**Description:**
Simplify the `lazy_load` method in the DataFrame document loader by
combining text extraction and metadata cleanup into a single operation.
This makes the code more concise while maintaining the same
functionality.
**Issue:** N/A
**Dependencies:** None
**Twitter handle:** N/A
- **Description:** The aload function, contrary to its name, is not an
asynchronous function, so it cannot work concurrently with other
asynchronous functions.
- **Issue:** #28336
- **Test: **: Done
- **Docs: **
[here](e0a95e5646/docs/docs/integrations/document_loaders/web_base.ipynb (L201))
- **Lint: ** All checks passed
If no one reviews your PR within a few days, please @-mention one of
baskaryan, efriis, eyurtsev, ccurme, vbarda, hwchase17.
---------
Co-authored-by: Chester Curme <chester.curme@gmail.com>
**Description:**
Added ability to set `prefix` attribute to prevent error :
```
httpx.HTTPStatusError: Error response 400 while fetching https://api.mistral.ai/v1/chat/completions: {"object":"error","message":"Expected last role User or Tool (or Assistant with prefix True) for serving but got assistant","type":"invalid_request_error","param":null,"code":null}
```
Co-authored-by: Sylvain DEPARTE <sylvain.departe@wizbii.com>
- *[x] **PR title**: "community: adding langchain-predictionguard
partner package documentation"
- *[x] **PR message**:
- **Description:** This PR adds documentation for the
langchain-predictionguard package to main langchain repo, along with
deprecating current Prediction Guard LLMs package. The LLMs package was
previously broken, so I also updated it one final time to allow it to
continue working from this point onward. . This enables users to chat
with LLMs through the Prediction Guard ecosystem.
- **Package Links**:
- [PyPI](https://pypi.org/project/langchain-predictionguard/)
- [Github
Repo](https://www.github.com/predictionguard/langchain-predictionguard)
- **Issue:** None
- **Dependencies:** None
- **Twitter handle:** [@predictionguard](https://x.com/predictionguard)
- *[x] **Add tests and docs**: All docs have been added for the partner
package, and the current LLMs package test was updated to reflect
changes.
- *[x] **Lint and test**: Linting tests are all passing.
---------
Co-authored-by: ccurme <chester.curme@gmail.com>
When using tools with optional parameters, the parameter `type` is not
longer available since langchain update to 0.3 (because of the pydantic
upgrade?) and there is now an `anyOf` field instead. This results in the
`type` being `None` in the chat request for the tool parameter, and the
LLM call fails with the error:
```
oci.exceptions.ServiceError: {'target_service': 'generative_ai_inference',
'status': 400, 'code': '400',
'opc-request-id': '...',
'message': 'Parameter definition must have a type.',
'operation_name': 'chat'
...
}
```
Example code that fails:
```
from langchain_community.chat_models.oci_generative_ai import ChatOCIGenAI
from langchain_core.tools import tool
from typing import Optional
llm = ChatOCIGenAI(
model_id="cohere.command-r-plus",
service_endpoint="https://inference.generativeai.us-chicago-1.oci.oraclecloud.com",
compartment_id="ocid1.compartment.oc1...",
auth_profile="your_profile",
auth_type="API_KEY",
model_kwargs={"temperature": 0, "max_tokens": 3000},
)
@tool
def test(example: Optional[str] = None):
"""This is the tool to use to test things
Args:
example: example variable, defaults to None
"""
return "this is a test"
llm_with_tools = llm.bind_tools([test])
result = llm_with_tools.invoke("can you make a test for g")
```
This PR sets the param type to `any` in that case, and fixes the
problem.
Co-authored-by: Erick Friis <erick@langchain.dev>
## Description
(This PR has contributions from @khushiDesai, @ashvini8, and
@ssumaiyaahmed).
This PR addresses **Issue #11229** which addresses the need for SQL
support in document parsing. This is integrated into the generic
TreeSitter parsing library, allowing LangChain users to easily load
codebases in SQL into smaller, manageable "documents."
This pull request adds a new ```SQLSegmenter``` class, which provides
the SQL integration.
## Issue
**Issue #11229**: Add support for a variety of languages to
LanguageParser
## Testing
We created a file ```test_sql.py``` with several tests to ensure the
```SQLSegmenter``` is functional. Below are the tests we added:
- ```def test_is_valid```: Checks SQL validity.
- ```def test_extract_functions_classes```: Extracts individual SQL
statements.
- ```def test_simplify_code```: Simplifies SQL code with comments.
---------
Co-authored-by: Syeda Sumaiya Ahmed <114104419+ssumaiyaahmed@users.noreply.github.com>
Co-authored-by: ashvini hunagund <97271381+ashvini8@users.noreply.github.com>
Co-authored-by: Khushi Desai <khushi.desai@advantawitty.com>
Co-authored-by: Khushi Desai <59741309+khushiDesai@users.noreply.github.com>
Co-authored-by: ccurme <chester.curme@gmail.com>
When using `create_xml_agent` or `create_json_chat_agent` to create a
agent, and the function corresponding to the tool is a parameterless
function, the `XMLAgentOutputParser` or `JSONAgentOutputParser` will
parse the tool input into an empty string, `BaseTool` will parse it into
a positional argument.
So, the program will crash finally because we invoke a parameterless
function but with a positional argument.Specially, below code will raise
StopIteration in
[_parse_input](https://github.com/langchain-ai/langchain/blob/master/libs/core/langchain_core/tools/base.py#L419)
```python
from langchain import hub
from langchain.agents import AgentExecutor, create_json_chat_agent, create_xml_agent
from langchain_openai import ChatOpenAI
prompt = hub.pull("hwchase17/react-chat-json")
llm = ChatOpenAI()
# agent = create_xml_agent(llm, tools, prompt)
agent = create_json_chat_agent(llm, tools, prompt)
agent_executor = AgentExecutor(agent=agent, tools=tools, verbose=True)
agent_executor.invoke(......)
