jyapayne/langflow
1
0
Fork 0
Code Issues Pull requests Projects Releases 1 Packages Wiki Activity Actions

Merge remote-tracking branch 'origin/dev' into streaming

Browse source
This commit is contained in:
Gabriel Almeida 2023-05-10 11:19:04 -03:00
commit fc007e4349
25 changed files with 384 additions and 937 deletions
Download patch file Download diff file Expand all files Collapse all files

2
.github/workflows/test.yml vendored
View file

@ -30,4 +30,4 @@ jobs:
run: poetry install
- name: Run unit tests
run: |
make test
make tests

7
Makefile
View file

@ -1,4 +1,4 @@
.PHONY: all format lint build
.PHONY: all format lint build build_frontend install_frontend run_frontend run_backend dev help tests coverage
all: help
@ -8,7 +8,7 @@ coverage:
--cov-report xml \
--cov-report term-missing:skip-covered
test:
tests:
poetry run pytest tests
format:
@ -71,3 +71,6 @@ help:
@echo 'build - build the frontend static files and package the project'
@echo 'publish - build the frontend static files and package the project and publish it to PyPI'
@echo 'dev - run the project in development mode with docker compose'
@echo 'tests - run the tests'
@echo 'coverage - run the tests and generate a coverage report'
@echo '----'

39
poetry.lock generated
View file

@ -1708,14 +1708,14 @@ test = ["ipykernel", "pre-commit", "pytest", "pytest-cov", "pytest-timeout"]
[[package]]
name = "langchain"
version = "0.0.158"
version = "0.0.160"
description = "Building applications with LLMs through composability"
category = "main"
optional = false
python-versions = ">=3.8.1,<4.0"
files = [
{file = "langchain-0.0.158-py3-none-any.whl", hash = "sha256:2121874e3c72db8987467e77c1890d53e146e5f90ab249e82194187568285e43"},
{file = "langchain-0.0.158.tar.gz", hash = "sha256:29b578a6c3ccb97b63ffa93451c590e03aa55c19a834c0a2d2f8573139ca5e90"},
{file = "langchain-0.0.160-py3-none-any.whl", hash = "sha256:e09310cc07c38a5e6777bd3d30c51227e2da775d4267d3fb72697a4de3931da3"},
{file = "langchain-0.0.160.tar.gz", hash = "sha256:427c142e2fdb9f9ef9f2352a2c82db28db4d11c61f02b20dd33e77d850fe81cc"},
]
[package.dependencies]
@ -1728,12 +1728,12 @@ openapi-schema-pydantic = ">=1.2,<2.0"
pydantic = ">=1,<2"
PyYAML = ">=5.4.1"
requests = ">=2,<3"
SQLAlchemy = ">=1.3,<3"
SQLAlchemy = ">=1.4,<3"
tenacity = ">=8.1.0,<9.0.0"
tqdm = ">=4.48.0"
[package.extras]
all = ["O365 (>=2.0.26,<3.0.0)", "aleph-alpha-client (>=2.15.0,<3.0.0)", "anthropic (>=0.2.6,<0.3.0)", "arxiv (>=1.4,<2.0)", "atlassian-python-api (>=3.36.0,<4.0.0)", "azure-cosmos (>=4.4.0b1,<5.0.0)", "azure-identity (>=1.12.0,<2.0.0)", "beautifulsoup4 (>=4,<5)", "clickhouse-connect (>=0.5.14,<0.6.0)", "cohere (>=3,<4)", "deeplake (>=3.3.0,<4.0.0)", "duckduckgo-search (>=2.8.6,<3.0.0)", "elasticsearch (>=8,<9)", "faiss-cpu (>=1,<2)", "google-api-python-client (==2.70.0)", "google-search-results (>=2,<3)", "gptcache (>=0.1.7)", "html2text (>=2020.1.16,<2021.0.0)", "huggingface_hub (>=0,<1)", "jina (>=3.14,<4.0)", "jinja2 (>=3,<4)", "lancedb (>=0.1,<0.2)", "lark (>=1.1.5,<2.0.0)", "manifest-ml (>=0.0.1,<0.0.2)", "networkx (>=2.6.3,<3.0.0)", "nlpcloud (>=1,<2)", "nltk (>=3,<4)", "nomic (>=1.0.43,<2.0.0)", "openai (>=0,<1)", "opensearch-py (>=2.0.0,<3.0.0)", "pexpect (>=4.8.0,<5.0.0)", "pgvector (>=0.1.6,<0.2.0)", "pinecone-client (>=2,<3)", "pinecone-text (>=0.4.2,<0.5.0)", "psycopg2-binary (>=2.9.5,<3.0.0)", "pyowm (>=3.3.0,<4.0.0)", "pypdf (>=3.4.0,<4.0.0)", "pytesseract (>=0.3.10,<0.4.0)", "pyvespa (>=0.33.0,<0.34.0)", "qdrant-client (>=1.1.2,<2.0.0)", "redis (>=4,<5)", "sentence-transformers (>=2,<3)", "spacy (>=3,<4)", "tensorflow-text (>=2.11.0,<3.0.0)", "tiktoken (>=0.3.2,<0.4.0)", "torch (>=1,<3)", "transformers (>=4,<5)", "weaviate-client (>=3,<4)", "wikipedia (>=1,<2)", "wolframalpha (==5.0.0)"]
all = ["O365 (>=2.0.26,<3.0.0)", "aleph-alpha-client (>=2.15.0,<3.0.0)", "anthropic (>=0.2.6,<0.3.0)", "arxiv (>=1.4,<2.0)", "atlassian-python-api (>=3.36.0,<4.0.0)", "azure-cosmos (>=4.4.0b1,<5.0.0)", "azure-identity (>=1.12.0,<2.0.0)", "beautifulsoup4 (>=4,<5)", "clickhouse-connect (>=0.5.14,<0.6.0)", "cohere (>=3,<4)", "deeplake (>=3.3.0,<4.0.0)", "duckduckgo-search (>=2.8.6,<3.0.0)", "elasticsearch (>=8,<9)", "faiss-cpu (>=1,<2)", "google-api-python-client (==2.70.0)", "google-search-results (>=2,<3)", "gptcache (>=0.1.7)", "html2text (>=2020.1.16,<2021.0.0)", "huggingface_hub (>=0,<1)", "jina (>=3.14,<4.0)", "jinja2 (>=3,<4)", "jq (>=1.4.1,<2.0.0)", "lancedb (>=0.1,<0.2)", "lark (>=1.1.5,<2.0.0)", "manifest-ml (>=0.0.1,<0.0.2)", "networkx (>=2.6.3,<3.0.0)", "nlpcloud (>=1,<2)", "nltk (>=3,<4)", "nomic (>=1.0.43,<2.0.0)", "openai (>=0,<1)", "opensearch-py (>=2.0.0,<3.0.0)", "pexpect (>=4.8.0,<5.0.0)", "pgvector (>=0.1.6,<0.2.0)", "pinecone-client (>=2,<3)", "pinecone-text (>=0.4.2,<0.5.0)", "psycopg2-binary (>=2.9.5,<3.0.0)", "pyowm (>=3.3.0,<4.0.0)", "pypdf (>=3.4.0,<4.0.0)", "pytesseract (>=0.3.10,<0.4.0)", "pyvespa (>=0.33.0,<0.34.0)", "qdrant-client (>=1.1.2,<2.0.0)", "redis (>=4,<5)", "sentence-transformers (>=2,<3)", "spacy (>=3,<4)", "tensorflow-text (>=2.11.0,<3.0.0)", "tiktoken (>=0.3.2,<0.4.0)", "torch (>=1,<3)", "transformers (>=4,<5)", "weaviate-client (>=3,<4)", "wikipedia (>=1,<2)", "wolframalpha (==5.0.0)"]
azure = ["azure-core (>=1.26.4,<2.0.0)", "azure-cosmos (>=4.4.0b1,<5.0.0)", "azure-identity (>=1.12.0,<2.0.0)", "openai (>=0,<1)"]
cohere = ["cohere (>=3,<4)"]
embeddings = ["sentence-transformers (>=2,<3)"]
@ -1743,13 +1743,13 @@ qdrant = ["qdrant-client (>=1.1.2,<2.0.0)"]
[[package]]
name = "lit"
version = "16.0.2"
version = "16.0.3"
description = "A Software Testing Tool"
category = "main"
optional = false
python-versions = "*"
files = [
{file = "lit-16.0.2.tar.gz", hash = "sha256:d743ef55cb58764bba85768c502e2d68d87aeb4303d508a18abaa8a35077ab25"},
{file = "lit-16.0.3.tar.gz", hash = "sha256:25524fe51fa3261212cfd86a8891429ed0460e247384c5a2001612d08e362e00"},
]
[[package]]
@ -4555,18 +4555,18 @@ files = [
[[package]]
name = "types-requests"
version = "2.29.0.0"
version = "2.30.0.0"
description = "Typing stubs for requests"
category = "dev"
optional = false
python-versions = "*"
files = [
{file = "types-requests-2.29.0.0.tar.gz", hash = "sha256:c86f4a955d943d2457120dbe719df24ef0924e11177164d10a0373cf311d7b4d"},
{file = "types_requests-2.29.0.0-py3-none-any.whl", hash = "sha256:4cf6e323e856c779fbe8815bb977a5bf5d6c5034713e4c17ff2a9a20610f5b27"},
{file = "types-requests-2.30.0.0.tar.gz", hash = "sha256:dec781054324a70ba64430ae9e62e7e9c8e4618c185a5cb3f87a6738251b5a31"},
{file = "types_requests-2.30.0.0-py3-none-any.whl", hash = "sha256:c6cf08e120ca9f0dc4fa4e32c3f953c3fba222bcc1db6b97695bce8da1ba9864"},
]
[package.dependencies]
types-urllib3 = "<1.27"
types-urllib3 = "*"
[[package]]
name = "types-urllib3"
@ -4891,6 +4891,21 @@ files = [
[package.extras]
test = ["pytest (>=6.0.0)"]
[[package]]
name = "wikipedia"
version = "1.4.0"
description = "Wikipedia API for Python"
category = "main"
optional = false
python-versions = "*"
files = [
{file = "wikipedia-1.4.0.tar.gz", hash = "sha256:db0fad1829fdd441b1852306e9856398204dc0786d2996dd2e0c8bb8e26133b2"},
]
[package.dependencies]
beautifulsoup4 = "*"
requests = ">=2.0.0,<3.0.0"
[[package]]
name = "xlsxwriter"
version = "3.1.0"
@ -5069,4 +5084,4 @@ cffi = ["cffi (>=1.11)"]
[metadata]
lock-version = "2.0"
python-versions = "^3.9"
content-hash = "5e7c877648cffc95b5312d87c82afa7bb6a9ce331492769c93efcb8f99252d9a"
content-hash = "5aca7cac07e5b678d4b01d994227da5ef06f4e8da6c2a854bab7e7e6cd17ccb5"

