**Description**
Adding different threshold types to the semantic chunker. I’ve had much
better and predictable performance when using standard deviations
instead of percentiles.
For all the documents I’ve tried, the distribution of distances look
similar to the above: positively skewed normal distribution. All skews
I’ve seen are less than 1 so that explains why standard deviations
perform well, but I’ve included IQR if anyone wants something more
robust.
Also, using the percentile method backwards, you can declare the number
of clusters and use semantic chunking to get an ‘optimal’ splitting.
---------
Co-authored-by: Harrison Chase <hw.chase.17@gmail.com>
## Amazon Personalize support on Langchain
This PR is a successor to this PR -
https://github.com/langchain-ai/langchain/pull/13216
This PR introduces an integration with [Amazon
Personalize](https://aws.amazon.com/personalize/) to help you to
retrieve recommendations and use them in your natural language
applications. This integration provides two new components:
1. An `AmazonPersonalize` client, that provides a wrapper around the
Amazon Personalize API.
2. An `AmazonPersonalizeChain`, that provides a chain to pull in
recommendations using the client, and then generating the response in
natural language.
We have added this to langchain_experimental since there was feedback
from the previous PR about having this support in experimental rather
than the core or community extensions.
Here is some sample code to explain the usage.
```python
from langchain_experimental.recommenders import AmazonPersonalize
from langchain_experimental.recommenders import AmazonPersonalizeChain
from langchain.llms.bedrock import Bedrock
recommender_arn = "<insert_arn>"
client=AmazonPersonalize(
credentials_profile_name="default",
region_name="us-west-2",
recommender_arn=recommender_arn
)
bedrock_llm = Bedrock(
model_id="anthropic.claude-v2",
region_name="us-west-2"
)
chain = AmazonPersonalizeChain.from_llm(
llm=bedrock_llm,
client=client
)
response = chain({'user_id': '1'})
```
Reviewer: @3coins
Noticed and fixed a few typos in the SmartLLMChain default ideation and
critique prompts
---------
Co-authored-by: Harrison Chase <hw.chase.17@gmail.com>
- **Description:**
[AS-IS] When dealing with a yaml file, the extension must be .yaml.
[TO-BE] In the absence of extension length constraints in the OS, the
extension of the YAML file is yaml, but control over the yml extension
must still be made.
It's as if it's an error because it's a .jpg extension in jpeg support.
- **Issue:** -
- **Dependencies:**
no dependencies required for this change,
## Summary
This PR upgrades LangChain's Ruff configuration in preparation for
Ruff's v0.2.0 release. (The changes are compatible with Ruff v0.1.5,
which LangChain uses today.) Specifically, we're now warning when
linter-only options are specified under `[tool.ruff]` instead of
`[tool.ruff.lint]`.
---------
Co-authored-by: Erick Friis <erick@langchain.dev>
Co-authored-by: Bagatur <baskaryan@gmail.com>
As described in issue #17060, in the case in which text has only one
sentence the following function fails. Checking for that and adding a
return case fixed the issue.
```python
def split_text(self, text: str) -> List[str]:
"""Split text into multiple components."""
# Splitting the essay on '.', '?', and '!'
single_sentences_list = re.split(r"(?<=[.?!])\s+", text)
sentences = [
{"sentence": x, "index": i} for i, x in enumerate(single_sentences_list)
]
sentences = combine_sentences(sentences)
embeddings = self.embeddings.embed_documents(
[x["combined_sentence"] for x in sentences]
)
for i, sentence in enumerate(sentences):
sentence["combined_sentence_embedding"] = embeddings[i]
distances, sentences = calculate_cosine_distances(sentences)
start_index = 0
# Create a list to hold the grouped sentences
chunks = []
breakpoint_percentile_threshold = 95
breakpoint_distance_threshold = np.percentile(
distances, breakpoint_percentile_threshold
) # If you want more chunks, lower the percentile cutoff
indices_above_thresh = [
i for i, x in enumerate(distances) if x > breakpoint_distance_threshold
] # The indices of those breakpoints on your list
# Iterate through the breakpoints to slice the sentences
for index in indices_above_thresh:
# The end index is the current breakpoint
end_index = index
# Slice the sentence_dicts from the current start index to the end index
group = sentences[start_index : end_index + 1]
combined_text = " ".join([d["sentence"] for d in group])
chunks.append(combined_text)
# Update the start index for the next group
start_index = index + 1
# The last group, if any sentences remain
if start_index < len(sentences):
combined_text = " ".join([d["sentence"] for d in sentences[start_index:]])
chunks.append(combined_text)
return chunks
```
Co-authored-by: Giulio Zani <salamanderxing@Giulios-MBP.homenet.telecomitalia.it>
- **Description:** Presidio-based anonymizers are not working because
`_remove_conflicts_and_get_text_manipulation_data` was being called
without a conflict resolution strategy. This PR fixes this issue. In
addition, it removes some mutable default arguments (antipattern).
