Stripped-down version of
[OpenAICallbackHandler](https://github.com/langchain-ai/langchain/blob/master/libs/community/langchain_community/callbacks/openai_info.py)
that just tracks `AIMessage.usage_metadata`.
```python
from langchain_core.callbacks import get_usage_metadata_callback
from langgraph.prebuilt import create_react_agent
def get_weather(location: str) -> str:
"""Get the weather at a location."""
return "It's sunny."
tools = [get_weather]
agent = create_react_agent("openai:gpt-4o-mini", tools)
with get_usage_metadata_callback() as cb:
result = await agent.ainvoke({"messages": "What's the weather in Boston?"})
print(cb.usage_metadata)
```
- Support thinking blocks in core's `convert_to_openai_messages` (pass
through instead of error)
- Ignore thinking blocks in ChatOpenAI (instead of error)
- Support Anthropic-style image blocks in ChatOpenAI
---
Standard integration tests include a `supports_anthropic_inputs`
property which is currently enabled only for tests on `ChatAnthropic`.
This test enforces compatibility with message histories of the form:
```
- system message
- human message
- AI message with tool calls specified only through `tool_use` content blocks
- human message containing `tool_result` and an additional `text` block
```
It additionally checks support for Anthropic-style image inputs if
`supports_image_inputs` is enabled.
Here we change this test, such that if you enable
`supports_anthropic_inputs`:
- You support AI messages with text and `tool_use` content blocks
- You support Anthropic-style image inputs (if `supports_image_inputs`
is enabled)
- You support thinking content blocks.
That is, we add a test case for thinking content blocks, but we also
remove the requirement of handling tool results within HumanMessages
(motivated by existing agent abstractions, which should all return
ToolMessage). We move that requirement to a ChatAnthropic-specific test.
See https://docs.astral.sh/ruff/rules/#flake8-type-checking-tc
Some fixes done for TC001,TC002 and TC003 but these rules are excluded
since they don't play well with Pydantic.
---------
Co-authored-by: Chester Curme <chester.curme@gmail.com>
Resolves https://github.com/langchain-ai/langchain/issues/29003,
https://github.com/langchain-ai/langchain/issues/27264
Related: https://github.com/langchain-ai/langchain-redis/issues/52
```python
from langchain.chat_models import init_chat_model
from langchain.globals import set_llm_cache
from langchain_community.cache import SQLiteCache
from pydantic import BaseModel
cache = SQLiteCache()
set_llm_cache(cache)
class Temperature(BaseModel):
value: int
city: str
llm = init_chat_model("openai:gpt-4o-mini")
structured_llm = llm.with_structured_output(Temperature)
```
```python
# 681 ms
response = structured_llm.invoke("What is the average temperature of Rome in May?")
```
```python
# 6.98 ms
response = structured_llm.invoke("What is the average temperature of Rome in May?")
```
See https://docs.astral.sh/ruff/rules/#flake8-annotations-ann
The interest compared to only mypy is that ruff is very fast at
detecting missing annotations.
ANN101 and ANN102 are deprecated so we ignore them
ANN401 (no Any type) ignored to be in sync with mypy config
---------
Co-authored-by: ccurme <chester.curme@gmail.com>
- **Description:** Add tests for respecting max_concurrency and
implement it for abatch_as_completed so that test passes
- **Issue:** #29425
- **Dependencies:** none
- **Twitter handle:** keenanpepper
**Description**
Currently, when parsing a partial JSON, if a string ends with the escape
character, the whole key/value is removed. For example:
```
>>> from langchain_core.utils.json import parse_partial_json
>>> my_str = '{"foo": "bar", "baz": "qux\\'
>>>
>>> parse_partial_json(my_str)
{'foo': 'bar'}
```
My expectation (and with this fix) would be for `parse_partial_json()`
to return:
```
>>> from langchain_core.utils.json import parse_partial_json
>>>
>>> my_str = '{"foo": "bar", "baz": "qux\\'
>>> parse_partial_json(my_str)
{'foo': 'bar', 'baz': 'qux'}
```
Notes:
1. It could be argued that current behavior is still desired.
2. I have experienced this issue when the streaming output from an LLM
and the chunk happens to end with `\\`
3. I haven't included tests. Will do if change is accepted.
4. This is specially troublesome when this function is used by
187131c55c/libs/core/langchain_core/output_parsers/transform.py (L111)
since what happens is that, for example, if the received sequence of
chunks are: `{"foo": "b` , `ar\\` :
Then, the result of calling `self.parse_result()` is:
```
{"foo": "b"}
```
and the second time:
```
{}
```
Co-authored-by: Erick Friis <erick@langchain.dev>
This PR uses the [blockbuster](https://github.com/cbornet/blockbuster)
library in langchain-core to detect blocking calls made in the asyncio
event loop during unit tests.
Avoiding blocking calls is hard as these can be deeply buried in the
code or made in 3rd party libraries.
Blockbuster makes it easier to detect them by raising an exception when
a call is made to a known blocking function (eg: `time.sleep`).
Adding blockbuster allowed to find a blocking call in
`aconfig_with_context` (it ends up calling `get_function_nonlocals`
which loads function code).
**Dependencies:**
- blockbuster (test)
**Twitter handle:** cbornet_
TRY004 ("use TypeError rather than ValueError") existing errors are
marked as ignore to preserve backward compatibility.
LMK if you prefer to fix some of them.
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>
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>
- 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>
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>
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>
I reported the bug 2 weeks ago here:
https://github.com/langchain-ai/langchain/issues/28447
I believe this is a critical bug for the indexer, so I submitted a PR to
revert the change and added unit tests to prevent similar bugs from
being introduced in the future.
@eyurtsev Could you check this?
Thank you for contributing to LangChain!
- [ ] **PR title**: "core: google docstring parsing fix"
- [x] **PR message**:
- **Description:** Added a solution for invalid parsing of google
docstring such as:
Args:
net_annual_income (float): The user's net annual income (in current year
dollars).
- **Issue:** Previous code would return arg = "net_annual_income
(float)" which would cause exception in
_validate_docstring_args_against_annotations
- **Dependencies:** None
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:**
- Trim functions were incorrectly deleting nodes with more than 1
outgoing/incoming edge, so an extra condition was added to check for
this directly. A unit test "test_trim_multi_edge" was written to test
this test case specifically.
- **Issue:**
- Fixes#28411
- Fixes https://github.com/langchain-ai/langgraph/issues/1676
- **Dependencies:**
- No changes were made to the dependencies
- [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: Tasif Hussain <tasif006@gmail.com>
We have a test
[test_structured_few_shot_examples](ad4333ca03/libs/standard-tests/langchain_tests/integration_tests/chat_models.py (L546))
in standard integration tests that implements a version of tool-calling
few shot examples that works with ~all tested providers. The formulation
supported by ~all providers is: `human message, tool call, tool message,
AI reponse`.
Here we update
`langchain_core.utils.function_calling.tool_example_to_messages` to
support this formulation.
The `tool_example_to_messages` util is undocumented outside of our API
reference. IMO, if we are testing that this function works across all
providers, it can be helpful to feature it in our guides. The structured
few-shot examples we document at the moment require users to implement
this function and can be simplified.
Last week Anthropic released version 0.39.0 of its python sdk, which
enabled support for Python 3.13. This release deleted a legacy
`client.count_tokens` method, which we currently access during init of
the `Anthropic` LLM. Anthropic has replaced this functionality with the
[client.beta.messages.count_tokens()
API](https://github.com/anthropics/anthropic-sdk-python/pull/726).
To enable support for `anthropic >= 0.39.0` and Python 3.13, here we
drop support for the legacy token counting method, and add support for
the new method via `ChatAnthropic.get_num_tokens_from_messages`.
To fully support the token counting API, we update the signature of
`get_num_tokens_from_message` to accept tools everywhere.