1
pyproject.toml
View file

@ -49,6 +49,7 @@ psycopg2-binary = "^2.9.6"
pyarrow = "^11.0.0"
websockets = "^11.0.2"
tiktoken = "^0.3.3"
wikipedia = "^1.4.0"
[tool.poetry.group.dev.dependencies]
black = "^23.1.0"

16
src/backend/langflow/api/endpoints.py
View file

@ -5,6 +5,12 @@ from fastapi import APIRouter, HTTPException
from langflow.interface.run import process_graph_cached
from langflow.interface.types import build_langchain_types_dict
from langflow.api.schemas import (
ExportedFlow,
GraphData,
PredictRequest,
PredictResponse,
)
# build router
router = APIRouter()
@ -16,10 +22,14 @@ def get_all():
return build_langchain_types_dict()
@router.post("/predict")
def get_load(data: Dict[str, Any]):
@router.post("/predict", response_model=PredictResponse)
async def get_load(predict_request: PredictRequest):
try:
return process_graph_cached(data)
exported_flow: ExportedFlow = predict_request.exported_flow
graph_data: GraphData = exported_flow.data
data = graph_data.dict()
response = process_graph_cached(data, predict_request.message)
return PredictResponse(result=response.get("result", ""))
except Exception as e:
# Log stack trace
logger.exception(e)

31
src/backend/langflow/api/schemas.py
View file

@ -1,8 +1,37 @@
from typing import Any, Union
from typing import Any, Union, Dict, List
from pydantic import BaseModel, validator
class GraphData(BaseModel):
"""Data inside the exported flow."""
nodes: List[Dict[str, Any]]
edges: List[Dict[str, Any]]
class ExportedFlow(BaseModel):
"""Exported flow from LangFlow."""
description: str
name: str
id: str
data: GraphData
class PredictRequest(BaseModel):
"""Predict request schema."""
message: str
exported_flow: ExportedFlow
class PredictResponse(BaseModel):
"""Predict response schema."""
result: str
class ChatMessage(BaseModel):
"""Chat message schema."""

