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

Implement Memories, validation and other fixes

Browse source
This commit is contained in:
Gabriel Luiz Freitas Almeida 2023-04-05 07:08:45 -03:00 • committed by GitHub
commit 54add63b40
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
29 changed files with 525 additions and 121 deletions
Download patch file Download diff file Expand all files Collapse all files

15
Makefile
View file

@ -41,14 +41,13 @@ build:
dev:
make install_frontend
ifeq ($(build),1)
@echo 'Running docker compose up with build'
docker compose up --build
else
@echo 'Running docker compose up without build'
docker compose up
endif
ifeq ($(build),1)
@echo 'Running docker compose up with build'
docker compose up $(if $(debug),-f docker-compose.debug.yml) --build
else
@echo 'Running docker compose up without build'
docker compose up $(if $(debug),-f docker-compose.debug.yml)
endif
publish:
make build

28
docker-compose.debug.yml Normal file
View file

@ -0,0 +1,28 @@
version: '3.4'
services:
backend:
volumes:
- ./:/app
build:
context: ./
dockerfile: ./dev.Dockerfile
command: ["sh", "-c", "pip install debugpy -t /tmp && python /tmp/debugpy --wait-for-client --listen 0.0.0.0:5678 -m uvicorn langflow.main:app --host 0.0.0.0 --port 7860 --reload"]
ports:
- 7860:7860
- 5678:5678
restart: on-failure
frontend:
build:
context: ./src/frontend
dockerfile: ./dev.Dockerfile
args:
- BACKEND_URL=http://backend:7860
ports:
- "3000:3000"
volumes:
- ./src/frontend/public:/home/node/app/public
- ./src/frontend/src:/home/node/app/src
- ./src/frontend/package.json:/home/node/app/package.json
restart: on-failure

19
poetry.lock generated
View file

@ -1199,14 +1199,14 @@ test = ["ipykernel", "pre-commit", "pytest", "pytest-cov", "pytest-timeout"]
[[package]]
name = "langchain"
version = "0.0.127"
version = "0.0.131"
description = "Building applications with LLMs through composability"
category = "main"
optional = false
python-versions = ">=3.8.1,<4.0"
files = [
{file = "langchain-0.0.127-py3-none-any.whl", hash = "sha256:04ba053881e6098e80e0f4afc8922f3fe78923b160fd12d856aebce49c261918"},
{file = "langchain-0.0.127.tar.gz", hash = "sha256:e8a3b67fd86a6f79c4334f0a7588c9476fcb57b27a8fb0e617f47c01eaab8be8"},
{file = "langchain-0.0.131-py3-none-any.whl", hash = "sha256:3564a759e85095c9d71a78817da9cec1e2a8a0cda1bdd94ef8ac7008e432717a"},
{file = "langchain-0.0.131.tar.gz", hash = "sha256:61baf67fbec561ce38d187915a46e1c41139270826453600951760fde1a5d98a"},
]
[package.dependencies]
@ -1220,8 +1220,11 @@ SQLAlchemy = ">=1,<2"
tenacity = ">=8.1.0,<9.0.0"
[package.extras]
all = ["aleph-alpha-client (>=2.15.0,<3.0.0)", "anthropic (>=0.2.4,<0.3.0)", "beautifulsoup4 (>=4,<5)", "boto3 (>=1.26.96,<2.0.0)", "cohere (>=3,<4)", "deeplake (>=3.2.9,<4.0.0)", "elasticsearch (>=8,<9)", "faiss-cpu (>=1,<2)", "google-api-python-client (==2.70.0)", "google-search-results (>=2,<3)", "huggingface_hub (>=0,<1)", "jina (>=3.14,<4.0)", "jinja2 (>=3,<4)", "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)", "pgvector (>=0.1.6,<0.2.0)", "pinecone-client (>=2,<3)", "psycopg2-binary (>=2.9.5,<3.0.0)", "pyowm (>=3.3.0,<4.0.0)", "pypdf (>=3.4.0,<4.0.0)", "qdrant-client (>=1.0.4,<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,<2)", "transformers (>=4,<5)", "weaviate-client (>=3,<4)", "wikipedia (>=1,<2)", "wolframalpha (==5.0.0)"]
all = ["aleph-alpha-client (>=2.15.0,<3.0.0)", "anthropic (>=0.2.4,<0.3.0)", "beautifulsoup4 (>=4,<5)", "boto3 (>=1.26.96,<2.0.0)", "cohere (>=3,<4)", "deeplake (>=3.2.9,<4.0.0)", "elasticsearch (>=8,<9)", "faiss-cpu (>=1,<2)", "google-api-python-client (==2.70.0)", "google-search-results (>=2,<3)", "huggingface_hub (>=0,<1)", "jina (>=3.14,<4.0)", "jinja2 (>=3,<4)", "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)", "pgvector (>=0.1.6,<0.2.0)", "pinecone-client (>=2,<3)", "psycopg2-binary (>=2.9.5,<3.0.0)", "pyowm (>=3.3.0,<4.0.0)", "pypdf (>=3.4.0,<4.0.0)", "qdrant-client (>=1.1.1,<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,<2)", "transformers (>=4,<5)", "weaviate-client (>=3,<4)", "wikipedia (>=1,<2)", "wolframalpha (==5.0.0)"]
cohere = ["cohere (>=3,<4)"]
llms = ["anthropic (>=0.2.4,<0.3.0)", "cohere (>=3,<4)", "huggingface_hub (>=0,<1)", "manifest-ml (>=0.0.1,<0.0.2)", "nlpcloud (>=1,<2)", "openai (>=0,<1)", "torch (>=1,<2)", "transformers (>=4,<5)"]
openai = ["openai (>=0,<1)"]
qdrant = ["qdrant-client (>=1.1.1,<2.0.0)"]
[[package]]
name = "markdown-it-py"
@ -1506,14 +1509,14 @@ files = [
[[package]]
name = "openai"
version = "0.27.2"
version = "0.27.4"
description = "Python client library for the OpenAI API"
category = "main"
optional = false
python-versions = ">=3.7.1"
files = [
{file = "openai-0.27.2-py3-none-any.whl", hash = "sha256:6df674cf257e9e0504f1fd191c333d3f6a2442b13218d0eccf06230eb24d320e"},
{file = "openai-0.27.2.tar.gz", hash = "sha256:5869fdfa34b0ec66c39afa22f4a0fb83a135dff81f6505f52834c6ab3113f762"},
{file = "openai-0.27.4-py3-none-any.whl", hash = "sha256:3b82c867d531e1fd2003d9de2131e1c4bfd4c70b1a3149e0543a555b30807b70"},
{file = "openai-0.27.4.tar.gz", hash = "sha256:9f9d27d26e62c6068f516c0729449954b5ef6994be1a6cbfe7dbefbc84423a04"},
]
[package.dependencies]
@ -2712,4 +2715,4 @@ testing = ["big-O", "flake8 (<5)", "jaraco.functools", "jaraco.itertools", "more
[metadata]
lock-version = "2.0"
python-versions = "^3.9"
content-hash = "2b523f3d737ef8f7082e8156f096bce6f4f84a8bee9d07bd4ed23a29d3dcfab1"
content-hash = "91c68c5a5673f7b2bd0833af35da1262afd21d631cc62ec6ff9c65f69a96af0a"

