The huggingface pipeline in langchain (used for locally hosted models)
does not support batching. If you send in a batch of prompts, it just
processes them serially using the base implementation of _generate:
https://github.com/docugami/langchain/blob/master/libs/langchain/langchain/llms/base.py#L1004C2-L1004C29
This PR adds support for batching in this pipeline, so that GPUs can be
fully saturated. I updated the accompanying notebook to show GPU batch
inference.
---------
Co-authored-by: Taqi Jaffri <tjaffri@docugami.com>
<!-- Thank you for contributing to LangChain!
Replace this entire comment with:
- **Description:** a description of the change,
- **Issue:** the issue # it fixes (if applicable),
- **Dependencies:** any dependencies required for this change,
- **Tag maintainer:** for a quicker response, tag the relevant
maintainer (see below),
- **Twitter handle:** we announce bigger features on Twitter. If your PR
gets announced, and you'd like a mention, we'll gladly shout you out!
Please make sure your PR is passing linting and testing before
submitting. Run `make format`, `make lint` and `make test` to check this
locally.
See contribution guidelines for more information on how to write/run
tests, lint, etc:
https://github.com/hwchase17/langchain/blob/master/.github/CONTRIBUTING.md
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/extras`
directory.
If no one reviews your PR within a few days, please @-mention one of
@baskaryan, @eyurtsev, @hwchase17.
-->
Closes#8842
<!-- Thank you for contributing to LangChain!
Replace this entire comment with:
- Description: a description of the change,
- Issue: the issue # it fixes (if applicable),
- Dependencies: any dependencies required for this change,
- Tag maintainer: for a quicker response, tag the relevant maintainer
(see below),
- Twitter handle: we announce bigger features on Twitter. If your PR
gets announced and you'd like a mention, we'll gladly shout you out!
Please make sure your PR is passing linting and testing before
submitting. Run `make format`, `make lint` and `make test` to check this
locally.
See contribution guidelines for more information on how to write/run
tests, lint, etc:
https://github.com/hwchase17/langchain/blob/master/.github/CONTRIBUTING.md
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. These live is docs/extras
directory.
If no one reviews your PR within a few days, please @-mention one of
@baskaryan, @eyurtsev, @hwchase17, @rlancemartin.
-->
- Description: fix `ChatMessageChunk` concat error
- Issue: #10173
- Dependencies: None
- Tag maintainer: @baskaryan, @eyurtsev, @rlancemartin
- Twitter handle: None
---------
Co-authored-by: wangshuai.scotty <wangshuai.scotty@bytedance.com>
Co-authored-by: Nuno Campos <nuno@boringbits.io>
This PR aims at showcasing how to use vLLM's OpenAI-compatible chat API.
### Context
Lanchain already supports vLLM and its OpenAI-compatible `Completion`
API. However, the `ChatCompletion` API was not aligned with OpenAI and
for this reason I've waited for this
[PR](https://github.com/vllm-project/vllm/pull/852) to be merged before
adding this notebook to langchain.
### Description
This PR makes the following changes to OpenSearch:
1. Pass optional ids with `from_texts`
2. Pass an optional index name with `add_texts` and `search` instead of
using the same index name that was used during `from_texts`
### Issue
https://github.com/langchain-ai/langchain/issues/10967
### Maintainers
@rlancemartin, @eyurtsev, @navneet1v
Signed-off-by: Naveen Tatikonda <navtat@amazon.com>
LLMRails Embedding Integration
This PR provides integration with LLMRails. Implemented here are:
langchain/embeddings/llm_rails.py
docs/extras/integrations/text_embedding/llm_rails.ipynb
Hi @hwchase17 after adding our vectorstore integration to langchain with
confirmation of you and @baskaryan, now we want to add our embedding
integration
---------
Co-authored-by: Anar Aliyev <aaliyev@mgmt.cloudnet.services>
Co-authored-by: Bagatur <baskaryan@gmail.com>
Adds support for gradient.ai's embedding model.
This will remain a Draft, as the code will likely be refactored with the
`pip install gradientai` python sdk.
<!-- Thank you for contributing to LangChain!
Replace this entire comment with:
- **Description:** a description of the change,
- **Issue:** the issue # it fixes (if applicable),
- **Dependencies:** any dependencies required for this change,
- **Tag maintainer:** for a quicker response, tag the relevant
maintainer (see below),
- **Twitter handle:** we announce bigger features on Twitter. If your PR
gets announced, and you'd like a mention, we'll gladly shout you out!
Please make sure your PR is passing linting and testing before
submitting. Run `make format`, `make lint` and `make test` to check this
locally.
See contribution guidelines for more information on how to write/run
tests, lint, etc:
https://github.com/hwchase17/langchain/blob/master/.github/CONTRIBUTING.md
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/extras`
directory.
