- StopIteration can't be set on an asyncio.Future it raises a TypeError
and leaves the Future pending forever so we need to convert it to a
RuntimeError
This PR adds deduplication of callback handlers in merge_configs.
Fix for this issue:
https://github.com/langchain-ai/langchain/issues/22227
The issue appears when the code is:
1) running python >=3.11
2) invokes a runnable from within a runnable
3) binds the callbacks to the child runnable from the parent runnable
using with_config
In this case, the same callbacks end up appearing twice: (1) the first
time from with_config, (2) the second time with langchain automatically
propagating them on behalf of the user.
Prior to this PR this will emit duplicate events:
```python
@tool
async def get_items(question: str, callbacks: Callbacks): # <--- Accept callbacks
"""Ask question"""
template = ChatPromptTemplate.from_messages(
[
(
"human",
"'{question}"
)
]
)
chain = template | chat_model.with_config(
{
"callbacks": callbacks, # <-- Propagate callbacks
}
)
return await chain.ainvoke({"question": question})
```
Prior to this PR this will work work correctly (no duplicate events):
```python
@tool
async def get_items(question: str, callbacks: Callbacks): # <--- Accept callbacks
"""Ask question"""
template = ChatPromptTemplate.from_messages(
[
(
"human",
"'{question}"
)
]
)
chain = template | chat_model
return await chain.ainvoke({"question": question}, {"callbacks": callbacks})
```
This will also work (as long as the user is using python >= 3.11) -- as
langchain will automatically propagate callbacks
```python
@tool
async def get_items(question: str,):
"""Ask question"""
template = ChatPromptTemplate.from_messages(
[
(
"human",
"'{question}"
)
]
)
chain = template | chat_model
return await chain.ainvoke({"question": question})
```
- This is a pattern that shows up occasionally in langgraph questions,
people chain a graph to something else after, and want to pass the graph
some kwargs (eg. stream_mode)
LangSmith and LangChain context var handling evolved in parallel since
originally we didn't expect people to want to interweave the decorator
and langchain code.
Once we get a new langsmith release, this PR will let you seemlessly
hand off between @traceable context and runnable config context so you
can arbitrarily nest code.
It's expected that this fails right now until we get another release of
the SDK
# Description
## Problem
`Runnable.get_graph` fails when `InputType` or `OutputType` property
raises `TypeError`.
-
003c98e5b4/libs/core/langchain_core/runnables/base.py (L250-L274)
-
003c98e5b4/libs/core/langchain_core/runnables/base.py (L394-L396)
This problem prevents getting a graph of `Runnable` objects whose
`InputType` or `OutputType` property raises `TypeError` but whose
`invoke` works well, such as `langchain.output_parsers.RegexParser`,
which I have already pointed out in #19792 that a `TypeError` would
occur.
## Solution
- Add `try-except` syntax to handle `TypeError` to the codes which get
`input_node` and `output_node`.
# Issue
- #19801
# Twitter Handle
- [hmdev3](https://twitter.com/hmdev3)
---------
Co-authored-by: Bagatur <baskaryan@gmail.com>
```python
class UsageMetadata(TypedDict):
"""Usage metadata for a message, such as token counts.
Attributes:
input_tokens: (int) count of input (or prompt) tokens
output_tokens: (int) count of output (or completion) tokens
total_tokens: (int) total token count
"""
input_tokens: int
output_tokens: int
total_tokens: int
```
```python
class AIMessage(BaseMessage):
...
usage_metadata: Optional[UsageMetadata] = None
"""If provided, token usage information associated with the message."""
...
```
- if tap_output_iter/aiter is called multiple times for the same run
issue events only once
- if chat model run is tapped don't issue duplicate on_llm_new_token
events
- if first chunk arrives after run has ended do not emit it as a stream
event
---------
Co-authored-by: Eugene Yurtsev <eyurtsev@gmail.com>
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, hwchase17.
To permit proper coercion of objects like the following:
```python
class MyAsyncCallable:
async def __call__(self, foo):
return await ...
class MyAsyncGenerator:
async def __call__(self, foo):
await ...
yield
```
This PR introduces a v2 implementation of astream events that removes
intermediate abstractions and fixes some issues with v1 implementation.