To reproduce the issue, just run the very first cell of this
[notebook](https://python.langchain.com/docs/guides/privacy/2/) from
langchain's documentation.
<!-- Thank you for contributing to LangChain!
Please title your PR "<package>: <description>", where <package> is
whichever of langchain, community, core, experimental, etc. is being
modified.
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,
- **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` from the root
of the package you've modified to check this locally.
See contribution guidelines for more information on how to write/run
tests, lint, etc: https://python.langchain.com/docs/contributing/
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.
If no one reviews your PR within a few days, please @-mention one of
@baskaryan, @eyurtsev, @hwchase17.
-->
experimental relies on `from langchain_core.runnables.config import
run_in_executor` which was introduced in core 0.1.5.
Updated pyproject dependency as well as minimum version test.
…tch]: import models from community
ran
```bash
git grep -l 'from langchain\.chat_models' | xargs -L 1 sed -i '' "s/from\ langchain\.chat_models/from\ langchain_community.chat_models/g"
git grep -l 'from langchain\.llms' | xargs -L 1 sed -i '' "s/from\ langchain\.llms/from\ langchain_community.llms/g"
git grep -l 'from langchain\.embeddings' | xargs -L 1 sed -i '' "s/from\ langchain\.embeddings/from\ langchain_community.embeddings/g"
git checkout master libs/langchain/tests/unit_tests/llms
git checkout master libs/langchain/tests/unit_tests/chat_models
git checkout master libs/langchain/tests/unit_tests/embeddings/test_imports.py
make format
cd libs/langchain; make format
cd ../experimental; make format
cd ../core; make format
```
- **Description:** This PR fixes test failures on Windows caused by path
handling differences and unescaped special characters in regex. The
failing tests are:
```
FAILED tests/unit_tests/storage/test_filesystem.py::test_yield_keys - AssertionError: assert ['key1', 'subdir\\key2'] == ['key1', 'subdir/key2']
FAILED tests/unit_tests/test_imports.py::test_importable_all - ModuleNotFoundError: No module named 'langchain_community.langchain_community\\adapters'
FAILED tests/unit_tests/tools/file_management/test_utils.py::test_get_validated_relative_path_errs_on_absolute - re.error: incomplete escape \U at position 53
FAILED tests/unit_tests/tools/file_management/test_utils.py::test_get_validated_relative_path_errs_on_parent_dir - re.error: incomplete escape \U at position 69
FAILED tests/unit_tests/tools/file_management/test_utils.py::test_get_validated_relative_path_errs_for_symlink_outside_root - re.error: incomplete escape \U at position 64
```
- **Issue:** fixes
https://github.com/langchain-ai/langchain/issues/11775 (partially)
- **Dependencies:** none
Addded missed docstrings. Fixed inconsistency in docstrings.
**Note** CC @efriis
There were PR errors on
`langchain_experimental/prompt_injection_identifier/hugging_face_identifier.py`
But, I didn't touch this file in this PR! Can it be some cache problems?
I fixed this error.
- **Description:** This is addition to [my previous
PR](https://github.com/langchain-ai/langchain/pull/13930) with
improvements to flexibility allowing different models and notebook to
use ONNX runtime for faster speed. Since the last PR, [our
model](https://huggingface.co/laiyer/deberta-v3-base-prompt-injection)
got more than 660k downloads, and with the [public
benchmark](https://huggingface.co/spaces/laiyer/prompt-injection-benchmark)
showed much fewer false-positives than the previous one from deepset.
