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

Merge branch 'dev' into feat/aws

Browse source
This commit is contained in:
Yuki Sekiya 2023-10-18 22:43:44 +09:00 • committed by GitHub
commit ef7bd714b2
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
15 changed files with 4059 additions and 614 deletions
Download patch file Download diff file Expand all files Collapse all files

8
.env.example
View file

@ -56,6 +56,14 @@ LANGFLOW_REMOVE_API_KEYS=
# LANGFLOW_REDIS_CACHE_EXPIRE (default: 3600)
LANGFLOW_CACHE_TYPE=
# Auto login
# If set to true then a superuser will be logged in automatically
# and the login page will be skipped, keeping the
# default experience of Langflow
# Values: true, false
# Example: LANGFLOW_AUTO_LOGIN=true
LANGFLOW_AUTO_LOGIN=
# Superuser username
# Example: LANGFLOW_SUPERUSER=admin
LANGFLOW_SUPERUSER=

4
.gitignore vendored
View file

@ -254,4 +254,6 @@ langflow.db
/tmp/*
src/backend/langflow/frontend/
.docker
.docker
.idea

6
Makefile
View file

@ -23,7 +23,11 @@ tests:
poetry run pytest tests
tests_frontend:
cd src/frontend && ./run-tests.sh
ifeq ($(UI), true)
cd src/frontend && ./run-tests.sh --ui
else
cd src/frontend && ./run-tests.sh
endif
format:
poetry run black .

38
docs/docs/components/llms.mdx
View file

@ -217,4 +217,40 @@ Vertex AI is a cloud computing platform offered by Google Cloud Platform (GCP).
- **top_k:** How the model selects tokens for output, the next token is selected from – defaults to `40`.
- **top_p:** Tokens are selected from most probable to least until the sum of their – defaults to `0.95`.
- **tuned_model_name:** The name of a tuned model. If provided, model_name is ignored.
- **verbose:** This parameter is used to control the level of detail in the output of the chain. When set to True, it will print out some internal states of the chain while it is being run, which can help debug and understand the chain's behavior. If set to False, it will suppress the verbose output – defaults to `False`.
- **verbose:** This parameter is used to control the level of detail in the output of the chain. When set to True, it will print out some internal states of the chain while it is being run, which can help debug and understand the chain's behavior. If set to False, it will suppress the verbose output – defaults to `False`.
---
### QianfanLLMEndpoint
Wrapper around [Baidu Qianfan](https://cloud.baidu.com/doc/WENXINWORKSHOP/index.html) large language models.
:::info
The Qianfan Big Model Platform is a one-stop platform for enterprise developers to develop and operate large models and services. It provides data management based on ERNIE Bot's underlying model (Ernie Bot), automatic model customization and fine-tuning, and one-stop large-scale model customization services for cloud deployment of prediction services, and provides ERNIE Bot's enterprise level service API that can be quickly called, helping to implement the demand for generative AI applications in various industries.
:::
- **Model Name:** Model name. you could get from https://cloud.baidu.com/doc/WENXINWORKSHOP/s/Nlks5zkzu preset models are mapping to an endpoint. `Model Name` will be ignored if `Endpoint` is set.
- **Qianfan Ak:** which you could get from https://cloud.baidu.com/product/wenxinworkshop.
- **Qianfan Sk:** which you could get from https://cloud.baidu.com/product/wenxinworkshop.
- **Top p:** Model params, only supported in ERNIE-Bot and ERNIE-Bot-turbo. The diversity of the output text is affected, and the larger the value, the stronger the diversity of the generated text - defaults to `0.8`.
- **Temperature:** Model params, only supported in ERNIE-Bot and ERNIE-Bot-turbo. Higher values make the output more random, while lower values make it more concentrated and deterministic - defaults to `0.95`.
- **Penalty Score:** Model params, only supported in ERNIE-Bot and ERNIE-Bot-turbo. By increasing the penalty for generated tokens, the phenomenon of duplicate generation is reduced. A higher value indicates a higher penalty - defaults to `1.0`.
- **Endpoint:** Endpoint of the Qianfan LLM, required if custom model used.
---
### QianfanChatEndpoint
Wrapper around [Baidu Qianfan](https://cloud.baidu.com/doc/WENXINWORKSHOP/index.html) chat large language models. This component supports some of the LLMs (Large Language Models) available by Baidu qianfan and is used for tasks such as chatbots, Generative Question-Answering (GQA), and summarization.
:::info
The Qianfan Big Model Platform is a one-stop platform for enterprise developers to develop and operate large models and services. It provides data management based on ERNIE Bot's underlying model (Ernie Bot), automatic model customization and fine-tuning, and one-stop large-scale model customization services for cloud deployment of prediction services, and provides ERNIE Bot's enterprise level service API that can be quickly called, helping to implement the demand for generative AI applications in various industries.
:::
- **Model Name:** Model name. you could get from https://cloud.baidu.com/doc/WENXINWORKSHOP/s/Nlks5zkzu preset models are mapping to an endpoint. `Model Name` will be ignored if `Endpoint` is set.
- **Qianfan Ak:** which you could get from https://cloud.baidu.com/product/wenxinworkshop.
- **Qianfan Sk:** which you could get from https://cloud.baidu.com/product/wenxinworkshop.
- **Top p:** Model params, only supported in ERNIE-Bot and ERNIE-Bot-turbo. The diversity of the output text is affected, and the larger the value, the stronger the diversity of the generated text - defaults to `0.8`.
- **Temperature:** Model params, only supported in ERNIE-Bot and ERNIE-Bot-turbo. Higher values make the output more random, while lower values make it more concentrated and deterministic - defaults to `0.95`.
- **Penalty Score:** Model params, only supported in ERNIE-Bot and ERNIE-Bot-turbo. By increasing the penalty for generated tokens, the phenomenon of duplicate generation is reduced. A higher value indicates a higher penalty - defaults to `1.0`.
- **Endpoint:** Endpoint of the Qianfan LLM, required if custom model used.