```
---------
Co-authored-by: Erick Friis <erick@langchain.dev>
Co-authored-by: Bagatur <22008038+baskaryan@users.noreply.github.com>
Co-authored-by: Chester Curme <chester.curme@gmail.com>
**Description:**
With current HTML splitters, they rely on secondary use of the
`RecursiveCharacterSplitter` to further chunk the document into
manageable chunks. The issue with this is it fails to maintain important
structures such as tables, lists, etc within HTML.
This Implementation of a HTML splitter, allows the user to define a
maximum chunk size, HTML elements to preserve in full, options to
preserve `<a>` href links in the output and custom handlers.
The core splitting begins with headers, similar to `HTMLHeaderSplitter`.
If these sections exceed the length of the `max_chunk_size` further
recursive splitting is triggered. During this splitting, elements listed
to preserve, will be excluded from the splitting process. This can cause
chunks to be slightly larger then the max size, depending on preserved
length. However, all contextual relevance of the preserved item remains
intact.
**Custom Handlers**: Sometimes, companies such as Atlassian have custom
HTML elements, that are not parsed by default with `BeautifulSoup`.
Custom handlers allows a user to provide a function to be ran whenever a
specific html tag is encountered. This allows the user to preserve and
gather information within custom html tags that `bs4` will potentially
miss during extraction.
**Dependencies:** User will need to install `bs4` in their project to
utilise this class
I have also added in `how_to` and unit tests, which require `bs4` to
run, otherwise they will be skipped.
Flowchart of process:
---------
Co-authored-by: Bagatur <baskaryan@gmail.com>
Co-authored-by: Chester Curme <chester.curme@gmail.com>
Simple modification to add support for anthropic models deployed in
Google Vertex AI model garden in `init_chat_model` importing
`ChatAnthropicVertex`
- [v] **Lint and test**
Thank you for contributing to LangChain!
- [x] **PR title**: "package: description"
- "community: 1. add new parameter `default_headers` for oci model
deployments and oci chat model deployments. 2. updated k parameter in
OCIModelDeploymentLLM class."
- [x] **PR message**:
- **Description:** 1. add new parameters `default_headers` for oci model
deployments and oci chat model deployments. 2. updated k parameter in
OCIModelDeploymentLLM class.
- [x] **Add tests and docs**:
1. unit tests
2. notebook
---------
Co-authored-by: Erick Friis <erick@langchain.dev>
- Convert developer openai messages to SystemMessage
- store additional_kwargs={"__openai_role__": "developer"} so that the
correct role can be reconstructed if needed
- update ChatOpenAI to read in openai_role
---------
Co-authored-by: Erick Friis <erick@langchain.dev>
- **Description:**
This PR resolves an issue with the
`ExperimentalMarkdownSyntaxTextSplitter` class, which retains the
internal state across multiple calls to the `split_text` method. This
behaviour caused an unintended accumulation of chunks in `self`
variables, leading to incorrect outputs when processing multiple
Markdown files sequentially.
- Modified `libs\text-splitters\langchain_text_splitters\markdown.py` to
reset the relevant internal attributes at the start of each `split_text`
invocation. This ensures each call processes the input independently.
- Added unit tests in
`libs\text-splitters\tests\unit_tests\test_text_splitters.py` to verify
the fix and ensure the state does not persist across calls.
- **Issue:**
Fixes [#26440](https://github.com/langchain-ai/langchain/issues/26440).
- **Dependencies:**
No additional dependencies are introduced with this change.
- [x] Unit tests were added to verify the changes.
- [x] Updated documentation where necessary.
- [x] Ran `make format`, `make lint`, and `make test` to ensure
compliance with project standards.
---------
Co-authored-by: Angel Chen <angelchen396@gmail.com>
Co-authored-by: Chester Curme <chester.curme@gmail.com>
- **Description:** The `kwargs` was being checked as None object which
was causing the rest of code in `with_structured_output` not getting
executed. The checking part has been fixed in this PR.
- **Issue:** #28776
Thank you for contributing to LangChain!
- [x] **PR title**: Add float Message into Dynamo DB
- community
- Example: "community: Chat Message History
- [x] **PR message**:
- **Description:** pushing float values into dynamo db creates error ,
solved that by converting to str type
- **Issue:** Float values are not getting pushed
- **Twitter handle:** VpkPrasanna
Have added an utility function for str conversion , let me know where to
place it happy to do an commit.
This PR is from an discussion of #26543
@hwchase17 @baskaryan @efriis
---------
Co-authored-by: Chester Curme <chester.curme@gmail.com>
## Description
This pull request introduces the `DocumentLoaderAsParser` class, which
acts as an adapter to transform document loaders into parsers within the
LangChain framework. The class enables document loaders that accept a
`file_path` parameter to be utilized as blob parsers. This is
particularly useful for integrating various document loading
capabilities seamlessly into the LangChain ecosystem.
When merged in together with PR
https://github.com/langchain-ai/langchain/pull/27716 It opens options
for `SharePointLoader` / `OneDriveLoader` to process any filetype that
has a document loader.
### Features
- **Flexible Parsing**: The `DocumentLoaderAsParser` class can adapt any
document loader that meets the criteria of accepting a `file_path`
argument, allowing for lazy parsing of documents.
- **Compatibility**: The class has been designed to work with various
document loaders, making it versatile for different use cases.
### Usage Example
To use the `DocumentLoaderAsParser`, you would initialize it with a
suitable document loader class and any required parameters. Here’s an
example of how to do this with the `UnstructuredExcelLoader`:
```python
from langchain_community.document_loaders.blob_loaders import Blob
from langchain_community.document_loaders.parsers.documentloader_adapter import DocumentLoaderAsParser
from langchain_community.document_loaders.excel import UnstructuredExcelLoader
# Initialize the parser adapter with UnstructuredExcelLoader
xlsx_parser = DocumentLoaderAsParser(UnstructuredExcelLoader, mode="paged")
# Use parser, for ex. pass it to MimeTypeBasedParser
MimeTypeBasedParser(
handlers={
"application/vnd.openxmlformats-officedocument.spreadsheetml.sheet": xlsx_parser
}
)
```
- **Dependencies:** None
- **Twitter handle:** @martintriska1
If no one reviews your PR within a few days, please @-mention one of
baskaryan, efriis, eyurtsev, ccurme, vbarda, hwchase17.