The v2 implementation significantly reduces relevant code that's
associated with the astream events implementation together with
overhead.
After this PR, the astream events implementation:
- Uses an async callback handler
- No longer relies on BaseTracer
- No longer relies on json patch
As a result of this re-write, a number of issues were discovered with
the existing implementation.
## Changes in V2 vs. V1
### on_chat_model_end `output`
The outputs associated with `on_chat_model_end` changed depending on
whether it was within a chain or not.
As a root level runnable the output was:
```python
"data": {"output": AIMessageChunk(content="hello world!", id='some id')}
```
As part of a chain the output was:
```
"data": {
"output": {
"generations": [
[
{
"generation_info": None,
"message": AIMessageChunk(
content="hello world!", id=AnyStr()
),
"text": "hello world!",
"type": "ChatGenerationChunk",
}
]
],
"llm_output": None,
}
},
```
After this PR, we will always use the simpler representation:
```python
"data": {"output": AIMessageChunk(content="hello world!", id='some id')}
```
**NOTE** Non chat models (i.e., regular LLMs) are still associated with
the more verbose format.
### Remove some `_stream` events
`on_retriever_stream` and `on_tool_stream` events were removed -- these
were not real events, but created as an artifact of implementing on top
of astream_log.
The same information is already available in the `x_on_end` events.
### Propagating Names
Names of runnables have been updated to be more consistent
```python
model = GenericFakeChatModel(messages=infinite_cycle).configurable_fields(
messages=ConfigurableField(
id="messages",
name="Messages",
description="Messages return by the LLM",
)
)
```
Before:
```python
"name": "RunnableConfigurableFields",
```
After:
```python
"name": "GenericFakeChatModel",
```
### on_retriever_end
on_retriever_end will always return `output` which is a list of
documents (rather than a dict containing a key called "documents")
### Retry events
Removed the `on_retry` callback handler. It was incorrectly showing that
the failed function being retried has invoked `on_chain_end`
https://github.com/langchain-ai/langchain/pull/21638/files#diff-e512e3f84daf23029ebcceb11460f1c82056314653673e450a5831147d8cb84dL1394
Add unit tests that show differences between sync / async versions when
streaming.
The inner on_chain_chunk event is missing if mixing sync and async
functionality. Likely due to missing tap_output_iter implementation on
the sync variant of `_transform_stream_with_config`
- it's only node ids that are limited
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, hwchase17.
Causes an issue for this code
```python
from langchain.chat_models.openai import ChatOpenAI
from langchain.output_parsers.openai_tools import JsonOutputToolsParser
from langchain.schema import SystemMessage
prompt = SystemMessage(content="You are a nice assistant.") + "{question}"
llm = ChatOpenAI(
model_kwargs={
"tools": [
{
"type": "function",
"function": {
"name": "web_search",
"description": "Searches the web for the answer to the question.",
"parameters": {
"type": "object",
"properties": {
"query": {
"type": "string",
"description": "The question to search for.",
},
},
},
},
}
],
},
streaming=True,
)
parser = JsonOutputToolsParser(first_tool_only=True)
llm_chain = prompt | llm | parser | (lambda x: x)
for chunk in llm_chain.stream({"question": "tell me more about turtles"}):
print(chunk)
# message = llm_chain.invoke({"question": "tell me more about turtles"})
# print(message)
```
Instead by definition, we'll assume that RunnableLambdas consume the
entire stream and that if the stream isn't addable then it's the last
message of the stream that's in the usable format.
---
If users want to use addable dicts, they can wrap the dict in an
AddableDict class.
---
Likely, need to follow up with the same change for other places in the
code that do the upgrade
- would happen when user's code tries to access attritbute that doesnt
exist, we prefer to let this crash in the user's code, rather than here
- also catch more cases where a runnable is invoked/streamed inside a
lambda. before we weren't seeing these as deps
- Add conditional: bool property to json representation of the graphs
- Add option to generate mermaid graph stripped of styles (useful as a
text representation of graph)
…s arg too
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, hwchase17.
core[minor], langchain[patch], openai[minor], anthropic[minor], fireworks[minor], groq[minor], mistralai[minor]
```python
class ToolCall(TypedDict):
name: str
args: Dict[str, Any]
id: Optional[str]
class InvalidToolCall(TypedDict):
name: Optional[str]
args: Optional[str]
id: Optional[str]
error: Optional[str]
class ToolCallChunk(TypedDict):
name: Optional[str]
args: Optional[str]
id: Optional[str]
index: Optional[int]
class AIMessage(BaseMessage):
...
tool_calls: List[ToolCall] = []
invalid_tool_calls: List[InvalidToolCall] = []
...
class AIMessageChunk(AIMessage, BaseMessageChunk):
...
tool_call_chunks: Optional[List[ToolCallChunk]] = None
...