Additionally, on the ONNX runtime, it can be running 3x faster on the
CPU, which might be handy for builders using Langchain.
**Issue:** N/A
- **Dependencies:** N/A
- **Tag maintainer:** N/A
- **Twitter handle:** `@laiyer_ai`
TIL `**` globstar doesn't work in make
Makefile changes fix that.
`__getattr__` changes allow import of all files, but raise error when
accessing anything from the module.
file deletions were corresponding libs change from #14559
**Description**
The `SmartLLMChain` was was fixed to output key "resolution".
Unfortunately, this prevents the ability to use multiple `SmartLLMChain`
in a `SequentialChain` because of colliding output keys. This change
simply gives the option the customize the output key to allow for
sequential chaining. The default behavior is the same as the current
behavior.
Now, it's possible to do the following:
```
from langchain.chat_models import ChatOpenAI
from langchain.prompts import PromptTemplate
from langchain_experimental.smart_llm import SmartLLMChain
from langchain.chains import SequentialChain
joke_prompt = PromptTemplate(
input_variables=["content"],
template="Tell me a joke about {content}.",
)
review_prompt = PromptTemplate(
input_variables=["scale", "joke"],
template="Rate the following joke from 1 to {scale}: {joke}"
)
llm = ChatOpenAI(temperature=0.9, model_name="gpt-4-32k")
joke_chain = SmartLLMChain(llm=llm, prompt=joke_prompt, output_key="joke")
review_chain = SmartLLMChain(llm=llm, prompt=review_prompt, output_key="review")
chain = SequentialChain(
chains=[joke_chain, review_chain],
input_variables=["content", "scale"],
output_variables=["review"],
verbose=True
)
response = chain.run({"content": "chickens", "scale": "10"})
print(response)
```
---------
Co-authored-by: Erick Friis <erick@langchain.dev>
<!-- 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/langchain-ai/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:** Fix#11737 issue (extra_tools option of
create_pandas_dataframe_agent is not working),
- **Issue:** #11737 ,
- **Dependencies:** no,
- **Tag maintainer:** @baskaryan, @eyurtsev, @hwchase17 I needed this
method at work, so I modified it myself and used it. There is a similar
issue(#11737) and PR(#13018) of @PyroGenesis, so I combined my code at
the original PR.
You may be busy, but it would be great help for me if you checked. Thank
you.
- **Twitter handle:** @lunara_x
If you need an .ipynb example about this, please tag me.
I will share what I am working on after removing any work-related
content.
---------
Co-authored-by: Harrison Chase <hw.chase.17@gmail.com>
See PR title.
From what I can see, `poetry` will auto-include this. Please let me know
if I am missing something here.
---------
Co-authored-by: Bagatur <baskaryan@gmail.com>
CC @baskaryan @hwchase17 @jmorganca
Having a bit of trouble importing `langchain_experimental` from a
notebook, will figure it out tomorrow
~Ah and also is blocked by #13226~
---------
Co-authored-by: Lance Martin <lance@langchain.dev>
Co-authored-by: Bagatur <baskaryan@gmail.com>
It was :
`from langchain.schema.prompts import BasePromptTemplate`
but because of the breaking change in the ns, it is now
`from langchain.schema.prompt_template import BasePromptTemplate`
This bug prevents building the API Reference for the langchain_experimental
- **Description:** Updates to `AnthropicFunctions` to be compatible with
the OpenAI `function_call` functionality.
- **Issue:** The functionality to indicate `auto`, `none` and a forced
function_call was not completely implemented in the existing code.
- **Dependencies:** None
- **Tag maintainer:** @baskaryan , and any of the other maintainers if
needed.
- **Twitter handle:** None
I have specifically tested this functionality via AWS Bedrock with the
Claude-2 and Claude-Instant models.
- **Description:** Existing model used for Prompt Injection is quite
outdated but we fine-tuned and open-source a new model based on the same
model deberta-v3-base from Microsoft -
[laiyer/deberta-v3-base-prompt-injection](https://huggingface.co/laiyer/deberta-v3-base-prompt-injection).