1529
poetry.lock generated
View file

File diff suppressed because it is too large Load diff

5
pyproject.toml
View file

@ -33,7 +33,7 @@ google-search-results = "^2.4.1"
google-api-python-client = "^2.79.0"
typer = "^0.9.0"
gunicorn = "^21.2.0"
langchain = "^0.0.308"
langchain = "^0.0.312"
openai = "^0.27.8"
pandas = "2.0.3"
chromadb = "^0.3.21"
@ -91,7 +91,8 @@ pillow = "^10.0.0"
metal-sdk = "^2.2.0"
markupsafe = "^2.1.3"
boto3 = "^1.28.63"
numexpr = "^2.8.6"
qianfan = "0.0.5"
[tool.poetry.group.dev.dependencies]
types-redis = "^4.6.0.5"

92
src/backend/langflow/components/llms/BaiduQianfanChatEndpoints.py Normal file
View file

@ -0,0 +1,92 @@
from typing import Optional
from langflow import CustomComponent
from langchain.chat_models.baidu_qianfan_endpoint import QianfanChatEndpoint
from langchain.llms.base import BaseLLM
class QianfanChatEndpointComponent(CustomComponent):
display_name: str = "QianfanChatEndpoint"
description: str = (
"Baidu Qianfan chat models. Get more detail from "
"https://python.langchain.com/docs/integrations/chat/baidu_qianfan_endpoint."
)
def build_config(self):
return {
"model": {
"display_name": "Model Name",
"options": [
"ERNIE-Bot",
"ERNIE-Bot-turbo",
"BLOOMZ-7B",
"Llama-2-7b-chat",
"Llama-2-13b-chat",
"Llama-2-70b-chat",
"Qianfan-BLOOMZ-7B-compressed",
"Qianfan-Chinese-Llama-2-7B",
"ChatGLM2-6B-32K",
"AquilaChat-7B",
],
"info": "https://python.langchain.com/docs/integrations/chat/baidu_qianfan_endpoint",
"required": True,
},
"qianfan_ak": {
"display_name": "Qianfan Ak",
"required": True,
"password": True,
"info": "which you could get from https://cloud.baidu.com/product/wenxinworkshop",
},
"qianfan_sk": {
"display_name": "Qianfan Sk",
"required": True,
"password": True,
"info": "which you could get from https://cloud.baidu.com/product/wenxinworkshop",
},
"top_p": {
"display_name": "Top p",
"field_type": "float",
"info": "Model params, only supported in ERNIE-Bot and ERNIE-Bot-turbo",
"value": 0.8,
},
"temperature": {
"display_name": "Temperature",
"field_type": "float",
"info": "Model params, only supported in ERNIE-Bot and ERNIE-Bot-turbo",
"value": 0.95,
},
"penalty_score": {
"display_name": "Penalty Score",
"field_type": "float",
"info": "Model params, only supported in ERNIE-Bot and ERNIE-Bot-turbo",
"value": 1.0,
},
"endpoint": {
"display_name": "Endpoint",
"info": "Endpoint of the Qianfan LLM, required if custom model used.",
},
"code": {"show": False},
}
def build(
self,
model: str = "ERNIE-Bot-turbo",
qianfan_ak: Optional[str] = None,
qianfan_sk: Optional[str] = None,
top_p: Optional[float] = None,
temperature: Optional[float] = None,
penalty_score: Optional[float] = None,
endpoint: Optional[str] = None,
) -> BaseLLM:
try:
output = QianfanChatEndpoint( # type: ignore
model=model,
qianfan_ak=qianfan_ak,
qianfan_sk=qianfan_sk,
top_p=top_p,
temperature=temperature,
penalty_score=penalty_score,
endpoint=endpoint,
)
except Exception as e:
raise ValueError("Could not connect to Baidu Qianfan API.") from e
return output # type: ignore