2
pyproject.toml
View file

@ -29,7 +29,7 @@ google-search-results = "^2.4.1"
google-api-python-client = "^2.79.0"
typer = "^0.7.0"
gunicorn = "^20.1.0"
langchain = "^0.0.127"
langchain = "^0.0.131"
openai = "^0.27.2"
types-pyyaml = "^6.0.12.8"
dill = "^0.3.6"

12
src/backend/langflow/api/base.py
View file

@ -4,13 +4,13 @@ from pydantic import BaseModel, validator
class Code(BaseModel):
code: str
@validator("code")
def validate_code(cls, v):
return v
class Prompt(BaseModel):
template: str
# Build ValidationResponse class for {"imports": {"errors": []}, "function": {"errors": []}}
class ValidationResponse(BaseModel):
class CodeValidationResponse(BaseModel):
imports: dict
function: dict
@ -21,3 +21,7 @@ class ValidationResponse(BaseModel):
@validator("function")
def validate_function(cls, v):
return v or {"errors": []}
class PromptValidationResponse(BaseModel):
input_variables: list

14
src/backend/langflow/api/endpoints.py
View file

@ -3,10 +3,8 @@ from typing import Any, Dict
from fastapi import APIRouter, HTTPException
from langflow.api.base import Code, ValidationResponse
from langflow.interface.run import process_graph
from langflow.interface.types import build_langchain_types_dict
from langflow.utils.validate import validate_code
# build router
router = APIRouter()
@ -26,15 +24,3 @@ def get_load(data: Dict[str, Any]):
# Log stack trace
logger.exception(e)
raise HTTPException(status_code=500, detail=str(e)) from e
@router.post("/validate", status_code=200, response_model=ValidationResponse)
def post_validate_code(code: Code):
try:
errors = validate_code(code.code)
return ValidationResponse(
imports=errors.get("imports", {}),
function=errors.get("function", {}),
)
except Exception as e:
return HTTPException(status_code=500, detail=str(e))

36
src/backend/langflow/api/validate.py Normal file
View file

@ -0,0 +1,36 @@
from fastapi import APIRouter, HTTPException
from langflow.api.base import (
Code,
CodeValidationResponse,
Prompt,
PromptValidationResponse,
)
from langflow.graph.utils import extract_input_variables_from_prompt
from langflow.utils.logger import logger
from langflow.utils.validate import validate_code
# build router
router = APIRouter(prefix="/validate", tags=["validate"])
@router.post("/code", status_code=200, response_model=CodeValidationResponse)
def post_validate_code(code: Code):
try:
errors = validate_code(code.code)
return CodeValidationResponse(
imports=errors.get("imports", {}),
function=errors.get("function", {}),
)
except Exception as e:
return HTTPException(status_code=500, detail=str(e))
@router.post("/prompt", status_code=200, response_model=PromptValidationResponse)
def post_validate_prompt(prompt: Prompt):
try:
input_variables = extract_input_variables_from_prompt(prompt.template)
return PromptValidationResponse(input_variables=input_variables)
except Exception as e:
logger.exception(e)
return HTTPException(status_code=500, detail=str(e))