---------
Co-authored-by: Chester Curme <chester.curme@gmail.com>
- **Description:** One-Bit Images was raising error which has been fixed
in this PR for `PDFPlumberParser`
- **Issue:** #28480
---------
Co-authored-by: Chester Curme <chester.curme@gmail.com>
- [ x ] Fix when lancedb return table without metadata column
- **Description:** Check the table schema, if not has metadata column,
init the Document with metadata argument equal to empty dict
- **Issue:** https://github.com/langchain-ai/langchain/issues/27005
- [ x ] **Add tests and docs**
---------
Co-authored-by: ccurme <chester.curme@gmail.com>
## Overview
This PR adds documentation for the `langchain-yt-dlp` package, a YouTube
document loader that uses `yt-dlp` for Youtube videos metadata
extraaction.
## Changes
- Added documentation notebook for YoutubeLoader
- Updated packages.yml to include langchain-yt-dlp
## Motivation
The existing LangChain YoutubeLoader was unable to fetch YouTube
metadata due to changes in YouTube's structure. This package resolves
those issues by leveraging the `yt-dlp` library.
## Features
- Reliable YouTube metadata extraction
## Related
- Package Repository: https://github.com/aqib0770/langchain-yt-dlp
- PyPI Package: https://pypi.org/project/langchain-yt-dlp/
---------
Co-authored-by: Chester Curme <chester.curme@gmail.com>
Hi, langchain team! I'm a maintainer of
[OceanBase](https://github.com/oceanbase/oceanbase).
With the integration guidance, I create a python lib named
[langchain-oceanbase](https://github.com/oceanbase/langchain-oceanbase)
to integrate `Oceanbase Vector Store` with `Langchain`.
So I'd like to add the required docs. I will appreciate your feedback.
Thank you!
---------
Signed-off-by: shanhaikang.shk <shanhaikang.shk@oceanbase.com>
Co-authored-by: Erick Friis <erick@langchain.dev>
- [X] **PR title**:
community: Add new model and structured output support
- [X] **PR message**:
- **Description:** add support for meta llama 3.2 image handling, and
JSON mode for structured output
- **Issue:** NA
- **Dependencies:** NA
- **Twitter handle:** NA
- [x] **Add tests and docs**:
1. we have updated our unit tests,
2. no changes required for documentation.
- [x] **Lint and test**:
make format, make lint and make test we run successfully
---------
Co-authored-by: Arthur Cheng <arthur.cheng@oracle.com>
Co-authored-by: ccurme <chester.curme@gmail.com>
**Description:**
Adding new AWS Bedrock model and their respective costs to match
https://aws.amazon.com/bedrock/pricing/ for the Bedrock callback
**Issue:**
Missing models for those that wish to try them out
**Dependencies:**
Nothing added
**Twitter handle:**
@David_Pryce and / or @JamfSoftware
If no one reviews your PR within a few days, please @-mention one of
baskaryan, efriis, eyurtsev, ccurme, vbarda, hwchase17.
**Description**:
This PR addresses an issue where the DocumentAttributeValue class
properties did not have default values of None. By explicitly setting
the Optional attributes (DateValue, LongValue, StringListValue, and
StringValue) to default to None, this change ensures the class functions
as expected when no value is provided for these attributes.
**Changes Made**:
Added default None values to the following properties of the
DocumentAttributeValue class:
DateValue
LongValue
StringListValue
StringValue
Removed the invalid argument extra="allow" from the BaseModel
inheritance.
Dependencies: None.
**Twitter handle (optional)**: @__korikori1021
**Checklist**
- [x] Verified that KendraRetriever works as expected after the changes.
Co-authored-by: y1u0d2a1i <y.kotani@raksul.com>
Thank you for contributing to LangChain!
- [x] **PR title**: "package: description"
- Where "package" is whichever of langchain, community, core, etc. is
being modified. Use "docs: ..." for purely docs changes, "infra: ..."
for CI changes.
- Example: "community: add foobar LLM"
Description:
Improved the `_parse_google_docstring` function in `langchain/core` to
support parsing multi-paragraph descriptions before the `Args:` section
while maintaining compliance with Google-style docstring guidelines.
This change ensures better handling of docstrings with detailed function
descriptions.
Issue:
Fixes#28628
Dependencies:
None.
Twitter handle:
@isatyamks
---------
Co-authored-by: Erick Friis <erick@langchain.dev>
Co-authored-by: Chester Curme <chester.curme@gmail.com>
- **Description:** I am working to address a similar issue to the one
mentioned in https://github.com/langchain-ai/langchain/pull/19499.
Specifically, there is a problem with the Webbase loader used in
open-webui, where it fails to load the proxy configuration. This PR aims
to resolve that issue.
<!--If no one reviews your PR within a few days, please @-mention one of
baskaryan, efriis, eyurtsev, ccurme, vbarda, hwchase17.-->
---------
Co-authored-by: Chester Curme <chester.curme@gmail.com>
**Description**: Some confluence instances don't support personal access
token, then cookie is a convenient way to authenticate. This PR adds
support for Confluence cookies.
**Twitter handle**: soulmachine
…ent path given.
Thank you for contributing to LangChain!
- [ ] **PR title**: "package: description"
- Where "package" is whichever of langchain, community, core,
experimental, etc. is being modified. Use "docs: ..." for purely docs
changes, "templates: ..." for template changes, "infra: ..." for CI
changes.
- Example: "community: add foobar LLM"
- [ ] **PR message**: ***Delete this entire checklist*** and replace
with
- **Description:** a description of the change
- **Issue:** the issue # it fixes, if applicable
- **Dependencies:** any dependencies required for this change
- **Twitter handle:** if your PR gets announced, and you'd like a
mention, we'll gladly shout you out!