92
src/backend/langflow/components/llms/BaiduQianfanLLMEndpoints.py Normal file
View file

@ -0,0 +1,92 @@
from typing import Optional
from langflow import CustomComponent
from langchain.llms.baidu_qianfan_endpoint import QianfanLLMEndpoint
from langchain.llms.base import BaseLLM
class QianfanLLMEndpointComponent(CustomComponent):
display_name: str = "QianfanLLMEndpoint"
description: str = (
"Baidu Qianfan hosted open source or customized models. "
"Get more detail from https://python.langchain.com/docs/integrations/chat/baidu_qianfan_endpoint"
)
def build_config(self):
return {
"model": {
"display_name": "Model Name",
"options": [
"ERNIE-Bot",
"ERNIE-Bot-turbo",
"BLOOMZ-7B",
"Llama-2-7b-chat",
"Llama-2-13b-chat",
"Llama-2-70b-chat",
"Qianfan-BLOOMZ-7B-compressed",
"Qianfan-Chinese-Llama-2-7B",
"ChatGLM2-6B-32K",
"AquilaChat-7B",
],
"info": "https://python.langchain.com/docs/integrations/chat/baidu_qianfan_endpoint",
"required": True,
},
"qianfan_ak": {
"display_name": "Qianfan Ak",
"required": True,
"password": True,
"info": "which you could get from https://cloud.baidu.com/product/wenxinworkshop",
},
"qianfan_sk": {
"display_name": "Qianfan Sk",
"required": True,
"password": True,
"info": "which you could get from https://cloud.baidu.com/product/wenxinworkshop",
},
"top_p": {
"display_name": "Top p",
"field_type": "float",
"info": "Model params, only supported in ERNIE-Bot and ERNIE-Bot-turbo",
"value": 0.8,
},
"temperature": {
"display_name": "Temperature",
"field_type": "float",
"info": "Model params, only supported in ERNIE-Bot and ERNIE-Bot-turbo",
"value": 0.95,
},
"penalty_score": {
"display_name": "Penalty Score",
"field_type": "float",
"info": "Model params, only supported in ERNIE-Bot and ERNIE-Bot-turbo",
"value": 1.0,
},
"endpoint": {
"display_name": "Endpoint",
"info": "Endpoint of the Qianfan LLM, required if custom model used.",
},
"code": {"show": False},
}
def build(
self,
model: str = "ERNIE-Bot-turbo",
qianfan_ak: Optional[str] = None,
qianfan_sk: Optional[str] = None,
top_p: Optional[float] = None,
temperature: Optional[float] = None,
penalty_score: Optional[float] = None,
endpoint: Optional[str] = None,
) -> BaseLLM:
try:
output = QianfanLLMEndpoint( # type: ignore
model=model,
qianfan_ak=qianfan_ak,
qianfan_sk=qianfan_sk,
top_p=top_p,
temperature=temperature,
penalty_score=penalty_score,
endpoint=endpoint,
)
except Exception as e:
raise ValueError("Could not connect to Baidu Qianfan API.") from e
return output # type: ignore