10
src/backend/langflow/config.yaml
View file

@ -3,6 +3,9 @@ chains:
- LLMMathChain
- LLMCheckerChain
- ConversationChain
- SeriesCharacterChain
- MidJourneyPromptChain
- TimeTravelGuideChain
agents:
- ZeroShotAgent
@ -13,9 +16,15 @@ agents:
prompts:
- PromptTemplate
- FewShotPromptTemplate
- ZeroShotPrompt
# Wait more tests
# - ChatPromptTemplate
# - SystemMessagePromptTemplate
# - HumanMessagePromptTemplate
llms:
- OpenAI
- AzureOpenAI
- ChatOpenAI
tools:
@ -36,6 +45,7 @@ toolkits:
memories:
- ConversationBufferMemory
- ConversationSummaryMemory
embeddings: []

1
src/backend/langflow/custom/customs.py
View file

@ -1,5 +1,6 @@
from langflow.template import nodes
# These should always be instantiated
CUSTOM_NODES = {
"prompts": {"ZeroShotPrompt": nodes.ZeroShotPromptNode()},
"tools": {"PythonFunction": nodes.PythonFunctionNode(), "Tool": nodes.ToolNode()},

2
src/backend/langflow/graph/nodes.py
View file

@ -75,7 +75,9 @@ class PromptNode(Node):
for param in prompt_params:
prompt_text = self.params[param]
variables = extract_input_variables_from_prompt(prompt_text)
self.params["input_variables"].extend(variables)
self.params["input_variables"] = list(set(self.params["input_variables"]))
self._build()
return deepcopy(self._built_object)

3
src/backend/langflow/interface/agents/base.py
View file

@ -18,7 +18,8 @@ class AgentCreator(LangChainTypeCreator):
self.type_dict = loading.AGENT_TO_CLASS
# Add JsonAgent to the list of agents
for name, agent in CUSTOM_AGENTS.items():
self.type_dict[name] = agent
# TODO: validate AgentType
self.type_dict[name] = agent # type: ignore
return self.type_dict
def get_signature(self, name: str) -> Optional[Dict]:

3
src/backend/langflow/interface/agents/custom.py
View file

@ -111,7 +111,8 @@ class InitializeAgent(AgentExecutor):
return initialize_agent(
tools=tools,
llm=llm,
agent=agent,
# LangChain now uses Enum for agent, but we still support string
agent=agent, # type: ignore
memory=memory,
return_intermediate_steps=True,
)

9
src/backend/langflow/interface/base.py
View file

@ -1,5 +1,5 @@
from abc import ABC, abstractmethod
from typing import Any, Dict, List, Optional, Union
from typing import Any, Dict, List, Optional, Type, Union
from pydantic import BaseModel
@ -12,6 +12,11 @@ class LangChainTypeCreator(BaseModel, ABC):
type_name: str
type_dict: Optional[Dict] = None
@property
def frontend_node_class(self) -> Type[FrontendNode]:
"""The class type of the FrontendNode created in frontend_node."""
return FrontendNode
@property
@abstractmethod
def type_to_loader_dict(self) -> Dict:
@ -62,7 +67,7 @@ class LangChainTypeCreator(BaseModel, ABC):
if key != "_type"
]
template = Template(type_name=name, fields=fields)
return FrontendNode(
return self.frontend_node_class(
template=template,
description=signature.get("description", ""),
base_classes=signature["base_classes"],

19
src/backend/langflow/interface/chains/base.py
View file

@ -1,5 +1,6 @@
from typing import Dict, List, Optional
from langflow.custom.customs import get_custom_nodes
from langflow.interface.base import LangChainTypeCreator
from langflow.interface.custom_lists import chain_type_to_cls_dict
from langflow.settings import settings
@ -15,19 +16,27 @@ class ChainCreator(LangChainTypeCreator):
def type_to_loader_dict(self) -> Dict:
if self.type_dict is None:
self.type_dict = chain_type_to_cls_dict
from langflow.interface.chains.custom import CUSTOM_CHAINS
self.type_dict.update(CUSTOM_CHAINS)
return self.type_dict
def get_signature(self, name: str) -> Optional[Dict]:
try:
return build_template_from_class(name, chain_type_to_cls_dict)
if name in get_custom_nodes(self.type_name).keys():
return get_custom_nodes(self.type_name)[name]
return build_template_from_class(name, self.type_to_loader_dict)
except ValueError as exc:
raise ValueError("Memory not found") from exc
raise ValueError("Chain not found") from exc
def to_list(self) -> List[str]:
custom_chains = list(get_custom_nodes("chains").keys())
default_chains = list(self.type_to_loader_dict.keys())
# Check if the chain is in the settings
return [
chain.__name__
for chain in self.type_to_loader_dict.values()
if chain.__name__ in settings.chains or settings.dev
chain
for chain in default_chains + custom_chains
if chain in settings.chains or settings.dev
]