It supports more up-to-date injections and less prone to
false-positives.
- **Dependencies:** No
- **Tag maintainer:** -
- **Twitter handle:** @alex_yaremchuk
---------
Co-authored-by: Bagatur <baskaryan@gmail.com>
- **Description:** The experimental package needs to be compatible with
the usage of importing agents
For example, if i use `from langchain.agents import
create_pandas_dataframe_agent`, running the program will prompt the
following information:
```
Traceback (most recent call last):
File "/Users/dongwm/test/main.py", line 1, in <module>
from langchain.agents import create_pandas_dataframe_agent
File "/Users/dongwm/test/venv/lib/python3.11/site-packages/langchain/agents/__init__.py", line 87, in __getattr__
raise ImportError(
ImportError: create_pandas_dataframe_agent has been moved to langchain experimental. See https://github.com/langchain-ai/langchain/discussions/11680 for more information.
Please update your import statement from: `langchain.agents.create_pandas_dataframe_agent` to `langchain_experimental.agents.create_pandas_dataframe_agent`.
```
But when I changed to `from langchain_experimental.agents import
create_pandas_dataframe_agent`, it was actually wrong:
```python
Traceback (most recent call last):
File "/Users/dongwm/test/main.py", line 2, in <module>
from langchain_experimental.agents import create_pandas_dataframe_agent
ImportError: cannot import name 'create_pandas_dataframe_agent' from 'langchain_experimental.agents' (/Users/dongwm/test/venv/lib/python3.11/site-packages/langchain_experimental/agents/__init__.py)
```
I should use `from langchain_experimental.agents.agent_toolkits import
create_pandas_dataframe_agent`. In order to solve the problem and make
it compatible, I added additional import code to the
langchain_experimental package. Now it can be like this Used `from
langchain_experimental.agents import create_pandas_dataframe_agent`
- **Twitter handle:** [lin_bob57617](https://twitter.com/lin_bob57617)
Fix some circular deps:
- move PromptValue into top level module bc both PromptTemplates and
OutputParsers import
- move tracer context vars to `tracers.context` and import them in
functions in `callbacks.manager`
- add core import tests
<!-- 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/langchain-ai/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.
-->
## Update 2023-09-08
This PR now supports further models in addition to Lllama-2 chat models.
See [this comment](#issuecomment-1668988543) for further details. The
title of this PR has been updated accordingly.
## Original PR description
This PR adds a generic `Llama2Chat` model, a wrapper for LLMs able to
serve Llama-2 chat models (like `LlamaCPP`,
`HuggingFaceTextGenInference`, ...). It implements `BaseChatModel`,
converts a list of chat messages into the [required Llama-2 chat prompt
format](https://huggingface.co/blog/llama2#how-to-prompt-llama-2) and
forwards the formatted prompt as `str` to the wrapped `LLM`. Usage
example:
```python
# uses a locally hosted Llama2 chat model
llm = HuggingFaceTextGenInference(
inference_server_url="http://127.0.0.1:8080/",
max_new_tokens=512,
top_k=50,
temperature=0.1,
repetition_penalty=1.03,
)
# Wrap llm to support Llama2 chat prompt format.
# Resulting model is a chat model
model = Llama2Chat(llm=llm)
messages = [
SystemMessage(content="You are a helpful assistant."),
MessagesPlaceholder(variable_name="chat_history"),
HumanMessagePromptTemplate.from_template("{text}"),
]
prompt = ChatPromptTemplate.from_messages(messages)
memory = ConversationBufferMemory(memory_key="chat_history", return_messages=True)
chain = LLMChain(llm=model, prompt=prompt, memory=memory)
# use chat model in a conversation
# ...
```
Also part of this PR are tests and a demo notebook.
- Tag maintainer: @hwchase17
- Twitter handle: `@mrt1nz`
---------
Co-authored-by: Erick Friis <erick@langchain.dev>
## Description
This PR adds support for
[lm-format-enforcer](https://github.com/noamgat/lm-format-enforcer) to
LangChain.