2
src/backend/langflow/api/validate.py
View file

@ -49,5 +49,5 @@ def post_validate_node(node_id: str, data: dict):
return str(node.params)
raise Exception(f"Node {node_id} not found")
except Exception as e:
logger.exception(e)
logger.error(e)
raise HTTPException(status_code=500, detail=str(e)) from e

1
src/backend/langflow/cache/base.py vendored
View file

@ -48,6 +48,7 @@ def memoize_dict(maxsize=128):
cache.clear()
wrapper.clear_cache = clear_cache # type: ignore
wrapper.cache = cache # type: ignore
return wrapper
return decorator

1
src/backend/langflow/config.yaml
View file

@ -51,6 +51,7 @@ tools:
- BingSearchRun
- GoogleSearchRun
- GoogleSearchResults
- GoogleSerperRun
- JsonListKeysTool
- JsonGetValueTool
- PythonREPLTool

20
src/backend/langflow/interface/agents/custom.py
View file

@ -27,7 +27,7 @@ from langchain.agents.agent_toolkits.vectorstore.prompt import (
from langchain.agents.mrkl.prompt import FORMAT_INSTRUCTIONS
from langchain.agents.mrkl.prompt import FORMAT_INSTRUCTIONS as SQL_FORMAT_INSTRUCTIONS
from langchain.base_language import BaseLanguageModel
from langchain.llms.base import BaseLLM
from langchain.memory.chat_memory import BaseChatMemory
from langchain.sql_database import SQLDatabase
from langchain.tools.python.tool import PythonAstREPLTool
@ -63,7 +63,7 @@ class JsonAgent(AgentExecutor):
llm=llm,
prompt=prompt,
)
agent = ZeroShotAgent(llm_chain=llm_chain, allowed_tools=tool_names)
agent = ZeroShotAgent(llm_chain=llm_chain, allowed_tools=tool_names) # type: ignore
return cls.from_agent_and_tools(agent=agent, tools=tools, verbose=True)
def run(self, *args, **kwargs):
@ -110,7 +110,7 @@ class CSVAgent(AgentExecutor):
prompt=partial_prompt,
)
tool_names = {tool.name for tool in tools}
agent = ZeroShotAgent(llm_chain=llm_chain, allowed_tools=tool_names, **kwargs)
agent = ZeroShotAgent(llm_chain=llm_chain, allowed_tools=tool_names, **kwargs) # type: ignore
return cls.from_agent_and_tools(agent=agent, tools=tools, verbose=True)
@ -134,7 +134,7 @@ class VectorStoreAgent(AgentExecutor):
@classmethod
def from_toolkit_and_llm(
cls, llm: BaseLLM, vectorstoreinfo: VectorStoreInfo, **kwargs: Any
cls, llm: BaseLanguageModel, vectorstoreinfo: VectorStoreInfo, **kwargs: Any
):
"""Construct a vectorstore agent from an LLM and tools."""
@ -147,7 +147,7 @@ class VectorStoreAgent(AgentExecutor):
prompt=prompt,
)
tool_names = {tool.name for tool in tools}
agent = ZeroShotAgent(llm_chain=llm_chain, allowed_tools=tool_names, **kwargs)
agent = ZeroShotAgent(llm_chain=llm_chain, allowed_tools=tool_names, **kwargs) # type: ignore
return AgentExecutor.from_agent_and_tools(
agent=agent, tools=tools, verbose=True
)
@ -171,7 +171,9 @@ class SQLAgent(AgentExecutor):
super().__init__(*args, **kwargs)
@classmethod
def from_toolkit_and_llm(cls, llm: BaseLLM, database_uri: str, **kwargs: Any):
def from_toolkit_and_llm(
cls, llm: BaseLanguageModel, database_uri: str, **kwargs: Any
):
"""Construct a sql agent from an LLM and tools."""
db = SQLDatabase.from_uri(database_uri)
toolkit = SQLDatabaseToolkit(db=db, llm=llm)
@ -213,7 +215,7 @@ class SQLAgent(AgentExecutor):
prompt=prompt,
)
tool_names = {tool.name for tool in tools} # type: ignore
agent = ZeroShotAgent(llm_chain=llm_chain, allowed_tools=tool_names, **kwargs)
agent = ZeroShotAgent(llm_chain=llm_chain, allowed_tools=tool_names, **kwargs) # type: ignore
return AgentExecutor.from_agent_and_tools(
agent=agent,
tools=tools, # type: ignore
@ -256,7 +258,7 @@ class VectorStoreRouterAgent(AgentExecutor):
prompt=prompt,
)
tool_names = {tool.name for tool in tools}
agent = ZeroShotAgent(llm_chain=llm_chain, allowed_tools=tool_names, **kwargs)
agent = ZeroShotAgent(llm_chain=llm_chain, allowed_tools=tool_names, **kwargs) # type: ignore
return AgentExecutor.from_agent_and_tools(
agent=agent, tools=tools, verbose=True
)
@ -275,7 +277,7 @@ class InitializeAgent(AgentExecutor):
@classmethod
def initialize(
cls,
llm: BaseLLM,
llm: BaseLanguageModel,
tools: List[Tool],
agent: str,
memory: Optional[BaseChatMemory] = None,

2
src/backend/langflow/interface/agents/prebuilt.py
View file

@ -33,7 +33,7 @@ class MalfoyAgent(AgentExecutor):
llm=llm,
prompt=prompt,
)
agent = ZeroShotAgent(llm_chain=llm_chain, allowed_tools=tool_names)
agent = ZeroShotAgent(llm_chain=llm_chain, allowed_tools=tool_names) # type: ignore
return cls.from_agent_and_tools(agent=agent, tools=tools, verbose=True)
def run(self, *args, **kwargs):