101
src/backend/langflow/interface/chains/custom.py Normal file
View file

@ -0,0 +1,101 @@
from typing import Dict, Optional, Type
from langchain.chains import ConversationChain
from langchain.memory.buffer import ConversationBufferMemory
from langchain.schema import BaseMemory
from pydantic import Field, root_validator
from langflow.graph.utils import extract_input_variables_from_prompt
DEFAULT_SUFFIX = """"
Current conversation:
{history}
Human: {input}
{ai_prefix}"""
class BaseCustomChain(ConversationChain):
"""BaseCustomChain is a chain you can use to have a conversation with a custom character."""
template: Optional[str]
ai_prefix_key: Optional[str]
"""Field to use as the ai_prefix. It needs to be set and has to be in the template"""
@root_validator(pre=False)
def build_template(cls, values):
format_dict = {}
input_variables = extract_input_variables_from_prompt(values["template"])
if values.get("ai_prefix_key", None) is None:
values["ai_prefix_key"] = values["memory"].ai_prefix
for key in input_variables:
new_value = values.get(key, f"{{{key}}}")
format_dict[key] = new_value
if key == values.get("ai_prefix_key", None):
values["memory"].ai_prefix = new_value
values["template"] = values["template"].format(**format_dict)
values["template"] = values["template"]
values["input_variables"] = extract_input_variables_from_prompt(
values["template"]
)
values["prompt"].template = values["template"]
values["prompt"].input_variables = values["input_variables"]
return values
class SeriesCharacterChain(BaseCustomChain):
"""SeriesCharacterChain is a chain you can use to have a conversation with a character from a series."""
character: str
series: str
template: Optional[
str
] = """I want you to act like {character} from {series}.
I want you to respond and answer like {character}. do not write any explanations. only answer like {character}.
You must know all of the knowledge of {character}.
Current conversation:
{history}
Human: {input}
{character}:"""
memory: BaseMemory = Field(default_factory=ConversationBufferMemory)
ai_prefix_key: Optional[str] = "character"
"""Default memory store."""
class MidJourneyPromptChain(BaseCustomChain):
"""MidJourneyPromptChain is a chain you can use to generate new MidJourney prompts."""
template: Optional[
str
] = """I want you to act as a prompt generator for Midjourney's artificial intelligence program.
Your job is to provide detailed and creative descriptions that will inspire unique and interesting images from the AI.
Keep in mind that the AI is capable of understanding a wide range of language and can interpret abstract concepts, so feel free to be as imaginative and descriptive as possible.
For example, you could describe a scene from a futuristic city, or a surreal landscape filled with strange creatures.
The more detailed and imaginative your description, the more interesting the resulting image will be. Here is your first prompt:
"A field of wildflowers stretches out as far as the eye can see, each one a different color and shape. In the distance, a massive tree towers over the landscape, its branches reaching up to the sky like tentacles.\"
Current conversation:
{history}
Human: {input}
AI:""" # noqa: E501
class TimeTravelGuideChain(BaseCustomChain):
template: Optional[
str
] = """I want you to act as my time travel guide. You are helpful and creative. I will provide you with the historical period or future time I want to visit and you will suggest the best events, sights, or people to experience. Provide the suggestions and any necessary information.
Current conversation:
{history}
Human: {input}
AI:""" # noqa: E501
CUSTOM_CHAINS: Dict[str, Type[ConversationChain]] = {
"SeriesCharacterChain": SeriesCharacterChain,
"MidJourneyPromptChain": MidJourneyPromptChain,
"TimeTravelGuideChain": TimeTravelGuideChain,
}

14
src/backend/langflow/interface/importing/utils.py
View file

@ -1,7 +1,7 @@
# This module is used to import any langchain class by name.
import importlib
from typing import Any
from typing import Any, Type
from langchain import PromptTemplate
from langchain.agents import Agent
@ -65,10 +65,14 @@ def import_class(class_path: str) -> Any:
return getattr(module, class_name)
def import_prompt(prompt: str) -> PromptTemplate:
def import_prompt(prompt: str) -> Type[PromptTemplate]:
from langflow.interface.prompts.custom import CUSTOM_PROMPTS
"""Import prompt from prompt name"""
if prompt == "ZeroShotPrompt":
return import_class("langchain.prompts.PromptTemplate")
elif prompt in CUSTOM_PROMPTS:
return CUSTOM_PROMPTS[prompt]
return import_class(f"langchain.prompts.{prompt}")
@ -100,6 +104,10 @@ def import_tool(tool: str) -> BaseTool:
return get_tool_by_name(tool)
def import_chain(chain: str) -> Chain:
def import_chain(chain: str) -> Type[Chain]:
"""Import chain from chain name"""
from langflow.interface.chains.custom import CUSTOM_CHAINS
if chain in CUSTOM_CHAINS:
return CUSTOM_CHAINS[chain]
return import_class(f"langchain.chains.{chain}")