```
Important considerations:
- Parsing logic occurs within different providers;
- ~Changing output type is a breaking change for anyone doing explicit
type checking;~
- ~Langsmith rendering will need to be updated:
https://github.com/langchain-ai/langchainplus/pull/3561~
- ~Langserve will need to be updated~
- Adding chunks:
- ~AIMessage + ToolCallsMessage = ToolCallsMessage if either has
non-null .tool_calls.~
- Tool call chunks are appended, merging when having equal values of
`index`.
- additional_kwargs accumulate the normal way.
- During streaming:
- ~Messages can change types (e.g., from AIMessageChunk to
AIToolCallsMessageChunk)~
- Output parsers parse additional_kwargs (during .invoke they read off
tool calls).
Packages outside of `partners/`:
- https://github.com/langchain-ai/langchain-cohere/pull/7
- https://github.com/langchain-ai/langchain-google/pull/123/files
---------
Co-authored-by: Chester Curme <chester.curme@gmail.com>
This PR completes work for PR #18798 to expose raw tool output in
on_tool_end.
Affected APIs:
* astream_log
* astream_events
* callbacks sent to langsmith via langsmith-sdk
* Any other code that relies on BaseTracer!
---------
Co-authored-by: Eugene Yurtsev <eyurtsev@gmail.com>
Co-authored-by: Bagatur <baskaryan@gmail.com>
- This ensures ids are stable across streamed chunks
- Multiple messages in batch call get separate ids
- Also fix ids being dropped when combining message chunks
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, hwchase17.
- **Description:** Add functionality to generate Mermaid syntax and
render flowcharts from graph data. This includes support for custom node
colors and edge curve styles, as well as the ability to export the
generated graphs to PNG images using either the Mermaid.INK API or
Pyppeteer for local rendering.
- **Dependencies:** Optional dependencies are `pyppeteer` if rendering
wants to be done using Pypeteer and Javascript code.
---------
Co-authored-by: Angel Igareta <angel.igareta@klarna.com>
Co-authored-by: Bagatur <22008038+baskaryan@users.noreply.github.com>
- .stream() and .astream() call on_llm_new_token, removing the need for
subclasses to do so. Backwards compatible because now we don't pass
run_manager into ._stream and ._astream
- .generate() and .agenerate() now handle `stream: bool` kwarg for
_generate and _agenerate. Subclasses handle this arg by delegating to
._stream(), now one less thing they need to do. Backwards compat because
this is an optional arg that we now never pass to the subclasses
- .generate() and .agenerate() now inspect callback handlers to decide
on a default value for stream:bool if not passed in. This auto enables
streaming when using astream_events and astream_log
- as a result of these three changes any usage of .astream_events and
.astream_log should now yield chat model stream events
- In future PRs we can update all subclasses to reflect these two things
now handled by base class, but in meantime all will continue to work
The root run id (~trace id's) is useful for assigning feedback, but the
current recommended approach is to use callbacks to retrieve it, which
has some drawbacks:
1. Doesn't work for streaming until after the first event
2. Doesn't let you call other endpoints with the same trace ID in
parallel (since you have to wait until the call is completed/started to
use
This PR lets you provide = "run_id" in the runnable config.
Couple considerations:
1. For batch calls, we split the trace up into separate trees (to permit
better rendering). We keep the provided run ID for the first one and
generate a unique one for other elements of the batch.
2. For nested calls, the provided ID is ONLY used on the top root/trace.
### Example Usage
```
chain.invoke("foo", {"run_id": uuid.uuid4()})
```
This PR adds `batch as completed` method to the standard Runnable
interface. It takes in a list of inputs and yields the corresponding
outputs as the inputs are completed.