- [ ] **Add tests and docs**: If you're adding a new integration, please
include
1. a test for the integration, preferably unit tests that do not rely on
network access,
2. an example notebook showing its use. It lives in
`docs/docs/integrations` directory.
- [ ] **Lint and test**: Run `make format`, `make lint` and `make test`
from the root of the package(s) you've modified. See contribution
guidelines for more: https://python.langchain.com/docs/contributing/
Additional guidelines:
- Make sure optional dependencies are imported within a function.
- Please do not add dependencies to pyproject.toml files (even optional
ones) unless they are required for unit tests.
- Most PRs should not touch more than one package.
- Changes should be backwards compatible.
- If you are adding something to community, do not re-import it in
langchain.
If no one reviews your PR within a few days, please @-mention one of
baskaryan, efriis, eyurtsev, ccurme, vbarda, hwchase17.
---------
Co-authored-by: Erick Friis <erick@langchain.dev>
Co-authored-by: Chester Curme <chester.curme@gmail.com>
**Description:**
- Add _concatenate_rich_text method to combine all elements in rich text
arrays
- Update load_page method to use _concatenate_rich_text for rich text
properties
- Ensure all text content is captured, including inline code and
formatted text
- Add unit tests to verify correct handling of multi-element rich text
This fix prevents truncation of content after backticks or other
formatting elements.
**Issue:**
Using Notion DB Loader, the text for `richtext` and `title` is truncated
after 1st element was loaded as Notion Loader only read the first
element.
**Dependencies:** any dependencies required for this change
None.
---------
Co-authored-by: Bagatur <22008038+baskaryan@users.noreply.github.com>
Co-authored-by: Erick Friis <erick@langchain.dev>
This PR closes#27781
# Problem
The current implementation of `NLTKTextSplitter` is using
`sent_tokenize`. However, this `sent_tokenize` doesn't handle chars
between 2 tokenized sentences... hence, this behavior throws errors when
we are using `add_start_index=True`, as described in issue #27781. In
particular:
```python
from nltk.tokenize import sent_tokenize
output1 = sent_tokenize("Innovation drives our success. Collaboration fosters creative solutions. Efficiency enhances data management.", language="english")
print(output1)
output2 = sent_tokenize("Innovation drives our success. Collaboration fosters creative solutions. Efficiency enhances data management.", language="english")
print(output2)
>>> ['Innovation drives our success.', 'Collaboration fosters creative solutions.', 'Efficiency enhances data management.']
>>> ['Innovation drives our success.', 'Collaboration fosters creative solutions.', 'Efficiency enhances data management.']
```
# Solution
With this new `use_span_tokenize` parameter, we can use NLTK to create
sentences (with `span_tokenize`), but also add extra chars to be sure
that we still can map the chunks to the original text.
---------
Co-authored-by: Erick Friis <erick@langchain.dev>
Co-authored-by: Erick Friis <erickfriis@gmail.com>
**Description:** Added support for FalkorDB Vector Store, including its
implementation, unit tests, documentation, and an example notebook. The
FalkorDB integration allows users to efficiently manage and query
embeddings in a vector database, with relevance scoring and maximal
marginal relevance search. The following components were implemented:
- Core implementation for FalkorDBVector store.
- Unit tests ensuring proper functionality and edge case coverage.
- Example notebook demonstrating an end-to-end setup, search, and
retrieval using FalkorDB.
**Twitter handle:** @tariyekorogha
---------
Co-authored-by: Erick Friis <erick@langchain.dev>
Update to OpenLLM 0.6, which we decides to make use of OpenLLM's
OpenAI-compatible endpoint. Thus, OpenLLM will now just become a thin
wrapper around OpenAI wrapper.
Signed-off-by: Aaron Pham <contact@aarnphm.xyz>
---------
Signed-off-by: Aaron Pham <contact@aarnphm.xyz>
Co-authored-by: ccurme <chester.curme@gmail.com>
## description
- I refactor `Chathunyuan` using tencentcloud sdk because I found the
original one can't work in my application
- I add `HunyuanEmbeddings` using tencentcloud sdk
- Both of them are extend the basic class of langchain. I have fully
tested them in my application
## Dependencies
- tencentcloud-sdk-python
---------
Co-authored-by: centonhuang <centonhuang@tencent.com>
Co-authored-by: Erick Friis <erick@langchain.dev>
### About:
- **Description:** the _get_files_from_directory_ method return a
string, but it's used in other methods that expect a List[str]
- **Issue:** None
- **Dependencies:** None
This pull request import a new method _list_files_ with the old logic of
_get_files_from_directory_, but it return a List[str] at the end.
The behavior of _ get_files_from_directory_ is not changed.
Thank you for contributing to LangChain!
- [ ] **PR title**: community: Add configurable `VisualFeatures` to the
`AzureAiServicesImageAnalysisTool`
- [ ] **PR message**:
- **Description:** The `AzureAiServicesImageAnalysisTool` is a good
service and utilises the Azure AI Vision package under the hood.
However, since the creation of this tool, new `VisualFeatures` have been
added to allow the user to request other image specific information to
be returned. Currently, the tool offers neither configuration of which
features should be return nor does it offer any newer feature types. The
aim of this PR is to address this and expose more of the Azure Service
in this integration.
- **Dependencies:** no new dependencies in the main class file,
azure.ai.vision.imageanalysis added to extra test dependencies file.
- [ ] **Add tests and docs**: If you're adding a new integration, please
include
1. Although no tests exist for already implemented Azure Service tools,
I've created 3 unit tests for this class that test initialisation and
credentials, local file analysis and a test for the new changes/
features option.
- [ ] **Lint and test**: All linting has passed.
---------
Co-authored-by: Erick Friis <erick@langchain.dev>
Co-authored-by: Chester Curme <chester.curme@gmail.com>
This pull request addresses the issue with authenticating Azure National
Cloud using token (RBAC) in the AzureSearch vectorstore implementation.