9
src/backend/langflow/interface/memories/base.py
View file

@ -1,14 +1,21 @@
from typing import Dict, List, Optional
from typing import Dict, List, Optional, Type
from langflow.interface.base import LangChainTypeCreator
from langflow.interface.custom_lists import memory_type_to_cls_dict
from langflow.settings import settings
from langflow.template.base import FrontendNode
from langflow.template.nodes import MemoryFrontendNode
from langflow.utils.util import build_template_from_class
class MemoryCreator(LangChainTypeCreator):
type_name: str = "memories"
@property
def frontend_node_class(self) -> Type[FrontendNode]:
"""The class type of the FrontendNode created in frontend_node."""
return MemoryFrontendNode
@property
def type_to_loader_dict(self) -> Dict:
if self.type_dict is None:

39
src/backend/langflow/interface/prompts/base.py
View file

@ -1,39 +1,54 @@
from typing import Dict, List, Optional
from typing import Dict, List, Optional, Type
from langchain.prompts import loading
from langchain import prompts
from langflow.custom.customs import get_custom_nodes
from langflow.interface.base import LangChainTypeCreator
from langflow.interface.importing.utils import import_class
from langflow.settings import settings
from langflow.utils.util import build_template_from_function
from langflow.template.nodes import PromptFrontendNode
from langflow.utils.util import build_template_from_class
class PromptCreator(LangChainTypeCreator):
type_name: str = "prompts"
@property
def frontend_node_class(self) -> Type[PromptFrontendNode]:
return PromptFrontendNode
@property
def type_to_loader_dict(self) -> Dict:
if self.type_dict is None:
self.type_dict = loading.type_to_loader_dict
self.type_dict = {
prompt_name: import_class(f"langchain.prompts.{prompt_name}")
# if prompt_name is not lower case it is a class
for prompt_name in prompts.__all__
if not prompt_name.islower() and prompt_name in settings.prompts
}
# Merge CUSTOM_PROMPTS into self.type_dict
from langflow.interface.prompts.custom import CUSTOM_PROMPTS
self.type_dict.update(CUSTOM_PROMPTS)
return self.type_dict
def get_signature(self, name: str) -> Optional[Dict]:
try:
if name in get_custom_nodes(self.type_name).keys():
return get_custom_nodes(self.type_name)[name]
return build_template_from_function(name, self.type_to_loader_dict)
return build_template_from_class(name, self.type_to_loader_dict)
except ValueError as exc:
raise ValueError("Prompt not found") from exc
def to_list(self) -> List[str]:
custom_prompts = get_custom_nodes("prompts")
library_prompts = [
prompt.__annotations__["return"].__name__
for prompt in self.type_to_loader_dict.values()
if prompt.__annotations__["return"].__name__ in settings.prompts
or settings.dev
]
return library_prompts + list(custom_prompts.keys())
# library_prompts = [
# prompt.__annotations__["return"].__name__
# for prompt in self.type_to_loader_dict.values()
# if prompt.__annotations__["return"].__name__ in settings.prompts
# or settings.dev
# ]
return list(self.type_to_loader_dict.keys()) + list(custom_prompts.keys())
prompt_creator = PromptCreator()

66
src/backend/langflow/interface/prompts/custom.py
View file

@ -1,49 +1,52 @@
from typing import List, Optional
from typing import Dict, List, Optional, Type
from langchain.prompts import PromptTemplate
from pydantic import root_validator
from langflow.graph.utils import extract_input_variables_from_prompt
from langflow.template.base import Template, TemplateField
from langflow.template.nodes import PromptTemplateNode
CHARACTER_PROMPT = """I want you to act like {character} from {series}.
I want you to respond and answer like {character}. do not write any explanations. only answer like {character}.
You must know all of the knowledge of {character}."""
# Steps to create a BaseCustomPrompt:
# 1. Create a prompt template that endes with:
# Current conversation:
# {history}
# Human: {input}
# {ai_prefix}:
# 2. Create a class that inherits from BaseCustomPrompt
# 3. Add the following class attributes:
# template: str = ""
# description: Optional[str]
# ai_prefix: Optional[str] = "{ai_prefix}"
# 3.1. The ai_prefix should be a value in input_variables
# SeriesCharacterPrompt is a working example
# If used in a LLMChain, with a Memory module, it will work as expected
# We should consider creating ConversationalChains that expose custom parameters
# That way it will be easier to create custom prompts
class BaseCustomPrompt(PromptTemplate):
template: str = ""
description: Optional[str]
human_text: str = "\n {input}"
ai_prefix: Optional[str]
@root_validator(pre=False)
def build_template(cls, values):
format_dict = {}
ai_prefix_format_dict = {}
for key in values.get("input_variables", []):
new_value = values[key]
new_value = values.get(key, f"{{{key}}}")
format_dict[key] = new_value
if key in values["ai_prefix"]:
ai_prefix_format_dict[key] = new_value
values["ai_prefix"] = values["ai_prefix"].format(**ai_prefix_format_dict)
values["template"] = values["template"].format(**format_dict)
values["template"] = values["template"] + values["human_text"]
values["template"] = values["template"]
values["input_variables"] = extract_input_variables_from_prompt(
values["template"]
)
return values
def build_frontend_node(self) -> PromptTemplateNode:
return PromptTemplateNode(
template=Template(
type_name="test",
fields=[
TemplateField(name=field, field_type="str", required=True)
for field in self.input_variables
],
),
description=self.description or "",
)
class SeriesCharacterPrompt(BaseCustomPrompt):
# Add a very descriptive description for the prompt generator
@ -52,14 +55,23 @@ class SeriesCharacterPrompt(BaseCustomPrompt):
] = "A prompt that asks the AI to act like a character from a series."
character: str
series: str
human_text: str = "\n {input}"
template: str = CHARACTER_PROMPT
template: str = """I want you to act like {character} from {series}.
I want you to respond and answer like {character}. do not write any explanations. only answer like {character}.
You must know all of the knowledge of {character}.
Current conversation:
{history}
Human: {input}
{character}:"""
ai_prefix: str = "{character}"
input_variables: List[str] = ["character", "series"]
CUSTOM_PROMPTS: Dict[str, Type[BaseCustomPrompt]] = {
"SeriesCharacterPrompt": SeriesCharacterPrompt
}
if __name__ == "__main__":
prompt = SeriesCharacterPrompt(character="Walter White", series="Breaking Bad")
user_input = "I am the one who knocks"
full_prompt = prompt.format(input=user_input)
print(full_prompt)
prompt = SeriesCharacterPrompt(character="Harry Potter", series="Harry Potter")
print(prompt.template)

