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

add support for VertexAIEmbeddings node

Browse source
This commit is contained in:
Dave Morris 2023-08-08 17:35:50 -05:00
commit bb2b8fbb3d
8 changed files with 121 additions and 4 deletions
Download patch file Download diff file Expand all files Collapse all files

1
.dockerignore Normal file
View file

@ -0,0 +1 @@
.venv/

22
docs/docs/components/embeddings.mdx
View file

@ -73,3 +73,25 @@ Used to load [OpenAI’s](https://openai.com/) embedding models.
- **request_timeout:** Used to specify the maximum amount of time, in milliseconds, to wait for a response from the OpenAI API when generating embeddings for a given text.
- **tiktoken_model_name:** Used to count the number of tokens in documents to constrain them to be under a certain limit. By default, when set to None, this will be the same as the embedding model name.
---
### VertexAIEmbeddings
Wrapper around [Google Vertex AI](https://cloud.google.com/vertex-ai) [Embeddings API](https://cloud.google.com/vertex-ai/docs/generative-ai/embeddings/get-text-embeddings).
:::info
Vertex AI is a cloud computing platform offered by Google Cloud Platform (GCP). It provides access, management, and development of applications and services through global data centers. To use Vertex AI PaLM, you need to have the [google-cloud-aiplatform](https://pypi.org/project/google-cloud-aiplatform/) Python package installed and credentials configured for your environment.
:::
- **credentials:** The default custom credentials (google.auth.credentials.Credentials) to use.
- **location:** The default location to use when making API calls – defaults to `us-central1`.
- **max_output_tokens:** Token limit determines the maximum amount of text output from one prompt – defaults to `128`.
- **model_name:** The name of the Vertex AI large language model – defaults to `text-bison`.
- **project:** The default GCP project to use when making Vertex API calls.
- **request_parallelism:** The amount of parallelism allowed for requests issued to VertexAI models – defaults to `5`.
- **temperature:** Tunes the degree of randomness in text generations. Should be a non-negative value – defaults to `0`.
- **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`.

22
docs/docs/components/llms.mdx
View file

@ -185,6 +185,28 @@ Wrapper around [Google Vertex AI](https://cloud.google.com/vertex-ai) large lang
Vertex AI is a cloud computing platform offered by Google Cloud Platform (GCP). It provides access, management, and development of applications and services through global data centers. To use Vertex AI PaLM, you need to have the [google-cloud-aiplatform](https://pypi.org/project/google-cloud-aiplatform/) Python package installed and credentials configured for your environment.
:::
- **credentials:** The default custom credentials (google.auth.credentials.Credentials) to use.
- **location:** The default location to use when making API calls – defaults to `us-central1`.
- **max_output_tokens:** Token limit determines the maximum amount of text output from one prompt – defaults to `128`.
- **model_name:** The name of the Vertex AI large language model – defaults to `text-bison`.
- **project:** The default GCP project to use when making Vertex API calls.
- **request_parallelism:** The amount of parallelism allowed for requests issued to VertexAI models – defaults to `5`.
- **temperature:** Tunes the degree of randomness in text generations. Should be a non-negative value – defaults to `0`.
- **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`.
---
### ChatVertexAI
Wrapper around [Google Vertex AI](https://cloud.google.com/vertex-ai) large language models.
:::info
Vertex AI is a cloud computing platform offered by Google Cloud Platform (GCP). It provides access, management, and development of applications and services through global data centers. To use Vertex AI PaLM, you need to have the [google-cloud-aiplatform](https://pypi.org/project/google-cloud-aiplatform/) Python package installed and credentials configured for your environment.
:::
- **credentials:** The default custom credentials (google.auth.credentials.Credentials) to use.
- **location:** The default location to use when making API calls – defaults to `us-central1`.
- **max_output_tokens:** Token limit determines the maximum amount of text output from one prompt – defaults to `128`.