## Changes
- Modified the `_get_search_client` method in `azuresearch.py` to pass
`additional_search_client_options` to the `SearchIndexClient` instance.
## Implementation Details
The patch updates the `SearchIndexClient` initialization to include the
`additional_search_client_options` parameter:
```python
index_client: SearchIndexClient = SearchIndexClient(
endpoint=endpoint,
credential=credential,
user_agent=user_agent,
**additional_search_client_options
)
```
This change allows the `audience` parameter to be correctly passed when
using Azure National Cloud, fixing the authentication issues with
GovCloud & RBAC.
This patch was generated by [Ana - AI SDE](https://openana.ai/), an
AI-powered software development assistant.
This is a fix for [Issue
25823](https://github.com/langchain-ai/langchain/issues/25823)
---------
Co-authored-by: ccurme <chester.curme@gmail.com>
Thank you for contributing to LangChain!
- **PR title**: "community: Remove all other keys in ChatLiteLLM and add
api_key"
- **PR message**: Currently, no api_key are passed to LiteLLM, and
LiteLLM only takes on api_key parameter. Therefore I removed all current
`*_api_key` attributes (They are not used), and added `api_key` that is
passed to ChatLiteLLM.
- Should fix issue #27826
---------
Co-authored-by: Chester Curme <chester.curme@gmail.com>
Description:
Current AGEGraph() implementation does some custom wrapping for graph
queries. The method here is _wrap_query() as it parse the field from the
original query to add some SQL context to it.
This improves the current parsing logic to cover additional edge cases
that are added to the test coverage, basically if any Node property name
or value has the "return" literal in it will break the graph / SQL
query.
We discovered this while dealing with real world datasets, is not an
uncommon scenario and I think it needs to be covered.
**Description:**
This PR addresses the `TypeError: sequence item 0: expected str
instance, FluentValue found` error when invoking `WikidataQueryRun`. The
root cause was an incompatible version of the
`wikibase-rest-api-client`, which caused the tool to fail when handling
`FluentValue` objects instead of strings.
The current implementation only supports `wikibase-rest-api-client<0.2`,
but the latest version is `0.2.1`, where the current implementation
breaks. Additionally, the error message advises users to install the
latest version: [code
reference](https://github.com/langchain-ai/langchain/blob/master/libs/community/langchain_community/utilities/wikidata.py#L125C25-L125C32).
Therefore, this PR updates the tool to support the latest version of
`wikibase-rest-api-client`.
Key changes:
- Updated the handling of `FluentValue` objects to ensure compatibility
with the latest `wikibase-rest-api-client`.
- Removed the restriction to `wikibase-rest-api-client<0.2` and updated
to support the latest version (`0.2.1`).
**Issue:**
Fixes [#24093](https://github.com/langchain-ai/langchain/issues/24093) –
`TypeError: sequence item 0: expected str instance, FluentValue found`.
**Dependencies:**
- Upgraded `wikibase-rest-api-client` to the latest version to resolve
the issue.
---------
Co-authored-by: peiwen_zhang <peiwen_zhang@email.com>
Co-authored-by: Erick Friis <erick@langchain.dev>
Co-authored-by: Chester Curme <chester.curme@gmail.com>
- **Description:** Changed the comparator to use a wildcard query
instead of match. This modification allows for partial text matching on
analyzed fields, which improves the flexibility of the search by
performing full-text searches that aren't limited to exact matches.
- **Issue:** The previous implementation used a match query, which
performs exact matches on analyzed fields. This approach limited the
search capabilities by requiring the query terms to align with the
indexed text. The modification to use a wildcard query instead addresses
this limitation. The wildcard query allows for partial text matching,
which means the search can return results even if only a portion of the
term matches the text. This makes the search more flexible and suitable
for use cases where exact matches aren't necessary or expected, enabling
broader full-text searches across analyzed fields.
In short, the problem was that match queries were too restrictive, and
the change to wildcard queries enhances the ability to perform partial
matches.
- **Dependencies:** none
- **Twitter handle:** @Andre_Q_Pereira
---------
Co-authored-by: André Quintino <andre.quintino@tui.com>
Co-authored-by: Eugene Yurtsev <eyurtsev@gmail.com>
Co-authored-by: Chester Curme <chester.curme@gmail.com>
**Description:**
The current implementation of `DynamoDBChatMessageHistory` updates the
`History` attribute for a given chat history record by first extracting
the existing contents into memory, appending the new message, and then
using the `put_item` method to put the record back. This has the effect
of overwriting any additional attributes someone may want to include in
the record, like chat session metadata.
This PR suggests changing from using `put_item` to using `update_item`
instead which will keep any other attributes in the record untouched.
The change is backward compatible since
1. `update_item` is an "upsert" operation, creating the record if it
doesn't already exist, otherwise updating it
2. It only touches the db insert call and passes the exact same
information. The rest of the class is left untouched
**Dependencies:**
None
**Tests and docs:**
No unit tests currently exist for the `DynamoDBChatMessageHistory`
class. This PR adds the file
`libs/community/tests/unit_tests/chat_message_histories/test_dynamodb_chat_message_history.py`
to test the `add_message` and `clear` methods. I wanted to use the moto
library to mock DynamoDB calls but I could not get poetry to resolve it
so I mocked those calls myself in the test. Therefore, no test
dependencies were added.
The change was tested on a test DynamoDB table as well. The first three
images below show the current behavior. First a message is added to chat
history, then a value is inserted in the record in some other attribute,
and finally another message is added to the record, destroying the other
attribute.
The next three images show the new behavior. Once again a value is added
to an attribute other than the History attribute, but now when the
followup message is added it does not destroy that other attribute. The
History attribute itself is unaffected by this change.
The doc located at `docs/docs/integrations/memory/aws_dynamodb.ipynb`
required no changes and was tested as well.