---------
Co-authored-by: Bagatur <22008038+baskaryan@users.noreply.github.com>
- **Description:**
Currently CommaSeparatedListOutputParser can't handle strings that may
contain commas within a column. It would parse any commas as the
delimiter.
Ex.
"foo, foo2", "bar", "baz"
It will create 4 columns: "foo", "foo2", "bar", "baz"
This should be 3 columns:
"foo, foo2", "bar", "baz"
- **Dependencies:**
Added 2 additional imports, but they are built in python packages.
import csv
from io import StringIO
- **Twitter handle:** @jkyamog
- [ ] **Add tests and docs**:
1. added simple unit test test_multiple_items_with_comma
---------
Co-authored-by: Erick Friis <erick@langchain.dev>
Co-authored-by: Bagatur <22008038+baskaryan@users.noreply.github.com>
Co-authored-by: Bagatur <baskaryan@gmail.com>
Thank you for contributing to LangChain!
- [x] **PR title**: "core: use friendlier names for duplicated nodes in
mermaid output"
- **Description:** When generating the Mermaid visualization of a chain,
if the chain had multiple nodes of the same type, the reid function
would replace their names with the UUID node_id. This made the generated
graph difficult to understand. This change deduplicates the nodes in a
chain by appending an index to their names.
- **Issue:** None
- **Discussion:**
https://github.com/langchain-ai/langchain/discussions/27714
- **Dependencies:** None
- [ ] **Add tests and docs**:
- Currently this functionality is not covered by unit tests, happy to
add tests if you'd like
- [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.
# Example Code:
```python
from langchain_core.runnables import RunnablePassthrough
def fake_llm(prompt: str) -> str: # Fake LLM for the example
return "completion"
runnable = {
'llm1': fake_llm,
'llm2': fake_llm,
} | RunnablePassthrough.assign(
total_chars=lambda inputs: len(inputs['llm1'] + inputs['llm2'])
)
print(runnable.get_graph().draw_mermaid(with_styles=False))
```
# Before
```mermaid
graph TD;
Parallel_llm1_llm2_Input --> 0b01139db5ed4587ad37964e3a40c0ec;
0b01139db5ed4587ad37964e3a40c0ec --> Parallel_llm1_llm2_Output;
Parallel_llm1_llm2_Input --> a98d4b56bd294156a651230b9293347f;
a98d4b56bd294156a651230b9293347f --> Parallel_llm1_llm2_Output;
Parallel_total_chars_Input --> Lambda;
Lambda --> Parallel_total_chars_Output;
Parallel_total_chars_Input --> Passthrough;
Passthrough --> Parallel_total_chars_Output;
Parallel_llm1_llm2_Output --> Parallel_total_chars_Input;
```
# After
```mermaid
graph TD;
Parallel_llm1_llm2_Input --> fake_llm_1;
fake_llm_1 --> Parallel_llm1_llm2_Output;
Parallel_llm1_llm2_Input --> fake_llm_2;
fake_llm_2 --> Parallel_llm1_llm2_Output;
Parallel_total_chars_Input --> Lambda;
Lambda --> Parallel_total_chars_Output;
Parallel_total_chars_Input --> Passthrough;
Passthrough --> Parallel_total_chars_Output;
Parallel_llm1_llm2_Output --> Parallel_total_chars_Input;
```
**Description:**
This PR fixes an issue where non-ASCII characters in Pydantic field
descriptions were being escaped to their Unicode representations when
using `JsonOutputParser`. The change allows non-ASCII characters to be
preserved in the output, which is especially important for multilingual
support and when working with non-English languages.