8
src/backend/langflow/components/vectorstores/Vectara.py
View file

@ -5,7 +5,6 @@ from langchain.vectorstores import Vectara
from langchain.schema import Document
from langchain.vectorstores.base import VectorStore
from langchain.schema import BaseRetriever
from langchain.embeddings.base import Embeddings
class VectaraComponent(CustomComponent):
@ -22,7 +21,6 @@ class VectaraComponent(CustomComponent):
"vectara_api_key": {"display_name": "Vectara API Key", "password": True},
"code": {"show": False},
"documents": {"display_name": "Documents"},
"embedding": {"display_name": "Embedding"},
}
def build(
@ -30,21 +28,21 @@ class VectaraComponent(CustomComponent):
vectara_customer_id: str,
vectara_corpus_id: str,
vectara_api_key: str,
embedding: Optional[Embeddings] = None,
documents: Optional[Document] = None,
) -> Union[VectorStore, BaseRetriever]:
# If documents, then we need to create a Vectara instance using .from_documents
if documents is not None and embedding is not None:
if documents is not None:
return Vectara.from_documents(
documents=documents, # type: ignore
vectara_customer_id=vectara_customer_id,
vectara_corpus_id=vectara_corpus_id,
vectara_api_key=vectara_api_key,
embedding=embedding,
source="langflow",
)
return Vectara(
vectara_customer_id=vectara_customer_id,
vectara_corpus_id=vectara_corpus_id,
vectara_api_key=vectara_api_key,
source="langflow",
)

2
src/backend/langflow/template/frontend_node/base.py
View file

@ -44,7 +44,7 @@ class FieldFormatters(BaseModel):
class FrontendNode(BaseModel):
template: Template
description: str
description: Optional[str] = None
base_classes: List[str]
name: str = ""
display_name: str = ""

5
src/frontend/src/contexts/tabsContext.tsx
View file

@ -260,9 +260,6 @@ export function TabsProvider({ children }: { children: ReactNode }) {
// simulate a click on the link element to trigger the download
link.click();
setNoticeData({
title: "Warning: Critical data, JSON file may include API keys.",
});
}
function downloadFlows() {
@ -300,7 +297,7 @@ export function TabsProvider({ children }: { children: ReactNode }) {
let fileData = JSON.parse(text);
if (fileData.flows) {
fileData.flows.forEach((flow: FlowType) => {
id = addFlow(flow, newProject);
id = addFlow(newProject, flow);
});
}
// parse the text into a JSON object

17
src/frontend/src/modals/exportModal/index.tsx
View file

@ -4,6 +4,7 @@ import IconComponent from "../../components/genericIconComponent";
import { Button } from "../../components/ui/button";
import { Checkbox } from "../../components/ui/checkbox";
import { EXPORT_DIALOG_SUBTITLE } from "../../constants/constants";
import { alertContext } from "../../contexts/alertContext";
import { TabsContext } from "../../contexts/tabsContext";
import { removeApiKeys } from "../../utils/reactflowUtils";
import BaseModal from "../baseModal";
@ -11,7 +12,8 @@ import BaseModal from "../baseModal";
const ExportModal = forwardRef(
(props: { children: ReactNode }, ref): JSX.Element => {
const { flows, tabId, downloadFlow } = useContext(TabsContext);
const [checked, setChecked] = useState(false);
const { setNoticeData } = useContext(alertContext);
const [checked, setChecked] = useState(true);
const flow = flows.find((f) => f.id === tabId);
useEffect(() => {
setName(flow!.name);
@ -44,6 +46,7 @@ const ExportModal = forwardRef(
<div className="mt-3 flex items-center space-x-2">
<Checkbox
id="terms"
checked={checked}
onCheckedChange={(event: boolean) => {
setChecked(event);
}}
@ -52,18 +55,26 @@ const ExportModal = forwardRef(
Save with my API keys
</label>
</div>
<span className="text-xs text-destructive">
Caution: Uncheck this box only removes API keys from fields
specifically designated for API keys.
</span>
</BaseModal.Content>
<BaseModal.Footer>
<Button
onClick={() => {
if (checked)
if (checked) {
downloadFlow(
flows.find((flow) => flow.id === tabId)!,
name!,
description
);
else
setNoticeData({
title:
"Warning: Critical data, JSON file may include API keys.",
});
} else
downloadFlow(
removeApiKeys(flows.find((flow) => flow.id === tabId)!),
name!,

