In collaboration with @rlouf I build an
[outlines](https://dottxt-ai.github.io/outlines/latest/) integration for
langchain!
I think this is really useful for doing any type of structured output
locally.
[Dottxt](https://dottxt.co) spend alot of work optimising this process
at a lower level
([outlines-core](https://pypi.org/project/outlines-core/0.1.14/) written
in rust) so I think this is a better alternative over all current
approaches in langchain to do structured output.
It also implements the `.with_structured_output` method so it should be
a drop in replacement for a lot of applications.
The integration includes:
- **Outlines LLM class**
- **ChatOutlines class**
- **Tutorial Cookbooks**
- **Documentation Page**
- **Validation and error messages**
- **Exposes Outlines Structured output features**
- **Support for multiple backends**
- **Integration and Unit Tests**
Dependencies: `outlines` + additional (depending on backend used)
I am not sure if the unit-tests comply with all requirements, if not I
suggest to just remove them since I don't see a useful way to do it
differently.
### Quick overview:
Chat Models:
<img width="698" alt="image"
src="https://github.com/user-attachments/assets/05a499b9-858c-4397-a9ff-165c2b3e7acc">
Structured Output:
<img width="955" alt="image"
src="https://github.com/user-attachments/assets/b9fcac11-d3e5-4698-b1ae-8c4cb3d54c45">
---------
Co-authored-by: Vadym Barda <vadym@langchain.dev>
## Description
This PR adds support for Memcached as a usable LLM model cache by adding
the ```MemcachedCache``` implementation relying on the
[pymemcache](https://github.com/pinterest/pymemcache) client.
Unit test-wise, the new integration is generally covered under existing
import testing. All new functionality depends on pymemcache if
instantiated and used, so to comply with the other cache implementations
the PR also adds optional integration tests for ```MemcachedCache```.
Since this is a new integration, documentation is added for Memcached as
an integration and as an LLM Cache.
## Issue
This PR closes#27275 which was originally raised as a discussion in
#27035
## Dependencies
There are no new required dependencies for langchain, but
[pymemcache](https://github.com/pinterest/pymemcache) is required to
instantiate the new ```MemcachedCache```.
## Example Usage
```python3
from langchain.globals import set_llm_cache
from langchain_openai import OpenAI
from langchain_community.cache import MemcachedCache
from pymemcache.client.base import Client
llm = OpenAI(model="gpt-3.5-turbo-instruct", n=2, best_of=2)
set_llm_cache(MemcachedCache(Client('localhost')))
# The first time, it is not yet in cache, so it should take longer
llm.invoke("Which city is the most crowded city in the USA?")
# The second time it is, so it goes faster
llm.invoke("Which city is the most crowded city in the USA?")
```
---------
Co-authored-by: Erick Friis <erick@langchain.dev>
Description:
This fixes an issue that mistakenly created in
https://github.com/langchain-ai/langchain/pull/27253. The issue
currently exists only in `langchain-community==0.3.4`.
Test cases were added to prevent this issue in the future.
Co-authored-by: Erick Friis <erick@langchain.dev>
**Description:**
- Fix bug in Replicate LLM class, where it was looking for parameter
names in a place where they no longer exist in pydantic 2, resulting in
the "Field required" validation error described in the issue.
- Fix Replicate LLM integration tests to:
- Use active models on Replicate.
- Use the correct model parameter `max_new_tokens` as shown in the
[Replicate
docs](https://replicate.com/docs/guides/language-models/how-to-use#minimum-and-maximum-new-tokens).
- Use callbacks instead of deprecated callback_manager.
**Issue:** #26937
**Dependencies:** n/a
**Twitter handle:** n/a
---------
Signed-off-by: Fayvor Love <fayvor@gmail.com>
Co-authored-by: Chester Curme <chester.curme@gmail.com>
Reopened as a personal repo outside the organization.
## Description
- Naver HyperCLOVA X community package
- Add chat model & embeddings
- Add unit test & integration test
- Add chat model & embeddings docs
- I changed partner
package(https://github.com/langchain-ai/langchain/pull/24252) to
community package on this PR
- Could this
embeddings(https://github.com/langchain-ai/langchain/pull/21890) be
deprecated? We are trying to replace it with embedding
model(**ClovaXEmbeddings**) in this PR.
Twitter handle: None. (if needed, contact with
joonha.jeon@navercorp.com)
---
you can check our previous discussion below:
> one question on namespaces - would it make sense to have these in
.clova namespaces instead of .naver?
I would like to keep it as is, unless it is essential to unify the
package name.
(ClovaX is a branding for the model, and I plan to add other models and
components. They need to be managed as separate classes.)
> also, could you clarify the difference between ClovaEmbeddings and
ClovaXEmbeddings?