**Issue:** Fixes#27256
**Example Code:**
```python
from pydantic import BaseModel, Field
from langchain_core.output_parsers import JsonOutputParser
class Article(BaseModel):
title: str = Field(description="科学文章的标题")
output_data_structure = Article
parser = JsonOutputParser(pydantic_object=output_data_structure)
print(parser.get_format_instructions())
```
**Previous Output**:
```... "title": {"description": "\\u79d1\\u5b66\\u6587\\u7ae0\\u7684\\u6807\\u9898", "title": "Title", "type": "string"}} ...```
**Current Output**:
```... "title": {"description": "科学文章的标题", "title": "Title", "type":
"string"}} ...```
**Changes made**:
- Modified `json.dumps()` call in
`langchain_core/output_parsers/json.py` to use `ensure_ascii=False`
- Added a unit test to verify Unicode handling
Co-authored-by: Harsimran-19 <harsimran1869@gmail.com>
Fixes#27411
**Description:** Adds `template_format` to the `ImagePromptTemplate`
class and updates passing in the `template_format` parameter from
ChatPromptTemplate instead of the hardcoded "f-string".
Also updated docs and typing related to `template_format` to be more
up-to-date and specific.
**Dependencies:** None
**Add tests and docs**: Added unit tests to validate fix. Needed to
update `test_chat` snapshot due to adding new attribute
`template_format` in `ImagePromptTemplate`.
---------
Co-authored-by: Vadym Barda <vadym@langchain.dev>
## Description
This PR fixes the context loss issue in `AsyncCallbackManager`,
specifically in `on_llm_start` and `on_chat_model_start` methods. It
properly honors the `run_inline` attribute of callback handlers,
preventing race conditions and ordering issues.
Key changes:
1. Separate handlers into inline and non-inline groups.
2. Execute inline handlers sequentially for each prompt.
3. Execute non-inline handlers concurrently across all prompts.
4. Preserve context for stateful handlers.
5. Maintain performance benefits for non-inline handlers.
**These changes are implemented in `AsyncCallbackManager` rather than
`ahandle_event` because the issue occurs at the prompt and message_list
levels, not within individual events.**
## Testing
- Test case implemented in #26857 now passes, verifying execution order
for inline handlers.
## Related Issues
- Fixes issue discussed in #23909
## Dependencies
No new dependencies are required.
---
@eyurtsev: This PR implements the discussed changes to respect
`run_inline` in `AsyncCallbackManager`. Please review and advise on any
needed changes.
Twitter handle: @parambharat
---------
Co-authored-by: Eugene Yurtsev <eyurtsev@gmail.com>
Added `**kwargs` parameters to the `index` and `aindex` functions in
`libs/core/langchain_core/indexing/api.py`. This allows users to pass
additional arguments to the `add_documents` and `aadd_documents`
methods, enabling the specification of a custom `vector_field`. For
example, users can now use `vector_field="embedding"` when indexing
documents in `OpenSearchVectorStore`
---------
Co-authored-by: Eugene Yurtsev <eyurtsev@gmail.com>
Given the current erroring behavior, every time we've moved a kwarg from
model_kwargs and made it its own field that was a breaking change.
Updating this behavior to support the old instantiations /
serializations.
Assuming build_extra_kwargs was not something that itself is being used
externally and needs to be kept backwards compatible
This adds support for inject tool args that are arbitrary types when
used with pydantic 2.
We'll need to add similar logic on the v1 path, and potentially mirror
the config from the original model when we're doing the subset.
- **Description:** prevent index function to re-index entire source
document even if nothing has changed.
- **Issue:** #22135
I worked on a solution to this issue that is a compromise between being
cheap and being fast.
In the previous code, when batch_size is greater than the number of docs
from a certain source almost the entire source is deleted (all documents
from that source except for the documents in the first batch)
My solution deletes documents from vector store and record manager only
if at least one document has changed for that source.
Hope this can help!
---------
Co-authored-by: Eugene Yurtsev <eyurtsev@gmail.com>
template_format is an init argument on ChatPromptTemplate but not an
attribute on the object so was getting shoved into
StructuredPrompt.structured_ouptut_kwargs
This prevents `trim_messages` from raising an `IndexError` when invoked
with `include_system=True`, `strategy="last"`, and an empty message
list.
Fixes#26895
Dependencies: none