The library is similar to jsonformer / RELLM which are supported in
Langchain, but has several advantages such as
- Batching and Beam search support
- More complete JSON Schema support
- LLM has control over whitespace, improving quality
- Better runtime performance due to only calling the LLM's generate()
function once per generate() call.
The integration is loosely based on the jsonformer integration in terms
of project structure.
## Dependencies
No compile-time dependency was added, but if `lm-format-enforcer` is not
installed, a runtime error will occur if it is trying to be used.
## Tests
Due to the integration modifying the internal parameters of the
underlying huggingface transformer LLM, it is not possible to test
without building a real LM, which requires internet access. So, similar
to the jsonformer and RELLM integrations, the testing is via the
notebook.
## Twitter Handle
[@noamgat](https://twitter.com/noamgat)
Looking forward to hearing feedback!
---------
Co-authored-by: Bagatur <baskaryan@gmail.com>
Best to review one commit at a time, since two of the commits are 100%
autogenerated changes from running `ruff format`:
- Install and use `ruff format` instead of black for code formatting.
- Output of `ruff format .` in the `langchain` package.
- Use `ruff format` in experimental package.
- Format changes in experimental package by `ruff format`.
- Manual formatting fixes to make `ruff .` pass.
This PR replaces the previous `Intent` check with the new `Prompt
Safety` check. The logic and steps to enable chain moderation via the
Amazon Comprehend service, allowing you to detect and redact PII, Toxic,
and Prompt Safety information in the LLM prompt or answer remains
unchanged.
This implementation updates the code and configuration types with
respect to `Prompt Safety`.
### Usage sample
```python
from langchain_experimental.comprehend_moderation import (BaseModerationConfig,
ModerationPromptSafetyConfig,
ModerationPiiConfig,
ModerationToxicityConfig
)
pii_config = ModerationPiiConfig(
labels=["SSN"],
redact=True,
mask_character="X"
)
toxicity_config = ModerationToxicityConfig(
threshold=0.5
)
prompt_safety_config = ModerationPromptSafetyConfig(
threshold=0.5
)
moderation_config = BaseModerationConfig(
filters=[pii_config, toxicity_config, prompt_safety_config]
)
comp_moderation_with_config = AmazonComprehendModerationChain(
moderation_config=moderation_config, #specify the configuration
client=comprehend_client, #optionally pass the Boto3 Client
verbose=True
)
template = """Question: {question}
Answer:"""
prompt = PromptTemplate(template=template, input_variables=["question"])
responses = [
"Final Answer: A credit card number looks like 1289-2321-1123-2387. A fake SSN number looks like 323-22-9980. John Doe's phone number is (999)253-9876.",
"Final Answer: This is a really shitty way of constructing a birdhouse. This is fucking insane to think that any birds would actually create their motherfucking nests here."
]
llm = FakeListLLM(responses=responses)
llm_chain = LLMChain(prompt=prompt, llm=llm)
chain = (
prompt
| comp_moderation_with_config
| {llm_chain.input_keys[0]: lambda x: x['output'] }
| llm_chain
| { "input": lambda x: x['text'] }
| comp_moderation_with_config
)
try:
response = chain.invoke({"question": "A sample SSN number looks like this 123-456-7890. Can you give me some more samples?"})
except Exception as e:
print(str(e))
else:
print(response['output'])
```
### Output
```python
> Entering new AmazonComprehendModerationChain chain...
Running AmazonComprehendModerationChain...
Running pii Validation...
Running toxicity Validation...
Running prompt safety Validation...
> Finished chain.
> Entering new AmazonComprehendModerationChain chain...
Running AmazonComprehendModerationChain...
Running pii Validation...
Running toxicity Validation...
Running prompt safety Validation...
> Finished chain.
Final Answer: A credit card number looks like 1289-2321-1123-2387. A fake SSN number looks like XXXXXXXXXXXX John Doe's phone number is (999)253-9876.
```
---------
Co-authored-by: Jha <nikjha@amazon.com>
Co-authored-by: Anjan Biswas <anjanavb@amazon.com>
Co-authored-by: Anjan Biswas <84933469+anjanvb@users.noreply.github.com>