84
src/backend/langflow/interface/run.py
View file

@ -52,6 +52,12 @@ def process_graph(data_graph: Dict[str, Any]):
)
logger.debug("Loaded langchain object")
if langchain_object is None:
# Raise user facing error
raise ValueError(
"There was an error loading the langchain_object. Please, check all the nodes and try again."
)
# Generate result and thought
logger.debug("Generating result and thought")
result, thought = get_result_and_thought_using_graph(langchain_object, message)
@ -66,34 +72,82 @@ def process_graph(data_graph: Dict[str, Any]):
return {"result": str(result), "thought": thought.strip()}
def get_result_and_thought_using_graph(loaded_langchain, message: str):
def fix_memory_inputs(langchain_object):
"""
Fix memory inputs by replacing the memory key with the input key.
"""
# Possible memory keys
# "chat_history", "history"
# if memory_key is "chat_history" and input_keys has "history"
# we need to replace "chat_history" with "history"
mem_key_dict = {
"chat_history": "history",
"history": "chat_history",
}
memory_key = langchain_object.memory.memory_key
possible_new_mem_key = mem_key_dict.get(memory_key)
if possible_new_mem_key is not None:
# get input_key
input_key = [
key
for key in langchain_object.input_keys
if key not in [memory_key, possible_new_mem_key]
][0]
# get output_key
output_key = [
key
for key in langchain_object.output_keys
if key not in [memory_key, possible_new_mem_key]
][0]
# set input_key and output_key in memory
langchain_object.memory.input_key = input_key
langchain_object.memory.output_key = output_key
for input_key in langchain_object.input_keys:
if input_key == possible_new_mem_key:
langchain_object.memory.memory_key = possible_new_mem_key
def get_result_and_thought_using_graph(langchain_object, message: str):
"""Get result and thought from extracted json"""
try:
loaded_langchain.verbose = True
if hasattr(langchain_object, "verbose"):
langchain_object.verbose = True
with io.StringIO() as output_buffer, contextlib.redirect_stdout(output_buffer):
chat_input = None
for key in loaded_langchain.input_keys:
if key == "chat_history" and hasattr(loaded_langchain, "memory"):
loaded_langchain.memory.memory_key = "chat_history"
else:
memory_key = ""
if (
hasattr(langchain_object, "memory")
and langchain_object.memory is not None
):
memory_key = langchain_object.memory.memory_key
for key in langchain_object.input_keys:
if key not in [memory_key, "chat_history"]:
chat_input = {key: message}
if hasattr(loaded_langchain, "return_intermediate_steps"):
if hasattr(langchain_object, "return_intermediate_steps"):
# https://github.com/hwchase17/langchain/issues/2068
loaded_langchain.return_intermediate_steps = False
# Deactivating until we have a frontend solution
# to display intermediate steps
langchain_object.return_intermediate_steps = False
fix_memory_inputs(langchain_object)
try:
output = loaded_langchain(chat_input)
output = langchain_object(chat_input)
except ValueError as exc:
logger.debug("Error: %s", str(exc))
output = loaded_langchain.run(chat_input)
# make the error message more informative
logger.debug(f"Error: {str(exc)}")
output = langchain_object.run(chat_input)
intermediate_steps = (
output.get("intermediate_steps", []) if isinstance(output, dict) else []
)
result = (
output.get(loaded_langchain.output_keys[0])
output.get(langchain_object.output_keys[0])
if isinstance(output, dict)
else output
)
@ -110,16 +164,16 @@ def get_result_and_thought_using_graph(loaded_langchain, message: str):
def get_result_and_thought(extracted_json: Dict[str, Any], message: str):
"""Get result and thought from extracted json"""
try:
loaded_langchain = loading.load_langchain_type_from_config(
langchain_object = loading.load_langchain_type_from_config(
config=extracted_json
)
with io.StringIO() as output_buffer, contextlib.redirect_stdout(output_buffer):
output = loaded_langchain(message)
output = langchain_object(message)
intermediate_steps = (
output.get("intermediate_steps", []) if isinstance(output, dict) else []
)
result = (
output.get(loaded_langchain.output_keys[0])
output.get(langchain_object.output_keys[0])
if isinstance(output, dict)
else output
)