The `FewShotSQLTool` gets some SQL query examples from a
`BaseExampleSelector` for a given question.
This is useful to provide [few-shot
examples](https://python.langchain.com/docs/how_to/sql_prompting/#few-shot-examples)
capability to an SQL agent.
Example usage:
```python
from langchain.agents.agent_toolkits.sql.prompt import SQL_PREFIX
embeddings = OpenAIEmbeddings()
example_selector = SemanticSimilarityExampleSelector.from_examples(
examples,
embeddings,
AstraDB,
k=5,
input_keys=["input"],
collection_name="lc_few_shots",
token=ASTRA_DB_APPLICATION_TOKEN,
api_endpoint=ASTRA_DB_API_ENDPOINT,
)
few_shot_sql_tool = FewShotSQLTool(
example_selector=example_selector,
description="Input to this tool is the input question, output is a few SQL query examples related to the input question. Always use this tool before checking the query with sql_db_query_checker!"
)
agent = create_sql_agent(
llm=llm,
db=db,
prefix=SQL_PREFIX + "\nYou MUST get some example queries before creating the query.",
extra_tools=[few_shot_sql_tool]
)
result = agent.invoke({"input": "How many artists are there?"})
```
---------
Co-authored-by: Chester Curme <chester.curme@gmail.com>
- **Description:** Added Support for `bind_tool` as requested in the
issue. Plus two issue in `_stream` were fixed:
- Corrected the Positional Argument Passing for `generate_step`
- Accountability if `token` returned by `generate_step` is integer.
- **Issue:** #28692
Description: Add tool calling and structured output support for
SambaStudio chat models, docs included
---------
Co-authored-by: Erick Friis <erick@langchain.dev>
**Description:**
AzureSearch vector store: create a wrapper class on
`azure.core.credentials.TokenCredential` (which is not-instantiable) to
fix Oauth usage with `azure_ad_access_token` argument
**Issue:** [the issue it
fixes](https://github.com/langchain-ai/langchain/issues/26216)
**Dependencies:** None
- [x] **Lint and test**
---------
Co-authored-by: Erick Friis <erick@langchain.dev>
**Description**: Fixed formatting start and end time
**Issue**: The old formatting resulted everytime in an timezone error
**Dependencies**: /
**Twitter handle**: /
---------
Co-authored-by: Yannick Opitz <yannick.opitz@gob.de>
Co-authored-by: Erick Friis <erick@langchain.dev>
Thank you for contributing to LangChain!
- Added [full
text](https://learn.microsoft.com/en-us/azure/cosmos-db/gen-ai/full-text-search)
and [hybrid
search](https://learn.microsoft.com/en-us/azure/cosmos-db/gen-ai/hybrid-search)
support for Azure CosmosDB NoSql Vector Store
- Added a new enum called CosmosDBQueryType which supports the following
values:
- VECTOR = "vector"
- FULL_TEXT_SEARCH = "full_text_search"
- FULL_TEXT_RANK = "full_text_rank"
- HYBRID = "hybrid"
- User now needs to provide this query_type to the similarity_search
method for the vectorStore to make the correct query api call.
- Added a couple of work arounds as for the FULL_TEXT_RANK and HYBRID
query functions we don't support parameterized queries right now. I have
added TODO's in place, and will remove these work arounds by end of
January.
- Added necessary test cases and updated the
- [x] **Add tests and docs**: If you're adding a new integration, please
include
1. a test for the integration, preferably unit tests that do not rely on
network access,
2. an example notebook showing its use. It lives in
`docs/docs/integrations` directory.
- [x] **Lint and test**: Run `make format`, `make lint` and `make test`
from the root of the package(s) you've modified. See contribution
guidelines for more: https://python.langchain.com/docs/contributing/
Additional guidelines:
- Make sure optional dependencies are imported within a function.
- Please do not add dependencies to pyproject.toml files (even optional
ones) unless they are required for unit tests.
- Most PRs should not touch more than one package.
- Changes should be backwards compatible.
- If you are adding something to community, do not re-import it in
langchain.
If no one reviews your PR within a few days, please @-mention one of
baskaryan, efriis, eyurtsev, ccurme, vbarda, hwchase17.
---------
Co-authored-by: Erick Friis <erickfriis@gmail.com>
- **Description:** `Model_Kwargs` was not being passed correctly to
`sentence_transformers.SentenceTransformer` which has been corrected
while maintaing backward compatability
- **Issue:** #28436
---------
Co-authored-by: MoosaTae <sadhis.tae@gmail.com>
Co-authored-by: Sadit Wongprayon <101176694+MoosaTae@users.noreply.github.com>
Co-authored-by: Erick Friis <erick@langchain.dev>
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:** Adds a helper that renders documents with the
GraphVectorStore metadata fields to Graphviz for visualization. This is
helpful for understanding and debugging.
---------
Co-authored-by: Erick Friis <erick@langchain.dev>
Thank you for contributing to LangChain!
**PR title**: "community: fix PDF Filter Type Error"
- **Description:** fix PDF Filter Type Error"
- **Issue:** the issue #27153 it fixes,
- **Dependencies:** no
- **Twitter handle:** if your PR gets announced, and you'd like a
mention, we'll gladly shout you out!
- [x] **Lint and test**: Run `make format`, `make lint` and `make test`
from the root of the package(s) you've modified. See contribution
guidelines for more: https://python.langchain.com/docs/contributing/
Additional guidelines:
- Make sure optional dependencies are imported within a function.
- Please do not add dependencies to pyproject.toml files (even optional
ones) unless they are required for unit tests.
- Most PRs should not touch more than one package.
- Changes should be backwards compatible.
- If you are adding something to community, do not re-import it in
langchain.
If no one reviews your PR within a few days, please @-mention one of
baskaryan, efriis, eyurtsev, ccurme, vbarda, hwchase17.
---------
Co-authored-by: Erick Friis <erick@langchain.dev>
# Description
Implements the `atransform_documents` method for
`MarkdownifyTransformer` using the `asyncio` built-in library for
concurrency.