If no one reviews your PR within a few days, please @-mention one of
@baskaryan, @eyurtsev, @hwchase17.
-->
<!-- Thank you for contributing to LangChain!
Replace this entire comment with:
- **Description:** a description of the change,
- **Issue:** the issue # it fixes (if applicable),
- **Dependencies:** any dependencies required for this change,
- **Tag maintainer:** for a quicker response, tag the relevant
maintainer (see below),
- **Twitter handle:** we announce bigger features on Twitter. If your PR
gets announced, and you'd like a mention, we'll gladly shout you out!
Please make sure your PR is passing linting and testing before
submitting. Run `make format`, `make lint` and `make test` to check this
locally.
See contribution guidelines for more information on how to write/run
tests, lint, etc:
https://github.com/hwchase17/langchain/blob/master/.github/CONTRIBUTING.md
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/extras`
directory.
If no one reviews your PR within a few days, please @-mention one of
@baskaryan, @eyurtsev, @hwchase17.
-->
- **Description:** a fix for `index`.
- **Issue:** Not applicable.
- **Dependencies:** None
- **Tag maintainer:**
- **Twitter handle:** richarddwang
# Problem
Replication code
```python
from pprint import pprint
from langchain.embeddings import OpenAIEmbeddings
from langchain.indexes import SQLRecordManager, index
from langchain.schema import Document
from langchain.vectorstores import Qdrant
from langchain_setup.qdrant import pprint_qdrant_documents, create_inmemory_empty_qdrant
# Documents
metadata1 = {"source": "fullhell.alchemist"}
doc1_1 = Document(page_content="1-1 I have a dog~", metadata=metadata1)
doc1_2 = Document(page_content="1-2 I have a daugter~", metadata=metadata1)
doc1_3 = Document(page_content="1-3 Ahh! O..Oniichan", metadata=metadata1)
doc2 = Document(page_content="2 Lancer died again.", metadata={"source": "fate.docx"})
# Create empty vectorstore
collection_name = "secret_of_D_disk"
vectorstore: Qdrant = create_inmemory_empty_qdrant()
# Create record Manager
import tempfile
from pathlib import Path
record_manager = SQLRecordManager(
namespace="qdrant/{collection_name}",
db_url=f"sqlite:///{Path(tempfile.gettempdir())/collection_name}.sql",
)
record_manager.create_schema() # 必須
sync_result = index(
[doc1_1, doc1_2, doc1_2, doc2],
record_manager,
vectorstore,
cleanup="full",
source_id_key="source",
)
print(sync_result, end="\n\n")
pprint_qdrant_documents(vectorstore)
```
<details>
<summary>Code of helper functions `pprint_qdrant_documents` and
`create_inmemory_empty_qdrant`</summary>
```python
def create_inmemory_empty_qdrant(**from_texts_kwargs):
# Qdrant requires vector size, which can be only know after applying embedder
vectorstore = Qdrant.from_texts(["dummy"], location=":memory:", embedding=OpenAIEmbeddings(), **from_texts_kwargs)
dummy_document_id = vectorstore.client.scroll(vectorstore.collection_name)[0][0].id
vectorstore.delete([dummy_document_id])
return vectorstore
def pprint_qdrant_documents(vectorstore, limit: int = 100, **scroll_kwargs):
document_ids, documents = [], []
for record in vectorstore.client.scroll(
vectorstore.collection_name, limit=100, **scroll_kwargs
)[0]:
document_ids.append(record.id)
documents.append(
Document(
page_content=record.payload["page_content"],
metadata=record.payload["metadata"] or {},
)
)
pprint_documents(documents, document_ids=document_ids)
def pprint_document(document: Document = None, document_id=None, return_string=False):
displayed_text = ""
if document_id:
displayed_text += f"Document {document_id}:\n\n"
displayed_text += f"{document.page_content}\n\n"
metadata_text = pformat(document.metadata, indent=1)
if "\n" in metadata_text:
displayed_text += f"Metadata:\n{metadata_text}"
else:
displayed_text += f"Metadata:{metadata_text}"
if return_string:
return displayed_text
else:
print(displayed_text)
def pprint_documents(documents, document_ids=None):
if not document_ids:
document_ids = [i + 1 for i in range(len(documents))]
displayed_texts = []
for document_id, document in zip(document_ids, documents):
displayed_text = pprint_document(
document_id=document_id, document=document, return_string=True
)
displayed_texts.append(displayed_text)
print(f"\n{'-' * 100}\n".join(displayed_texts))
```
</details>
You will get
```
{'num_added': 3, 'num_updated': 0, 'num_skipped': 0, 'num_deleted': 0}
Document 1b19816e-b802-53c0-ad60-5ff9d9b9b911:
1-2 I have a daugter~
Metadata:{'source': 'fullhell.alchemist'}
----------------------------------------------------------------------------------------------------
Document 3362f9bc-991a-5dd5-b465-c564786ce19c:
1-1 I have a dog~
Metadata:{'source': 'fullhell.alchemist'}
----------------------------------------------------------------------------------------------------
Document a4d50169-2fda-5339-a196-249b5f54a0de:
1-2 I have a daugter~
Metadata:{'source': 'fullhell.alchemist'}
```
This is not correct. We should be able to expect that the vectorsotre
now includes doc1_1, doc1_2, and doc2, but not doc1_1, doc1_2, and
doc1_2.