12
src/backend/langflow/interface/importing/utils.py
View file

@ -7,11 +7,9 @@ from langchain import PromptTemplate
from langchain.agents import Agent
from langchain.chains.base import Chain
from langchain.chat_models.base import BaseChatModel
from langchain.llms.base import BaseLLM
from langchain.base_language import BaseLanguageModel
from langchain.tools import BaseTool
from langflow.interface.tools.base import tool_creator
def import_module(module_path: str) -> Any:
"""Import module from module path"""
@ -100,15 +98,19 @@ def import_agent(agent: str) -> Agent:
return import_class(f"langchain.agents.{agent}")
def import_llm(llm: str) -> BaseLLM:
def import_llm(llm: str) -> BaseLanguageModel:
"""Import llm from llm name"""
return import_class(f"langchain.llms.{llm}")
def import_tool(tool: str) -> BaseTool:
"""Import tool from tool name"""
from langflow.interface.tools.base import tool_creator
return tool_creator.type_to_loader_dict[tool]["fcn"]
if tool in tool_creator.type_to_loader_dict:
return tool_creator.type_to_loader_dict[tool]["fcn"]
return import_class(f"langchain.tools.{tool}")
def import_chain(chain: str) -> Type[Chain]:

6
src/backend/langflow/interface/loading.py
View file

@ -15,7 +15,7 @@ from langchain.agents.loading import load_agent_from_config
from langchain.agents.tools import Tool
from langchain.callbacks.base import BaseCallbackManager
from langchain.chains.loading import load_chain_from_config
from langchain.llms.base import BaseLLM
from langchain.base_language import BaseLanguageModel
from langchain.llms.loading import load_llm_from_config
from pydantic import ValidationError
@ -74,12 +74,10 @@ def instantiate_class(node_type: str, base_type: str, params: Dict) -> Any:
return loaded_toolkit
elif base_type == "embeddings":
# ? Why remove model from params?
try:
params.pop("model")
except KeyError:
pass
# remove all params that are not in class_object.__fields__
try:
return class_object(**params)
@ -188,7 +186,7 @@ def load_langchain_type_from_config(config: Dict[str, Any]):
def load_agent_executor_from_config(
config: dict,
llm: Optional[BaseLLM] = None,
llm: Optional[BaseLanguageModel] = None,
tools: Optional[list[Tool]] = None,
callback_manager: Optional[BaseCallbackManager] = None,
**kwargs: Any,

35
src/backend/langflow/interface/run.py
View file

@ -101,13 +101,12 @@ def process_graph(data_graph: Dict[str, Any]):
return {"result": str(result), "thought": thought.strip()}
def process_graph_cached(data_graph: Dict[str, Any]):
def process_graph_cached(data_graph: Dict[str, Any], message: str):
"""
Process graph by extracting input variables and replacing ZeroShotPrompt
with PromptTemplate,then run the graph and return the result and thought.
"""
# Load langchain object
message = data_graph.pop("message", "")
is_first_message = len(data_graph.get("chatHistory", [])) == 0
langchain_object = load_or_build_langchain_object(data_graph, is_first_message)
logger.debug("Loaded langchain object")
@ -120,7 +119,7 @@ def process_graph_cached(data_graph: Dict[str, Any]):
# Generate result and thought
logger.debug("Generating result and thought")
result, thought = get_result_and_steps(langchain_object, message)
result, thought = get_result_and_thought(langchain_object, message)
logger.debug("Generated result and thought")
return {"result": str(result), "thought": thought.strip()}
@ -247,7 +246,7 @@ async def get_result_and_steps(langchain_object, message: str, **kwargs):
return result, thought
def async_get_result_and_steps(langchain_object, message: str):
def get_result_and_thought(langchain_object, message: str):
"""Get result and thought from extracted json"""
try:
if hasattr(langchain_object, "verbose"):
@ -302,34 +301,6 @@ def async_get_result_and_steps(langchain_object, message: str):
return result, thought
def get_result_and_thought(extracted_json: Dict[str, Any], message: str):
"""Get result and thought from extracted json"""
try:
langchain_object = loading.load_langchain_type_from_config(
config=extracted_json
)
with io.StringIO() as output_buffer, contextlib.redirect_stdout(output_buffer):
output = langchain_object(message)
intermediate_steps = (
output.get("intermediate_steps", []) if isinstance(output, dict) else []
)
result = (
output.get(langchain_object.output_keys[0])
if isinstance(output, dict)
else output
)
if intermediate_steps:
thought = format_intermediate_steps(intermediate_steps)
else:
thought = output_buffer.getvalue()
except Exception as e:
result = f"Error: {str(e)}"
thought = ""
return result, thought
def format_intermediate_steps(intermediate_steps):
formatted_chain = "> Entering new AgentExecutor chain...\n"
for step in intermediate_steps:

5
src/backend/langflow/interface/tools/base.py
View file

@ -29,7 +29,9 @@ TOOL_INPUTS = {
placeholder="",
value="",
),
"llm": TemplateField(field_type="BaseLLM", required=True, is_list=False, show=True),
"llm": TemplateField(
field_type="BaseLanguageModel", required=True, is_list=False, show=True
),
"func": TemplateField(
field_type="function",
required=True,
@ -65,6 +67,7 @@ class ToolCreator(LangChainTypeCreator):
def type_to_loader_dict(self) -> Dict:
if self.tools_dict is None:
all_tools = {}
for tool, tool_fcn in ALL_TOOLS_NAMES.items():
tool_params = get_tool_params(tool_fcn)
tool_name = tool_params.get("name", tool)