2684
src/frontend/tests/onlyFront/assets/collection.json Normal file
View file

File diff suppressed because it is too large Load diff

96
src/frontend/tests/onlyFront/assets/flow.json Normal file
View file

@ -0,0 +1,96 @@
{
"description": "Engineered for Excellence, Built for Business.",
"name": "Fluffy Sinoussi",
"data": {
"nodes": [
{
"id": "AgentInitializer-Zza0A",
"type": "genericNode",
"position": { "x": 595, "y": 224.25 },
"data": {
"type": "AgentInitializer",
"node": {
"template": {
"llm": {
"required": true,
"placeholder": "",
"show": true,
"multiline": false,
"password": false,
"name": "llm",
"display_name": "LLM",
"advanced": false,
"dynamic": false,
"info": "",
"type": "BaseLanguageModel",
"list": false
},
"memory": {
"required": false,
"placeholder": "",
"show": true,
"multiline": false,
"password": false,
"name": "memory",
"advanced": false,
"dynamic": false,
"info": "",
"type": "BaseChatMemory",
"list": false
},
"tools": {
"required": true,
"placeholder": "",
"show": true,
"multiline": false,
"password": false,
"name": "tools",
"advanced": false,
"dynamic": false,
"info": "",
"type": "Tool",
"list": true
},
"agent": {
"required": true,
"placeholder": "",
"show": true,
"multiline": false,
"value": "zero-shot-react-description",
"password": false,
"options": [
"zero-shot-react-description",
"react-docstore",
"self-ask-with-search",
"conversational-react-description",
"openai-functions",
"openai-multi-functions"
],
"name": "agent",
"advanced": false,
"dynamic": false,
"info": "",
"type": "str",
"list": true
},
"_type": "initialize_agent"
},
"description": "Construct a zero shot agent from an LLM and tools.",
"base_classes": ["AgentExecutor", "function"],
"display_name": "AgentInitializer",
"custom_fields": {},
"output_types": [],
"documentation": "https://python.langchain.com/docs/modules/agents/agent_types/",
"beta": false,
"error": null
},
"id": "AgentInitializer-Zza0A"
},
"positionAbsolute": { "x": 595, "y": 224.25 }
}
],
"edges": [],
"viewport": { "x": 0, "y": 0, "zoom": 1 }
},
"id": "84c4b46f-063b-4d48-bf7f-6c668013064f"
}

87
src/frontend/tests/onlyFront/dragAndDrop.spec.ts Normal file
View file

@ -0,0 +1,87 @@
import { expect, test } from "@playwright/test";
import { readFileSync } from "fs";
test.describe("drag and drop test", () => {
/// <reference lib="dom"/>
test("drop collection", async ({ page }) => {
await page.routeFromHAR("harFiles/langflow.har", {
url: "**/api/v1/**",
update: false,
});
await page.route("**/api/v1/flows/", async (route) => {
const json = {
id: "e9ac1bdc-429b-475d-ac03-d26f9a2a3210",
};
await route.fulfill({ json, status: 201 });
});
await page.goto("http:localhost:3000/");
await page.locator("span").filter({ hasText: "My Collection" }).isVisible();
// Read your file into a buffer.
const jsonContent = readFileSync(
"tests/onlyFront/assets/collection.json",
"utf-8"
);
// Create the DataTransfer and File
const dataTransfer = await page.evaluateHandle((data) => {
const dt = new DataTransfer();
// Convert the buffer to a hex array
const file = new File([data], "collection.json", {
type: "application/json",
});
dt.items.add(file);
return dt;
}, jsonContent);
// Now dispatch
await page.dispatchEvent('//*[@id="root"]/div/div[2]/div[2]', "drop", {
dataTransfer,
});
expect(
await page
.locator(".main-page-flows-display")
.evaluate((el) => el.children)
).toBeTruthy();
});
test("drop flow", async ({ page }) => {
await page.routeFromHAR("harFiles/langflow.har", {
url: "**/api/v1/**",
update: false,
});
await page.route("**/api/v1/flows/", async (route) => {
const json = {
id: "e9ac1bdc-429b-475d-ac03-d26f9a2a3210",
};
await route.fulfill({ json, status: 201 });
});
await page.goto("http:localhost:3000/");
await page.locator("span").filter({ hasText: "My Collection" }).isVisible();
// Read your file into a buffer.
const jsonContent = readFileSync(
"tests/onlyFront/assets/flow.json",
"utf-8"
);
// Create the DataTransfer and File
const dataTransfer = await page.evaluateHandle((data) => {
const dt = new DataTransfer();
// Convert the buffer to a hex array
const file = new File([data], "flow.json", {
type: "application/json",
});
dt.items.add(file);
return dt;
}, jsonContent);
// Now dispatch
await page.dispatchEvent('//*[@id="root"]/div/div[2]/div[2]', "drop", {
dataTransfer,
});
expect(
await page
.locator(".main-page-flows-display")
.evaluate((el) => el.children)
).toBeTruthy();
});
});