Note that this is mainly for API completeness when working with async
frameworks rather than for performance, since the `markdownify` function
is not I/O bound because it works with `Document` objects already in
memory.
# Issue
Fixes#27865
# Dependencies
No new dependencies added, but
[`markdownify`](https://github.com/matthewwithanm/python-markdownify) is
required since this PR updates the `markdownify` integration.
# Tests and docs
- Tests added
- I did not modify the docstrings since they already described the basic
functionality, and [the API docs also already included a
description](https://python.langchain.com/api_reference/community/document_transformers/langchain_community.document_transformers.markdownify.MarkdownifyTransformer.html#langchain_community.document_transformers.markdownify.MarkdownifyTransformer.atransform_documents).
If it would be helpful, I would be happy to update the docstrings and/or
the API docs.
# Lint and test
- [x] format
- [x] lint
- [x] test
I ran formatting with `make format`, linting with `make lint`, and
confirmed that tests pass using `make test`. Note that some unit tests
pass in CI but may fail when running `make_test`. Those unit tests are:
- `test_extract_html` (and `test_extract_html_async`)
- `test_strip_tags` (and `test_strip_tags_async`)
- `test_convert_tags` (and `test_convert_tags_async`)
The reason for the difference is that there are trailing spaces when the
tests are run in the CI checks, and no trailing spaces when run with
`make test`. I ensured that the tests pass in CI, but they may fail with
`make test` due to the addition of trailing spaces.
---------
Co-authored-by: Erick Friis <erick@langchain.dev>
**Description:** This PR introduces a `model` alias for the embedding
classes that contain the attribute `model_name`, to ensure consistency
across the codebase, as suggested by a moderator in a previous PR. The
change aligns the usage of attribute names across the project (see for
example
[here](65deeddd5d/libs/partners/groq/langchain_groq/chat_models.py (L304))).
**Issue:** This PR addresses the suggestion from the review of issue
#28269.
**Dependencies:** None
---------
Co-authored-by: Eugene Yurtsev <eyurtsev@gmail.com>
Co-authored-by: Erick Friis <erick@langchain.dev>
community: add hybrid search in opensearch
# Langchain OpenSearch Hybrid Search Implementation
## Implementation of Hybrid Search:
I have taken LangChain's OpenSearch integration to the next level by
adding hybrid search capabilities. Building on the existing
OpenSearchVectorSearch class, I have implemented Hybrid Search
functionality (which combines the best of both keyword and semantic
search). This new functionality allows users to harness the power of
OpenSearch's advanced hybrid search features without leaving the
familiar LangChain ecosystem. By blending traditional text matching with
vector-based similarity, the enhanced class delivers more accurate and
contextually relevant results. It's designed to seamlessly fit into
existing LangChain workflows, making it easy for developers to upgrade
their search capabilities.
In implementing the hybrid search for OpenSearch within the LangChain
framework, I also incorporated filtering capabilities. It's important to
note that according to the OpenSearch hybrid search documentation, only
post-filtering is supported for hybrid queries. This means that the
filtering is applied after the hybrid search results are obtained,
rather than during the initial search process.
**Note:** For the implementation of hybrid search, I strictly followed
the official OpenSearch Hybrid search documentation and I took
inspiration from
https://github.com/AndreasThinks/langchain/tree/feature/opensearch_hybrid_search
Thanks Mate!
### Experiments
I conducted few experiments to verify that the hybrid search
implementation is accurate and capable of reproducing the results of
both plain keyword search and vector search.
Experiment - 1
Hybrid Search
Keyword_weight: 1, vector_weight: 0
I conducted an experiment to verify the accuracy of my hybrid search
implementation by comparing it to a plain keyword search. For this test,
I set the keyword_weight to 1 and the vector_weight to 0 in the hybrid
search, effectively giving full weightage to the keyword component. The
results from this hybrid search configuration matched those of a plain
keyword search, confirming that my implementation can accurately
reproduce keyword-only search results when needed. It's important to
note that while the results were the same, the scores differed between
the two methods. This difference is expected because the plain keyword
search in OpenSearch uses the BM25 algorithm for scoring, whereas the
hybrid search still performs both keyword and vector searches before
normalizing the scores, even when the vector component is given zero
weight. This experiment validates that my hybrid search solution
correctly handles the keyword search component and properly applies the
weighting system, demonstrating its accuracy and flexibility in
emulating different search scenarios.
Experiment - 2
Hybrid Search
keyword_weight = 0.0, vector_weight = 1.0
For experiment-2, I took the inverse approach to further validate my
hybrid search implementation. I set the keyword_weight to 0 and the
vector_weight to 1, effectively giving full weightage to the vector
search component (KNN search). I then compared these results with a pure
vector search. The outcome was consistent with my expectations: the
results from the hybrid search with these settings exactly matched those
from a standalone vector search. This confirms that my implementation
accurately reproduces vector search results when configured to do so. As
with the first experiment, I observed that while the results were
identical, the scores differed between the two methods. This difference
in scoring is expected and can be attributed to the normalization
process in hybrid search, which still considers both components even
when one is given zero weight. This experiment further validates the
accuracy and flexibility of my hybrid search solution, demonstrating its
ability to effectively emulate pure vector search when needed while
maintaining the underlying hybrid search structure.
Experiment - 3
Hybrid Search - balanced
keyword_weight = 0.5, vector_weight = 0.5
For experiment-3, I adopted a balanced approach to further evaluate the
effectiveness of my hybrid search implementation. In this test, I set
both the keyword_weight and vector_weight to 0.5, giving equal
importance to keyword-based and vector-based search components. This
configuration aims to leverage the strengths of both search methods
simultaneously. By setting both weights to 0.5, I intended to create a
scenario where the hybrid search would consider lexical matches and
semantic similarity equally. This balanced approach is often ideal for
many real-world applications, as it can capture both exact keyword
matches and contextually relevant results that might not contain the
exact search terms.