45
src/backend/langflow/interface/tools/constants.py
View file

@ -1,3 +1,4 @@
from langchain import tools
from langchain.agents import Tool
from langchain.agents.load_tools import (
_BASE_TOOLS,
@ -5,50 +6,16 @@ from langchain.agents.load_tools import (
_EXTRA_OPTIONAL_TOOLS,
_LLM_TOOLS,
)
from langchain.tools.bing_search.tool import BingSearchRun
from langchain.tools.google_search.tool import GoogleSearchResults, GoogleSearchRun
from langchain.tools.json.tool import JsonGetValueTool, JsonListKeysTool, JsonSpec
from langchain.tools.python.tool import PythonAstREPLTool, PythonREPLTool
from langchain.tools.requests.tool import (
RequestsDeleteTool,
RequestsGetTool,
RequestsPatchTool,
RequestsPostTool,
RequestsPutTool,
)
from langchain.tools.sql_database.tool import (
InfoSQLDatabaseTool,
ListSQLDatabaseTool,
QueryCheckerTool,
QuerySQLDataBaseTool,
)
from langchain.tools.wikipedia.tool import WikipediaQueryRun
from langchain.tools.wolfram_alpha.tool import WolframAlphaQueryRun
from langchain.tools.json.tool import JsonSpec
from langflow.interface.importing.utils import import_class
from langflow.interface.tools.custom import PythonFunction
FILE_TOOLS = {"JsonSpec": JsonSpec}
CUSTOM_TOOLS = {"Tool": Tool, "PythonFunction": PythonFunction}
OTHER_TOOLS = {
"QuerySQLDataBaseTool": QuerySQLDataBaseTool,
"InfoSQLDatabaseTool": InfoSQLDatabaseTool,
"ListSQLDatabaseTool": ListSQLDatabaseTool,
"QueryCheckerTool": QueryCheckerTool,
"BingSearchRun": BingSearchRun,
"GoogleSearchRun": GoogleSearchRun,
"GoogleSearchResults": GoogleSearchResults,
"JsonListKeysTool": JsonListKeysTool,
"JsonGetValueTool": JsonGetValueTool,
"PythonREPLTool": PythonREPLTool,
"PythonAstREPLTool": PythonAstREPLTool,
"RequestsGetTool": RequestsGetTool,
"RequestsPostTool": RequestsPostTool,
"RequestsPatchTool": RequestsPatchTool,
"RequestsPutTool": RequestsPutTool,
"RequestsDeleteTool": RequestsDeleteTool,
"WikipediaQueryRun": WikipediaQueryRun,
"WolframAlphaQueryRun": WolframAlphaQueryRun,
}
OTHER_TOOLS = {tool: import_class(f"langchain.tools.{tool}") for tool in tools.__all__}
ALL_TOOLS_NAMES = {
**_BASE_TOOLS,
**_LLM_TOOLS, # type: ignore

5
src/frontend/dev.Dockerfile
View file

@ -15,7 +15,10 @@ WORKDIR /home/node/app
COPY --chown=node:node . ./
COPY ./set_proxy.sh .
RUN chmod +x set_proxy.sh && ./set_proxy.sh
RUN chmod +x set_proxy.sh && \
cat set_proxy.sh | tr -d '\r' > set_proxy_unix.sh && \
chmod +x set_proxy_unix.sh && \
./set_proxy_unix.sh
USER node

2
src/frontend/set_proxy.sh
View file

@ -7,4 +7,4 @@ packagejson=$(cat package.json)
packagejson=$(echo "$packagejson" | jq ".proxy = \"$backend_url\"")
echo "$packagejson" > package.json
echo "$packagejson" > package.json