There are 3 models that are being serviced by embedding, and all are
supported in the current PR. In addition, all the functionality of CLOVA
Studio that serves actual models, such as distinguishing between test
apps and service apps, is supported. The existing PR does not support
this content because it is hard-coded.
---------
Co-authored-by: Erick Friis <erick@langchain.dev>
Co-authored-by: Vadym Barda <vadym@langchain.dev>
**Description:**
This PR updates `CassandraGraphVectorStore` to be based off
`CassandraVectorStore`, instead of using a custom CQL implementation.
This allows users using a `CassandraVectorStore` to upgrade to a
`GraphVectorStore` without having to change their database schema or
re-embed documents.
This PR also updates the documentation of the `GraphVectorStore` base
class and contains native async implementations for the standard graph
methods: `traversal_search` and `mmr_traversal_search` in
`CassandraVectorStore`.
**Issue:** No issue number.
**Dependencies:** https://github.com/langchain-ai/langchain/pull/27078
(already-merged)
**Lint and test**:
- Lint and tests all pass, including existing
`CassandraGraphVectorStore` tests.
- Also added numerous additional tests based of the tests in
`langchain-astradb` which cover many more scenarios than the existing
tests for `Cassandra` and `CassandraGraphVectorStore`
** BREAKING CHANGE**
Note that this is a breaking change for existing users of
`CassandraGraphVectorStore`. They will need to wipe their database table
and restart.
However:
- The interfaces have not changed. Just the underlying storage
mechanism.
- Any one using `langchain_community.vectorstores.Cassandra` can instead
use `langchain_community.graph_vectorstores.CassandraGraphVectorStore`
and they will gain Graph capabilities without having to re-embed their
existing documents. This is the primary goal of this PR.
---------
Co-authored-by: Erick Friis <erick@langchain.dev>
**Description**:
this PR enable VectorStore TLS and authentication (digest, basic) with
HTTP/2 for Infinispan server.
Based on httpx.
Added docker-compose facilities for testing
Added documentation
**Dependencies:**
requires `pip install httpx[http2]` if HTTP2 is needed
**Twitter handle:**
https://twitter.com/infinispan
**Description:** this PR adds a set of methods to deal with metadata
associated to the vector store entries. These, while essential to the
Graph-related extension of the `Cassandra` vector store, are also useful
in themselves. These are (all come in their sync+async versions):
- `[a]delete_by_metadata_filter`
- `[a]replace_metadata`
- `[a]get_by_document_id`
- `[a]metadata_search`
Additionally, a `[a]similarity_search_with_embedding_id_by_vector`
method is introduced to better serve the store's internal working (esp.
related to reranking logic).
**Issue:** no issue number, but now all Document's returned bear their
`.id` consistently (as a consequence of a slight refactoring in how the
raw entries read from DB are made back into `Document` instances).
**Dependencies:** (no new deps: packaging comes through langchain-core
already; `cassio` is now required to be version 0.1.10+)
**Add tests and docs**
Added integration tests for the relevant newly-introduced methods.
(Docs will be updated in a separate PR).
**Lint and test** Lint and (updated) test all pass.
---------
Co-authored-by: Erick Friis <erick@langchain.dev>
**Description**:
Adds a vector store integration with
[sqlite-vec](https://alexgarcia.xyz/sqlite-vec/), the successor to
sqlite-vss that is a single C file with no external dependencies.
Pretty straightforward, just copy-pasted the sqlite-vss integration and
made a few tweaks and added integration tests. Only question is whether
all documentation should be directed away from sqlite-vss if it is
defacto deprecated (cc @asg017).
---------
Co-authored-by: Erick Friis <erick@langchain.dev>
Co-authored-by: philippe-oger <philippe.oger@adevinta.com>
Description:
- Add system templates and user templates in integration testing
- initialize the response id field value to request_id
- Adjust the default model to hunyuan-pro
- Remove the default values of Temperature and TopP
- Add SystemMessage
all the integration tests have passed.
1、Execute integration tests for the first time
<img width="1359" alt="71ca77a2-e9be-4af6-acdc-4d665002bd9b"
src="https://github.com/user-attachments/assets/9298dc3a-aa26-4bfa-968b-c011a4e699c9">
2、Run the integration test a second time
<img width="1501" alt="image"
src="https://github.com/user-attachments/assets/61335416-4a67-4840-bb89-090ba668e237">
Issue: None
Dependencies: None
Twitter handle: None
---------
Co-authored-by: Chester Curme <chester.curme@gmail.com>
**Description:** [IPEX-LLM](https://github.com/intel-analytics/ipex-llm)
is a PyTorch library for running LLM on Intel CPU and GPU (e.g., local
PC with iGPU, discrete GPU such as Arc, Flex and Max) with very low
latency. This PR adds Intel GPU support to `ipex-llm` llm integration.