# Reason
In `index`, the original code is
```python
uids = []
docs_to_index = []
for doc, hashed_doc, doc_exists in zip(doc_batch, hashed_docs, exists_batch):
if doc_exists:
# Must be updated to refresh timestamp.
record_manager.update([hashed_doc.uid], time_at_least=index_start_dt)
num_skipped += 1
continue
uids.append(hashed_doc.uid)
docs_to_index.append(doc)
```
In the aforementioned example, `len(doc_batch) == 4`, but
`len(hashed_docs) == len(exists_batch) == 3`. This is because the
deduplication of input documents [doc1_1, doc1_2, doc1_2, doc2] is
[doc1_1, doc1_2, doc2]. So `index` insert doc1_1, doc1_2, doc1_2 with
the uid of doc1_1, doc1_2, doc2.
---------
Co-authored-by: Eugene Yurtsev <eyurtsev@gmail.com>
This PR makes `ChatAnthropic.anthropic_api_key` a `pydantic.SecretStr`
to avoid inadvertently exposing API keys when the `ChatAnthropic` object
is represented as a str.
**Description**
Fixes broken link to `CONTRIBUTING.md` in `libs/langchain/README.md`.
Because`libs/langchain/README.md` was copied from the top level README,
and because the README contains a link to `.github/CONTRIBUTING.md`, the
copied README's link relative path must be updated. This commit fixes
that link.
**Description:**
Default refine template does not actually use the refine template
defined above, it uses a string with the variable name.
@baskaryan, @eyurtsev, @hwchase17
- chat vertex async
- vertex stream
- vertex full generation info
- vertex use server-side stopping
- model garden async
- update docs for all the above
in follow up will add
[] chat vertex full generation info
[] chat vertex retries
[] scheduled tests
- Description:
Updated JSONLoader usage documentation which was making it unusable
- Issue: JSONLoader if used with the documented arguments was failing on
various JSON documents.
- Dependencies:
no dependencies
- Twitter handle: @TheSlnArchitect
This adds a section on usage of `CassandraCache` and
`CassandraSemanticCache` to the doc notebook about caching LLMs, as
suggested in [this
comment](https://github.com/langchain-ai/langchain/pull/9772/#issuecomment-1710544100)
on a previous merged PR.
I also spotted what looks like a mismatch between different executions
and propose a fix (line 98).
Being the result of several runs, the cell execution numbers are
scrambled somewhat, so I volunteer to refine this PR by (manually)
re-numbering the cells to restore the appearance of a single, smooth
running (for the sake of orderly execution :)
**Description:**
This commit adds a vector store for the Postgres-based vector database
(`TimescaleVector`).
Timescale Vector(https://www.timescale.com/ai) is PostgreSQL++ for AI
applications. It enables you to efficiently store and query billions of
vector embeddings in `PostgreSQL`:
- Enhances `pgvector` with faster and more accurate similarity search on
1B+ vectors via DiskANN inspired indexing algorithm.
- Enables fast time-based vector search via automatic time-based
partitioning and indexing.
- Provides a familiar SQL interface for querying vector embeddings and
relational data.
Timescale Vector scales with you from POC to production:
- Simplifies operations by enabling you to store relational metadata,
vector embeddings, and time-series data in a single database.
- Benefits from rock-solid PostgreSQL foundation with enterprise-grade
feature liked streaming backups and replication, high-availability and
row-level security.
- Enables a worry-free experience with enterprise-grade security and
compliance.
Timescale Vector is available on Timescale, the cloud PostgreSQL
platform. (There is no self-hosted version at this time.) LangChain
users get a 90-day free trial for Timescale Vector.
---------
Co-authored-by: Bagatur <baskaryan@gmail.com>
Co-authored-by: Avthar Sewrathan <avthar@timescale.com>
- **Description:** This PR implements a new LLM API to
https://gradient.ai
- **Issue:** Feature request for LLM #10745
- **Dependencies**: No additional dependencies are introduced.
- **Tag maintainer:** I am opening this PR for visibility, once ready
for review I'll tag.
- ```make format && make lint && make test``` is running.
- added a `integration` and `mock unit` test.
Co-authored-by: michaelfeil <me@michaelfeil.eu>
Co-authored-by: Bagatur <baskaryan@gmail.com>