408
src/frontend/src/contexts/tabsContext.tsx
View file

@ -1,10 +1,10 @@
import {
createContext,
useEffect,
useState,
useRef,
ReactNode,
useContext,
createContext,
useEffect,
useState,
useRef,
ReactNode,
useContext,
} from "react";
import { FlowType } from "../types/flow";
import { LangFlowState, TabsContextType } from "../types/tabs";
@ -15,221 +15,221 @@ import { APITemplateType, TemplateVariableType } from "../types/api";
const { v4: uuidv4 } = require("uuid");
const TabsContextInitialValue: TabsContextType = {
save: () => {},
tabIndex: 0,
setTabIndex: (index: number) => {},
flows: [],
removeFlow: (id: string) => {},
addFlow: (flowData?: any) => {},
updateFlow: (newFlow: FlowType) => {},
incrementNodeId: () => 0,
downloadFlow: (flow: FlowType) => {},
uploadFlow: () => {},
hardReset: () => {},
save: () => {},
tabIndex: 0,
setTabIndex: (index: number) => {},
flows: [],
removeFlow: (id: string) => {},
addFlow: (flowData?: any) => {},
updateFlow: (newFlow: FlowType) => {},
incrementNodeId: () => 0,
downloadFlow: (flow: FlowType) => {},
uploadFlow: () => {},
hardReset: () => {},
};
export const TabsContext = createContext<TabsContextType>(
TabsContextInitialValue
TabsContextInitialValue
);
export function TabsProvider({ children }: { children: ReactNode }) {
const { setNoticeData } = useContext(alertContext);
const [tabIndex, setTabIndex] = useState(0);
const [flows, setFlows] = useState<Array<FlowType>>([]);
const [id, setId] = useState("");
const { templates } = useContext(typesContext);
const { setNoticeData } = useContext(alertContext);
const [tabIndex, setTabIndex] = useState(0);
const [flows, setFlows] = useState<Array<FlowType>>([]);
const [id, setId] = useState("");
const { templates } = useContext(typesContext);
const newNodeId = useRef(0);
function incrementNodeId() {
newNodeId.current = newNodeId.current + 1;
return newNodeId.current;
}
function save() {
if (flows.length !== 0)
window.localStorage.setItem(
"tabsData",
JSON.stringify({ tabIndex, flows, id, nodeId: newNodeId.current })
);
}
useEffect(() => {
//save tabs locally
save();
}, [flows, id, tabIndex, newNodeId]);
const newNodeId = useRef(0);
function incrementNodeId() {
newNodeId.current = newNodeId.current + 1;
return newNodeId.current;
}
function save() {
if (flows.length !== 0)
window.localStorage.setItem(
"tabsData",
JSON.stringify({ tabIndex, flows, id, nodeId: newNodeId.current })
);
}
useEffect(() => {
//save tabs locally
save();
}, [flows, id, tabIndex, newNodeId]);
useEffect(() => {
//get tabs locally saved
let cookie = window.localStorage.getItem("tabsData");
if (cookie && Object.keys(templates).length > 0) {
let cookieObject: LangFlowState = JSON.parse(cookie);
cookieObject.flows.forEach((flow) => {
flow.data.nodes.forEach((node) => {
if (Object.keys(templates[node.data.type]["template"]).length > 0) {
node.data.node.template = updateTemplate(
templates[node.data.type][
"template"
] as unknown as APITemplateType,
useEffect(() => {
//get tabs locally saved
let cookie = window.localStorage.getItem("tabsData");
if (cookie && Object.keys(templates).length > 0) {
let cookieObject: LangFlowState = JSON.parse(cookie);
cookieObject.flows.forEach((flow) => {
flow.data.nodes.forEach((node) => {
if (Object.keys(templates[node.data.type]["template"]).length > 0) {
node.data.node.template = updateTemplate(
templates[node.data.type][
"template"
] as unknown as APITemplateType,
node.data.node.template as APITemplateType
);
}
});
});
setTabIndex(cookieObject.tabIndex);
setFlows(cookieObject.flows);
setId(cookieObject.id);
newNodeId.current = cookieObject.nodeId;
}
}, [templates]);
function hardReset() {
newNodeId.current = 0;
setTabIndex(0);
setFlows([]);
setId("");
}
node.data.node.template as APITemplateType
);
}
});
});
setTabIndex(cookieObject.tabIndex);
setFlows(cookieObject.flows);
setId(cookieObject.id);
newNodeId.current = cookieObject.nodeId;
}
}, [templates]);
function hardReset() {
newNodeId.current = 0;
setTabIndex(0);
setFlows([]);
setId("");
}
/**
* Downloads the current flow as a JSON file
*/
function downloadFlow(flow: FlowType) {
// create a data URI with the current flow data
const jsonString = `data:text/json;chatset=utf-8,${encodeURIComponent(
JSON.stringify(flow)
)}`;
/**
* Downloads the current flow as a JSON file
*/
function downloadFlow(flow: FlowType) {
// create a data URI with the current flow data
const jsonString = `data:text/json;chatset=utf-8,${encodeURIComponent(
JSON.stringify(flow)
)}`;
// create a link element and set its properties
const link = document.createElement("a");
link.href = jsonString;
link.download = `${normalCaseToSnakeCase(flows[tabIndex].name)}.json`;
// create a link element and set its properties
const link = document.createElement("a");
link.href = jsonString;
link.download = `${flows[tabIndex].name}.json`;
// simulate a click on the link element to trigger the download
link.click();
setNoticeData({
title: "Warning: Critical data,JSON file may including API keys.",
});
}
// simulate a click on the link element to trigger the download
link.click();
setNoticeData({
title: "Warning: Critical data,JSON file may including API keys.",
});
}
/**
* Creates a file input and listens to a change event to upload a JSON flow file.
* If the file type is application/json, the file is read and parsed into a JSON object.
* The resulting JSON object is passed to the addFlow function.
*/
function uploadFlow() {
// create a file input
const input = document.createElement("input");
input.type = "file";
// add a change event listener to the file input
input.onchange = (e: Event) => {
// check if the file type is application/json
if ((e.target as HTMLInputElement).files[0].type === "application/json") {
// get the file from the file input
const file = (e.target as HTMLInputElement).files[0];
// read the file as text
file.text().then((text) => {
// parse the text into a JSON object
let flow: FlowType = JSON.parse(text);
/**
* Creates a file input and listens to a change event to upload a JSON flow file.
* If the file type is application/json, the file is read and parsed into a JSON object.
* The resulting JSON object is passed to the addFlow function.
*/
function uploadFlow() {
// create a file input
const input = document.createElement("input");
input.type = "file";
// add a change event listener to the file input
input.onchange = (e: Event) => {
// check if the file type is application/json
if ((e.target as HTMLInputElement).files[0].type === "application/json") {
// get the file from the file input
const file = (e.target as HTMLInputElement).files[0];
// read the file as text
file.text().then((text) => {
// parse the text into a JSON object
let flow: FlowType = JSON.parse(text);
addFlow(flow);
});
}
};
// trigger the file input click event to open the file dialog
input.click();
}
/**
* Removes a flow from an array of flows based on its id.
* Updates the state of flows and tabIndex using setFlows and setTabIndex hooks.
* @param {string} id - The id of the flow to remove.
*/
function removeFlow(id: string) {
setFlows((prevState) => {
const newFlows = [...prevState];
const index = newFlows.findIndex((flow) => flow.id === id);
if (index >= 0) {
if (index === tabIndex) {
setTabIndex(flows.length - 2);
newFlows.splice(index, 1);
} else {
let flowId = flows[tabIndex].id;
newFlows.splice(index, 1);
setTabIndex(newFlows.findIndex((flow) => flow.id === flowId));
}
}
return newFlows;
});
}
/**
* Add a new flow to the list of flows.
* @param flow Optional flow to add.
*/
function addFlow(flow?: FlowType) {
// Get data from the flow or set it to null if there's no flow provided.
const data = flow?.data ? flow.data : null;
const description = flow?.description ? flow.description : "";
addFlow(flow);
});
}
};
// trigger the file input click event to open the file dialog
input.click();
}
/**
* Removes a flow from an array of flows based on its id.
* Updates the state of flows and tabIndex using setFlows and setTabIndex hooks.
* @param {string} id - The id of the flow to remove.
*/
function removeFlow(id: string) {
setFlows((prevState) => {
const newFlows = [...prevState];
const index = newFlows.findIndex((flow) => flow.id === id);
if (index >= 0) {
if (index === tabIndex) {
setTabIndex(flows.length - 2);
newFlows.splice(index, 1);
} else {
let flowId = flows[tabIndex].id;
newFlows.splice(index, 1);
setTabIndex(newFlows.findIndex((flow) => flow.id === flowId));
}
}
return newFlows;
});
}
/**
* Add a new flow to the list of flows.
* @param flow Optional flow to add.
*/
function addFlow(flow?: FlowType) {
// Get data from the flow or set it to null if there's no flow provided.
const data = flow?.data ? flow.data : null;
const description = flow?.description ? flow.description : "";
if (data) {
data.nodes.forEach((node) => {
if (Object.keys(templates[node.data.type]["template"]).length > 0) {
node.data.node.template = updateTemplate(
templates[node.data.type]["template"] as unknown as APITemplateType,
node.data.node.template as APITemplateType
);
}
});
}
// Create a new flow with a default name if no flow is provided.
let newFlow: FlowType = {
description,
name: flow?.name ?? "New Flow",
id: id.toString(),
data,
};
if (data) {
data.nodes.forEach((node) => {
if (Object.keys(templates[node.data.type]["template"]).length > 0) {
node.data.node.template = updateTemplate(
templates[node.data.type]["template"] as unknown as APITemplateType,
node.data.node.template as APITemplateType
);
}
});
}
// Create a new flow with a default name if no flow is provided.
let newFlow: FlowType = {
description,
name: flow?.name ?? "New Flow",
id: id.toString(),
data,
};
// Increment the ID counter.
setId(uuidv4());
// Increment the ID counter.
setId(uuidv4());
// Add the new flow to the list of flows.
setFlows((prevState) => {
const newFlows = [...prevState, newFlow];
return newFlows;
});
// Add the new flow to the list of flows.
setFlows((prevState) => {
const newFlows = [...prevState, newFlow];
return newFlows;
});
// Set the tab index to the new flow.
setTabIndex(flows.length);
}
/**
* Updates an existing flow with new data
* @param newFlow - The new flow object containing the updated data
*/
function updateFlow(newFlow: FlowType) {
setFlows((prevState) => {
const newFlows = [...prevState];
const index = newFlows.findIndex((flow) => flow.id === newFlow.id);
if (index !== -1) {
newFlows[index].description = newFlow.description ?? "";
newFlows[index].data = newFlow.data;
newFlows[index].name = newFlow.name;
}
return newFlows;
});
}
// Set the tab index to the new flow.
setTabIndex(flows.length);
}
/**
* Updates an existing flow with new data
* @param newFlow - The new flow object containing the updated data
*/
function updateFlow(newFlow: FlowType) {
setFlows((prevState) => {
const newFlows = [...prevState];
const index = newFlows.findIndex((flow) => flow.id === newFlow.id);
if (index !== -1) {
newFlows[index].description = newFlow.description ?? "";
newFlows[index].data = newFlow.data;
newFlows[index].name = newFlow.name;
}
return newFlows;
});
}
return (
<TabsContext.Provider
value={{
save,
hardReset,
tabIndex,
setTabIndex,
flows,
incrementNodeId,
removeFlow,
addFlow,
updateFlow,
downloadFlow,
uploadFlow,
}}
>
{children}
</TabsContext.Provider>
);
return (
<TabsContext.Provider
value={{
save,
hardReset,
tabIndex,
setTabIndex,
flows,
incrementNodeId,
removeFlow,
addFlow,
updateFlow,
downloadFlow,
uploadFlow,
}}
>
{children}
</TabsContext.Provider>
);
}