**Dependencies:** `ipex-llm`
**Contribution maintainer**: @ivy-lv11 @Oscilloscope98
**tests and docs**:
- Add: langchain/docs/docs/integrations/llms/ipex_llm_gpu.ipynb
- Update: langchain/docs/docs/integrations/llms/ipex_llm_gpu.ipynb
- Update: langchain/libs/community/tests/llms/test_ipex_llm.py
---------
Co-authored-by: ivy-lv11 <zhicunlv@gmail.com>
it fixes two issues:
### YGPTs are broken #25575
```
File ....conda/lib/python3.11/site-packages/langchain_community/embeddings/yandex.py:211, in _make_request(self, texts, **kwargs)
..
--> 211 res = stub.TextEmbedding(request, metadata=self._grpc_metadata) # type: ignore[attr-defined]
AttributeError: 'YandexGPTEmbeddings' object has no attribute '_grpc_metadata'
```
My gut feeling that #23841 is the cause.
I have to drop leading underscore from `_grpc_metadata` for quickfix,
but I just don't know how to do it _pydantic_ enough.
### minor issue:
if we use `api_key`, which is not the best practice the code fails with
```
File ~/git/...../python3.11/site-packages/langchain_community/embeddings/yandex.py:119, in YandexGPTEmbeddings.validate_environment(cls, values)
...
AttributeError: 'tuple' object has no attribute 'append'
```
- Added new integration test. But it requires YGPT env available and
active account. I don't know how int tests dis\enabled in CI.
- added small unit tests with mocks. Should be fine.
---------
Co-authored-by: mikhail-khludnev <mikhail_khludnev@rntgroup.com>
Added Azure Search Access Token Authentication instead of API KEY auth.
Fixes Issue: https://github.com/langchain-ai/langchain/issues/24263
Dependencies: None
Twitter: @levalencia
@baskaryan
Could you please review? First time creating a PR that fixes some code.
---------
Co-authored-by: Erick Friis <erick@langchain.dev>
This PR introduces adjustments to ensure compatibility with the recently
released preview version of [TiDB Serverless Vector
Search](https://tidb.cloud/ai), aiming to prevent user confusion.
- TiDB Vector now supports vector indexing with cosine and l2 distance
strategies, although inner_product remains unsupported.
- Changing the distance strategy is currently not supported, so the test
cased should be adjusted.
- [x] NatbotChain: move to community, deprecate langchain version.
Update to use `prompt | llm | output_parser` instead of LLMChain.
- [x] LLMMathChain: deprecate + add langgraph replacement example to API
ref
- [x] HypotheticalDocumentEmbedder (retriever): update to use `prompt |
llm | output_parser` instead of LLMChain
- [x] FlareChain: update to use `prompt | llm | output_parser` instead
of LLMChain
- [x] ConstitutionalChain: deprecate + add langgraph replacement example
to API ref
- [x] LLMChainExtractor (document compressor): update to use `prompt |
llm | output_parser` instead of LLMChain
- [x] LLMChainFilter (document compressor): update to use `prompt | llm
| output_parser` instead of LLMChain
- [x] RePhraseQueryRetriever (retriever): update to use `prompt | llm |
output_parser` instead of LLMChain
- **Description:** Standardize SparkLLM, include:
- docs, the issue #24803
- to support stream
- update api url
- model init arg names, the issue #20085
- In the in ` embedding-3 ` and later models of Zhipu AI, it is
supported to specify the dimensions parameter of Embedding. Ref:
https://bigmodel.cn/dev/api#text_embedding-3 .
- Add test case for `embedding-3` model by assigning dimensions.
Change all usages of __fields__ with get_fields adapter merged into
langchain_core.
Code mod generated using the following grit pattern:
```
engine marzano(0.1)
language python
`$X.__fields__` => `get_fields($X)` where {
add_import(source="langchain_core.utils.pydantic", name="get_fields")
}
```
- description: I remove the limitation of mandatory existence of
`QIANFAN_AK` and default model name which langchain uses cause there is
already a default model nama underlying `qianfan` SDK powering langchain
component.
---------
Co-authored-by: Chester Curme <chester.curme@gmail.com>
This PR adds annotations in comunity package.
Annotations are only strictly needed in subclasses of BaseModel for
pydantic 2 compatibility.
This PR adds some unnecessary annotations, but they're not bad to have
regardless for documentation pages.
Description: The old method will be discontinued; use the official SDK
for more model options.
Issue: None
Dependencies: None
Twitter handle: None
Co-authored-by: trumanyan <trumanyan@tencent.com>
## Description
This PR:
- Fixes the validation error in `FastEmbedEmbeddings`.
- Adds support for `batch_size`, `parallel` params.
- Removes support for very old FastEmbed versions.
- Updates the FastEmbed doc with the new params.
Associated Issues:
- Resolves#24039
- Resolves #https://github.com/qdrant/fastembed/issues/296