2
src/backend/langflow/interface/wrappers/base.py
View file

@ -13,7 +13,7 @@ class WrapperCreator(LangChainTypeCreator):
def type_to_loader_dict(self) -> Dict:
if self.type_dict is None:
self.type_dict = {
wrapper.__name__: wrapper for wrapper in [requests.RequestsWrapper]
wrapper.__name__: wrapper for wrapper in [requests.TextRequestsWrapper]
}
return self.type_dict

2
src/backend/langflow/main.py
View file

@ -2,6 +2,7 @@ from fastapi import FastAPI
from fastapi.middleware.cors import CORSMiddleware
from langflow.api.endpoints import router as endpoints_router
from langflow.api.validate import router as validate_router
def create_app():
@ -21,6 +22,7 @@ def create_app():
)
app.include_router(endpoints_router)
app.include_router(validate_router)
return app

1
src/backend/langflow/template/base.py
View file

@ -22,6 +22,7 @@ class TemplateFieldCreator(BaseModel, ABC):
password: bool = False
options: list[str] = []
name: str = ""
display_name: Optional[str] = None
def to_dict(self):
result = self.dict()

21
src/backend/langflow/template/constants.py
View file

@ -9,3 +9,24 @@ FORCE_SHOW_FIELDS = [
"max_value_length",
"max_tokens",
]
DEFAULT_PROMPT = """
I want you to act as a naming consultant for new companies.
Here are some examples of good company names:
- search engine, Google
- social media, Facebook
- video sharing, YouTube
The name should be short, catchy and easy to remember.
What is a good name for a company that makes {product}?
"""
SYSTEM_PROMPT = """
You are a helpful assistant that talks casually about life in general.
You are a good listener and you can talk about anything.
"""
HUMAN_PROMPT = "{input}"

57
src/backend/langflow/template/nodes.py
View file

@ -4,10 +4,27 @@ from langchain.agents import loading
from langchain.agents.mrkl import prompt
from langflow.template.base import FrontendNode, Template, TemplateField
from langflow.template.constants import DEFAULT_PROMPT, HUMAN_PROMPT, SYSTEM_PROMPT
from langflow.utils.constants import DEFAULT_PYTHON_FUNCTION
NON_CHAT_AGENTS = {
agent_type: agent_class
for agent_type, agent_class in loading.AGENT_TO_CLASS.items()
if "chat" not in agent_type.value
}
class ZeroShotPromptNode(FrontendNode):
class BasePromptFrontendNode(FrontendNode):
name: str
template: Template
description: str
base_classes: list[str]
def to_dict(self):
return super().to_dict()
class ZeroShotPromptNode(BasePromptFrontendNode):
name: str = "ZeroShotPrompt"
template: Template = Template(
type_name="zero_shot",
@ -165,8 +182,8 @@ class InitializeAgentNode(FrontendNode):
is_list=True,
show=True,
multiline=False,
options=list(loading.AGENT_TO_CLASS.keys()),
value=list(loading.AGENT_TO_CLASS.keys())[0],
options=list(NON_CHAT_AGENTS.keys()),
value=list(NON_CHAT_AGENTS.keys())[0],
name="agent",
),
TemplateField(
@ -229,3 +246,37 @@ class CSVAgentNode(FrontendNode):
def to_dict(self):
return super().to_dict()
class PromptFrontendNode(FrontendNode):
@staticmethod
def format_field(field: TemplateField, name: Optional[str] = None) -> None:
# if field.field_type == "StringPromptTemplate"
# change it to str
if field.field_type == "StringPromptTemplate" and "Message" in str(name):
field.field_type = "str"
field.multiline = True
field.value = HUMAN_PROMPT if "Human" in field.name else SYSTEM_PROMPT
if field.name == "template" and field.value == "":
field.value = DEFAULT_PROMPT
if (
"Union" in field.field_type
and "BaseMessagePromptTemplate" in field.field_type
):
field.field_type = "BaseMessagePromptTemplate"
class MemoryFrontendNode(FrontendNode):
@staticmethod
def format_field(field: TemplateField, name: Optional[str] = None) -> None:
FrontendNode.format_field(field, name)
if not isinstance(field.value, str):
field.value = None
if field.name == "k":
field.required = True
field.show = True
field.field_type = "int"
field.value = 10
field.display_name = "Memory Size"

2
src/frontend/package.json
View file

@ -60,4 +60,4 @@
]
},
"proxy": "http://backend:7860"
}
}

2
src/frontend/src/controllers/API/index.ts
View file

@ -12,5 +12,5 @@ export async function sendAll(data:sendAllProps) {
export async function checkCode(code:string):Promise<AxiosResponse<errorsTypeAPI>>{
return await axios.post('/validate',{code})
return await axios.post('/validate/code',{code})
}