506
tests/data/basic_example.json
View file

File diff suppressed because one or more lines are too long

1
tests/test_agents_template.py
View file

@ -48,7 +48,6 @@ def test_zero_shot_agent(client: TestClient):
"type": "Tool",
"list": True,
"advanced": False,
"value": [],
}

65
tests/test_cache.py
View file

@ -1,10 +1,11 @@
import json
import tempfile
from pathlib import Path
import pytest
from langflow.cache.base import PREFIX, save_cache
from langflow.interface.run import load_langchain_object
from langflow.interface.run import (
build_graph,
build_langchain_object_with_caching,
load_or_build_langchain_object,
)
def get_graph(_type="basic"):
@ -40,26 +41,44 @@ def langchain_objects_are_equal(obj1, obj2):
return str(obj1) == str(obj2)
def test_cache_creation(basic_data_graph):
# Compute hash for the input data_graph
# Call process_graph function to build and cache the langchain_object
is_first_message = True
computed_hash, langchain_object = load_langchain_object(
basic_data_graph, is_first_message=is_first_message
)
save_cache(computed_hash, langchain_object, is_first_message)
# Check if the cache file exists
cache_file = Path(tempfile.gettempdir()) / f"{PREFIX}/{computed_hash}.dill"
assert cache_file.exists()
# Test load_or_build_langchain_object
def test_load_or_build_langchain_object_first_message_true(basic_data_graph):
build_langchain_object_with_caching.clear_cache()
graph = load_or_build_langchain_object(basic_data_graph, is_first_message=True)
assert graph is not None
def test_cache_reuse(basic_data_graph):
# Call process_graph function to build and cache the langchain_object
result1 = load_langchain_object(basic_data_graph)
def test_load_or_build_langchain_object_first_message_false(basic_data_graph):
graph = load_or_build_langchain_object(basic_data_graph, is_first_message=False)
assert graph is not None
# Call process_graph function again to use the cached langchain_object
result2 = load_langchain_object(basic_data_graph)
# Compare the results to ensure the same langchain_object was used
assert langchain_objects_are_equal(result1, result2)
# Test build_langchain_object_with_caching
def test_build_langchain_object_with_caching(basic_data_graph):
build_langchain_object_with_caching.clear_cache()
graph = build_langchain_object_with_caching(basic_data_graph)
assert graph is not None
# Test build_graph
def test_build_graph(basic_data_graph):
graph = build_graph(basic_data_graph)
assert graph is not None
assert len(graph.nodes) == len(basic_data_graph["nodes"])
assert len(graph.edges) == len(basic_data_graph["edges"])
# Test cache size limit
def test_cache_size_limit(basic_data_graph):
build_langchain_object_with_caching.clear_cache()
for i in range(11):
modified_data_graph = basic_data_graph.copy()
nodes = modified_data_graph["nodes"]
node_id = nodes[0]["id"]
# Now we replace all instances ode node_id with a new id in the json
json_string = json.dumps(modified_data_graph)
modified_json_string = json_string.replace(node_id, f"{node_id}_{i}")
modified_data_graph_new_id = json.loads(modified_json_string)
build_langchain_object_with_caching(modified_data_graph_new_id)
assert len(build_langchain_object_with_caching.cache) == 10

2
tests/test_chains_template.py
View file

@ -191,7 +191,7 @@ def test_llm_checker_chain(client: TestClient):
"multiline": False,
"password": False,
"name": "llm",
"type": "BaseLLM",
"type": "BaseLanguageModel",
"list": False,
"advanced": False,
}