1
src/backend/Dockerfile
View file

@ -12,4 +12,3 @@ RUN rm *.whl
EXPOSE 80
CMD [ "uvicorn", "--host", "0.0.0.0", "--port", "7860", "--factory", "langflow.main:create_app" ]

8
src/backend/langflow/interface/initialize/loading.py
View file

@ -88,7 +88,7 @@ def instantiate_based_on_type(class_object, base_type, node_type, params):
elif base_type == "toolkits":
return instantiate_toolkit(node_type, class_object, params)
elif base_type == "embeddings":
return instantiate_embedding(class_object, params)
return instantiate_embedding(node_type, class_object, params)
elif base_type == "vectorstores":
return instantiate_vectorstore(class_object, params)
elif base_type == "documentloaders":
@ -258,9 +258,13 @@ def instantiate_toolkit(node_type, class_object: Type[BaseToolkit], params: Dict
return loaded_toolkit
def instantiate_embedding(class_object, params: Dict):
def instantiate_embedding(node_type, class_object, params: Dict):
params.pop("model", None)
params.pop("headers", None)
if "VertexAI" in node_type:
return initialize_vertexai(class_object=class_object, params=params)
try:
return class_object(**params)
except ValidationError:

67
src/backend/langflow/template/frontend_node/embeddings.py
View file

@ -5,6 +5,47 @@ from langflow.template.frontend_node.base import FrontendNode
class EmbeddingFrontendNode(FrontendNode):
def add_extra_fields(self) -> None:
if "VertexAI" in self.template.type_name:
# Add credentials field which should of type file.
self.template.add_field(
TemplateField(
field_type="file",
required=False,
show=True,
name="credentials",
value="",
suffixes=[".json"],
file_types=["json"],
)
)
@staticmethod
def format_vertex_field(field: TemplateField, name: str):
if "VertexAI" in name:
advanced_fields = [
"verbose",
"top_p",
"top_k",
"max_output_tokens",
]
if field.name in advanced_fields:
field.advanced = True
show_fields = [
"verbose",
"project",
"location",
"credentials",
"max_output_tokens",
"model_name",
"temperature",
"top_p",
"top_k",
]
if field.name in show_fields:
field.show = True
@staticmethod
def format_jina_fields(field: TemplateField):
if "jina" in field.name:
@ -41,10 +82,36 @@ class EmbeddingFrontendNode(FrontendNode):
@staticmethod
def format_field(field: TemplateField, name: Optional[str] = None) -> None:
FrontendNode.format_field(field, name)
if name and "vertex" in name.lower():
EmbeddingFrontendNode.format_vertex_field(field, name)
field.advanced = not field.required
field.show = True
if field.name == "headers":
field.show = False
if field.name == "model_kwargs":
field.field_type = "code"
field.advanced = True
field.show = True
elif field.name in [
"model_name",
"temperature",
"model_file",
"model_type",
"deployment_name",
"credentials",
]:
field.advanced = False
field.show = True
if field.name == "credentials":
field.field_type = "file"
if name == "VertexAI" and field.name not in [
"callbacks",
"client",
"stop",
"tags",
"cache",
]:
field.show = True
# Format Jina fields
EmbeddingFrontendNode.format_jina_fields(field)

1
src/frontend/src/utils/styleUtils.ts
View file

@ -205,6 +205,7 @@ export const nodeIconsLucide = {
SupabaseVectorStore: SupabaseIcon,
VertexAI: VertexAIIcon,
ChatVertexAI: VertexAIIcon,
VertexAIEmbeddings: VertexAIIcon,
agents: Rocket,
WikipediaAPIWrapper: SvgWikipedia,
chains: Link,

3
src/frontend/vite.config.ts
View file

@ -5,6 +5,7 @@ const apiRoutes = ["^/api/v1/", "/health"];
// Use environment variable to determine the target.
const target = process.env.VITE_PROXY_TARGET || "http://127.0.0.1:7860";
const port = process.env.VITE_PROXY_PORT || 3000;
const proxyTargets = apiRoutes.reduce((proxyObj, route) => {
proxyObj[route] = {
@ -22,7 +23,7 @@ export default defineConfig(() => {
},
plugins: [react(), svgr()],
server: {
port: 3000,
port: port,
proxy: {
...proxyTargets,
},