14
tests/data/Openapi.json
View file

@ -74,7 +74,7 @@
"multiline": false,
"password": false,
"name": "requests_wrapper",
"type": "RequestsWrapper",
"type": "TextRequestsWrapper",
"list": false
},
"_type": "OpenAPIToolkit"
@ -154,7 +154,7 @@
"y": 532.9920887988924
},
"data": {
"type": "RequestsWrapper",
"type": "TextRequestsWrapper",
"node": {
"template": {
"headers": {
@ -178,11 +178,11 @@
"type": "ClientSession",
"list": false
},
"_type": "RequestsWrapper"
"_type": "TextRequestsWrapper"
},
"description": "Lightweight wrapper around requests library.",
"base_classes": [
"RequestsWrapper"
"TextRequestsWrapper"
]
},
"id": "dndnode_34",
@ -405,11 +405,11 @@
},
{
"source": "dndnode_34",
"sourceHandle": "RequestsWrapper|dndnode_34|RequestsWrapper",
"sourceHandle": "TextRequestsWrapper|dndnode_34|TextRequestsWrapper",
"target": "dndnode_32",
"targetHandle": "RequestsWrapper|requests_wrapper|dndnode_32",
"targetHandle": "TextRequestsWrapper|requests_wrapper|dndnode_32",
"className": "animate-pulse",
"id": "reactflow__edge-dndnode_34RequestsWrapper|dndnode_34|RequestsWrapper-dndnode_32RequestsWrapper|requests_wrapper|dndnode_32",
"id": "reactflow__edge-dndnode_34RequestsWrapper|dndnode_34|TextRequestsWrapper-dndnode_32RequestsWrapper|requests_wrapper|dndnode_32",
"selected": false
},
{

59
tests/test_endpoints.py
View file

@ -1,3 +1,4 @@
import pytest
from fastapi.testclient import TestClient
from langflow.interface.tools.constants import CUSTOM_TOOLS
@ -20,7 +21,7 @@ import math
def square(x):
return x ** 2
"""
response1 = client.post("/validate", json={"code": code1})
response1 = client.post("/validate/code", json={"code": code1})
assert response1.status_code == 200
assert response1.json() == {"imports": {"errors": []}, "function": {"errors": []}}
@ -31,7 +32,7 @@ import non_existent_module
def square(x):
return x ** 2
"""
response2 = client.post("/validate", json={"code": code2})
response2 = client.post("/validate/code", json={"code": code2})
assert response2.status_code == 200
assert response2.json() == {
"imports": {"errors": ["No module named 'non_existent_module'"]},
@ -45,7 +46,7 @@ import math
def square(x)
return x ** 2
"""
response3 = client.post("/validate", json={"code": code3})
response3 = client.post("/validate/code", json={"code": code3})
assert response3.status_code == 200
assert response3.json() == {
"imports": {"errors": []},
@ -53,11 +54,11 @@ def square(x)
}
# Test case with invalid JSON payload
response4 = client.post("/validate", json={"invalid_key": code1})
response4 = client.post("/validate/code", json={"invalid_key": code1})
assert response4.status_code == 422
# Test case with an empty code string
response5 = client.post("/validate", json={"code": ""})
response5 = client.post("/validate/code", json={"code": ""})
assert response5.status_code == 200
assert response5.json() == {"imports": {"errors": []}, "function": {"errors": []}}
@ -68,9 +69,55 @@ import math
def square(x)
return x ** 2
"""
response6 = client.post("/validate", json={"code": code6})
response6 = client.post("/validate/code", json={"code": code6})
assert response6.status_code == 200
assert response6.json() == {
"imports": {"errors": []},
"function": {"errors": ["expected ':' (<unknown>, line 4)"]},
}
VALID_PROMPT = """
I want you to act as a naming consultant for new companies.
Here are some examples of good company names:
- search engine, Google
- social media, Facebook
- video sharing, YouTube
The name should be short, catchy and easy to remember.
What is a good name for a company that makes {product}?
"""
INVALID_PROMPT = "This is an invalid prompt without any input variable."
def test_valid_prompt(client: TestClient):
response = client.post("/validate/prompt", json={"template": VALID_PROMPT})
assert response.status_code == 200
assert response.json() == {"input_variables": ["product"]}
def test_invalid_prompt(client: TestClient):
response = client.post("/validate/prompt", json={"template": INVALID_PROMPT})
assert response.status_code == 200
assert response.json() == {"input_variables": []}
@pytest.mark.parametrize(
"prompt,expected_input_variables",
[
("{color} is my favorite color.", ["color"]),
("The weather is {weather} today.", ["weather"]),
("This prompt has no variables.", []),
("{a}, {b}, and {c} are variables.", ["a", "b", "c"]),
],
)
def test_various_prompts(client, prompt, expected_input_variables):
response = client.post("/validate/prompt", json={"template": prompt})
assert response.status_code == 200
assert response.json() == {
"input_variables": expected_input_variables,
}