103
tests/test_graph.py
View file

@ -1,7 +1,7 @@
from typing import Type, Union
import pytest
from langchain.agents import AgentExecutor
from langchain.chains.base import Chain
from langchain.llms.fake import FakeListLLM
from langflow.graph import Edge, Graph, Node
from langflow.graph.nodes import (
@ -15,7 +15,7 @@ from langflow.graph.nodes import (
WrapperNode,
)
from langflow.interface.run import get_result_and_steps
from langflow.utils.payload import build_json, get_root_node
from langflow.utils.payload import get_root_node
# Test cases for the graph module
@ -102,32 +102,13 @@ def test_get_node_neighbors_basic(basic_graph):
# We need to check if there is a Chain in the one of the neighbors'
# data attribute in the type key
assert any(
"Chain" in neighbor.data["type"] for neighbor, val in neighbors.items() if val
)
# assert Serper Search is in the neighbors
assert any(
"Serper" in neighbor.data["type"] for neighbor, val in neighbors.items() if val
)
# Now on to the Chain's neighbors
chain = next(
neighbor
"ConversationBufferMemory" in neighbor.data["type"]
for neighbor, val in neighbors.items()
if "Chain" in neighbor.data["type"] and val
)
chain_neighbors = basic_graph.get_node_neighbors(chain)
assert chain_neighbors is not None
assert isinstance(chain_neighbors, dict)
# Check if there is a LLM in the chain's neighbors
assert any(
"OpenAI" in neighbor.data["type"]
for neighbor, val in chain_neighbors.items()
if val
)
# Chain should have a Prompt as a neighbor
assert any(
"Prompt" in neighbor.data["type"]
for neighbor, val in chain_neighbors.items()
if val
"OpenAI" in neighbor.data["type"] for neighbor, val in neighbors.items() if val
)
@ -209,7 +190,7 @@ def test_get_root_node(basic_graph, complex_graph):
root = get_root_node(basic_graph)
assert root is not None
assert isinstance(root, Node)
assert root.data["type"] == "ZeroShotAgent"
assert root.data["type"] == "TimeTravelGuideChain"
# For complex example, the root node is a ZeroShotAgent too
assert isinstance(complex_graph, Graph)
root = get_root_node(complex_graph)
@ -218,26 +199,6 @@ def test_get_root_node(basic_graph, complex_graph):
assert root.data["type"] == "ZeroShotAgent"
def test_build_json(basic_graph):
"""Test building JSON from graph"""
assert isinstance(basic_graph, Graph)
root = get_root_node(basic_graph)
json_data = build_json(root, basic_graph)
assert isinstance(json_data, dict)
assert json_data["_type"] == "zero-shot-react-description"
assert isinstance(json_data["llm_chain"], dict)
assert json_data["llm_chain"]["_type"] == "llm_chain"
assert json_data["llm_chain"]["memory"] is None
assert json_data["llm_chain"]["verbose"] is False
assert isinstance(json_data["llm_chain"]["prompt"], dict)
assert isinstance(json_data["llm_chain"]["llm"], dict)
assert json_data["llm_chain"]["output_key"] == "text"
assert isinstance(json_data["allowed_tools"], list)
assert all(isinstance(tool, dict) for tool in json_data["allowed_tools"])
assert isinstance(json_data["return_values"], list)
assert all(isinstance(val, str) for val in json_data["return_values"])
def test_validate_edges(basic_graph):
"""Test validating edges"""
@ -269,45 +230,11 @@ def test_build_params(basic_graph):
assert all(edge.matched_type in edge.source_types for edge in basic_graph.edges)
# Get the root node
root = get_root_node(basic_graph)
# Root node is a ZeroShotAgent
# which requires an llm_chain, allowed_tools and return_values
# Root node is a TimeTravelGuideChain
# which requires an llm and memory
assert isinstance(root.params, dict)
assert "llm_chain" in root.params
assert "allowed_tools" in root.params
assert "return_values" in root.params
# The llm_chain should be a Node
assert isinstance(root.params["llm_chain"], Node)
# The allowed_tools should be a list of Nodes
assert isinstance(root.params["allowed_tools"], list)
assert all(isinstance(tool, Node) for tool in root.params["allowed_tools"])
# The return_values is of type str so it should be a list of strings
assert isinstance(root.params["return_values"], list)
assert all(isinstance(val, str) for val in root.params["return_values"])
# The llm_chain should have a prompt and llm
llm_chain_node = root.params["llm_chain"]
assert isinstance(llm_chain_node.params, dict)
assert "prompt" in llm_chain_node.params
assert "llm" in llm_chain_node.params
# The prompt should be a Node
assert isinstance(llm_chain_node.params["prompt"], Node)
# The llm should be a Node
assert isinstance(llm_chain_node.params["llm"], Node)
# The prompt should have format_insctructions, suffix, prefix
prompt_node = llm_chain_node.params["prompt"]
assert isinstance(prompt_node.params, dict)
assert "format_instructions" in prompt_node.params
assert "suffix" in prompt_node.params
assert "prefix" in prompt_node.params
# All of them should be of type str
assert isinstance(prompt_node.params["format_instructions"], str)
assert isinstance(prompt_node.params["suffix"], str)
assert isinstance(prompt_node.params["prefix"], str)
# The llm should have a model
llm_node = llm_chain_node.params["llm"]
assert isinstance(llm_node.params, dict)
assert "model_name" in llm_node.params
# The model should be a str
assert isinstance(llm_node.params["model_name"], str)
assert "llm" in root.params
assert "memory" in root.params
def test_build(basic_graph, complex_graph, openapi_graph):
@ -324,18 +251,18 @@ def assert_agent_was_built(graph):
# Build the Agent
result = graph.build()
# The agent should be a AgentExecutor
assert isinstance(result, AgentExecutor)
assert isinstance(result, Chain)
def test_agent_node_build(basic_graph):
agent_node = get_node_by_type(basic_graph, AgentNode)
def test_agent_node_build(complex_graph):
agent_node = get_node_by_type(complex_graph, AgentNode)
assert agent_node is not None
built_object = agent_node.build()
assert built_object is not None
def test_tool_node_build(basic_graph):
tool_node = get_node_by_type(basic_graph, ToolNode)
def test_tool_node_build(complex_graph):
tool_node = get_node_by_type(complex_graph, ToolNode)
assert tool_node is not None
built_object = tool_node.build()
assert built_object is not None

4
tests/test_loading.py
View file

@ -1,7 +1,7 @@
import json
import pytest
from langchain.agents import AgentExecutor
from langchain.chains.base import Chain
from langflow import load_flow_from_json
from langflow.graph import Graph
from langflow.utils.payload import get_root_node
@ -11,7 +11,7 @@ def test_load_flow_from_json():
"""Test loading a flow from a json file"""
loaded = load_flow_from_json(pytest.BASIC_EXAMPLE_PATH)
assert loaded is not None
assert isinstance(loaded, AgentExecutor)
assert isinstance(loaded, Chain)
def test_get_root_node():