Kindly verify the notebook for the experiments conducted!
**Notebook:**
https://github.com/karthikbharadhwajKB/Langchain_OpenSearch_Hybrid_search/blob/main/Opensearch_Hybridsearch.ipynb
### Instructions to follow for Performing Hybrid Search:
**Step-1: Instantiating OpenSearchVectorSearch Class:**
```python
opensearch_vectorstore = OpenSearchVectorSearch(
index_name=os.getenv("INDEX_NAME"),
embedding_function=embedding_model,
opensearch_url=os.getenv("OPENSEARCH_URL"),
http_auth=(os.getenv("OPENSEARCH_USERNAME"),os.getenv("OPENSEARCH_PASSWORD")),
use_ssl=False,
verify_certs=False,
ssl_assert_hostname=False,
ssl_show_warn=False
)
```
**Parameters:**
1. **index_name:** The name of the OpenSearch index to use.
2. **embedding_function:** The function or model used to generate
embeddings for the documents. It's assumed that embedding_model is
defined elsewhere in the code.
3. **opensearch_url:** The URL of the OpenSearch instance.
4. **http_auth:** A tuple containing the username and password for
authentication.
5. **use_ssl:** Set to False, indicating that the connection to
OpenSearch is not using SSL/TLS encryption.
6. **verify_certs:** Set to False, which means the SSL certificates are
not being verified. This is often used in development environments but
is not recommended for production.
7. **ssl_assert_hostname:** Set to False, disabling hostname
verification in SSL certificates.
8. **ssl_show_warn:** Set to False, suppressing SSL-related warnings.
**Step-2: Configure Search Pipeline:**
To initiate hybrid search functionality, you need to configures a search
pipeline first.
**Implementation Details:**
This method configures a search pipeline in OpenSearch that:
1. Normalizes the scores from both keyword and vector searches using the
min-max technique.
2. Applies the specified weights to the normalized scores.
3. Calculates the final score using an arithmetic mean of the weighted,
normalized scores.
**Parameters:**
* **pipeline_name (str):** A unique identifier for the search pipeline.
It's recommended to use a descriptive name that indicates the weights
used for keyword and vector searches.
* **keyword_weight (float):** The weight assigned to the keyword search
component. This should be a float value between 0 and 1. In this
example, 0.3 gives 30% importance to traditional text matching.
* **vector_weight (float):** The weight assigned to the vector search
component. This should be a float value between 0 and 1. In this
example, 0.7 gives 70% importance to semantic similarity.
```python
opensearch_vectorstore.configure_search_pipelines(
pipeline_name="search_pipeline_keyword_0.3_vector_0.7",
keyword_weight=0.3,
vector_weight=0.7,
)
```
**Step-3: Performing Hybrid Search:**
After creating the search pipeline, you can perform a hybrid search
using the `similarity_search()` method (or) any methods that are
supported by `langchain`. This method combines both `keyword-based and
semantic similarity` searches on your OpenSearch index, leveraging the
strengths of both traditional information retrieval and vector embedding
techniques.
**parameters:**
* **query:** The search query string.
* **k:** The number of top results to return (in this case, 3).
* **search_type:** Set to `hybrid_search` to use both keyword and vector
search capabilities.
* **search_pipeline:** The name of the previously created search
pipeline.
```python
query = "what are the country named in our database?"
top_k = 3
pipeline_name = "search_pipeline_keyword_0.3_vector_0.7"
matched_docs = opensearch_vectorstore.similarity_search_with_score(
query=query,
k=top_k,
search_type="hybrid_search",
search_pipeline = pipeline_name
)
matched_docs
```
twitter handle: @iamkarthik98
---------
Co-authored-by: Karthik Kolluri <karthik.kolluri@eidosmedia.com>
Co-authored-by: Eugene Yurtsev <eyurtsev@gmail.com>
JSONparse, in _validate_metadata_func(), checks the consistency of the
_metadata_func() function. To do this, it invokes it and makes sure it
receives a dictionary in response. However, during the call, it does not
respect future calls, as shown on line 100. This generates errors if,
for example, the function is like this:
```python
def generate_metadata(json_node:Dict[str,Any],kwargs:Dict[str,Any]) -> Dict[str,Any]:
return {
"source": url,
"row": kwargs['seq_num'],
"question":json_node.get("question"),
}
loader = JSONLoader(
file_path=file_path,
content_key="answer",
jq_schema='.[]',
metadata_func=generate_metadata,
text_content=False)
```
To avoid this, the verification must comply with the specifications.
This patch does just that.
---------
Co-authored-by: Eugene Yurtsev <eyurtsev@gmail.com>
The delete methods in the VectorStore and DocumentIndex interfaces
return a status indicating the result. Therefore, we can assume that
their implementations don't throw exceptions but instead return a result
indicating whether the delete operations have failed. The current
implementation doesn't check the returned value, so I modified it to
throw an exception when the operation fails.
---------
Co-authored-by: Eugene Yurtsev <eyurtsev@gmail.com>
~Note that this PR is now Draft, so I didn't add change to `aindex`
function and didn't add test codes for my change.
After we have an agreement on the direction, I will add commits.~
`batch_size` is very difficult to decide because setting a large number
like >10000 will impact VectorDB and RecordManager, while setting a
small number will delete records unnecessarily, leading to redundant
work, as the `IMPORTANT` section says.
On the other hand, we can't use `full` because the loader returns just a
subset of the dataset in our use case.
I guess many people are in the same situation as us.
So, as one of the possible solutions for it, I would like to introduce a
new argument, `scoped_full_cleanup`.
This argument will be valid only when `claneup` is Full. If True, Full
cleanup deletes all documents that haven't been updated AND that are
associated with source ids that were seen during indexing. Default is
False.
This change keeps backward compatibility.
---------
Co-authored-by: Eugene Yurtsev <eugene@langchain.dev>
Co-authored-by: Eugene Yurtsev <eyurtsev@gmail.com>
