Artifact | Documentation

The Artifact component is a data component that allows users to manipulate and smart search files and data in the artifact store. It can carry out the following tasks:

Upload File
Upload Files
Get Files Metadata
Get Chunks Metadata
Get File In Markdown
Match File Status
Retrieve
Ask

To use Artifact Component, you will need to set up the OpenAI API key for self-hosted deployment of Instill Core. You can do this by setting the OPENAI_API_KEY environment variable. Please refer to configuring-the-embedding-feature p.s. In Instill Cloud case, you do not need to set up the OpenAI API key.

#Release Stage

Alpha

#Configuration

The component configuration is defined and maintained here.

#Supported Tasks

#Upload File

Upload and process the files into chunks into Catalog

Input	ID	Type	Description
Task ID (required)	`task`	string	`TASK_UPLOAD_FILE`
Options (required)	`options`	object	Choose to upload the files to existing catalog or create a new catalog

Output	ID	Type	Description
File	`file`	object	Result of uploading file into catalog
Status	`status`	boolean	The status of trigger file processing, if succeeded, return true

#Upload Files

Upload and process the files into chunks into Catalog

Input	ID	Type	Description
Task ID (required)	`task`	string	`TASK_UPLOAD_FILES`
Options (required)	`options`	object	Choose to upload the files to existing catalog or create a new catalog

Output	ID	Type	Description
Files	`files`	array[object]	Files metadata in catalog
Status	`status`	boolean	The status of trigger file processing, if ALL succeeded, return true

#Get Files Metadata

get the metadata of the files in the catalog

Input	ID	Type	Description
Task ID (required)	`task`	string	`TASK_GET_FILES_METADATA`
Namespace (required)	`namespace`	string	Fill in your namespace, you can get namespace through the tab of switching namespace
Catalog ID (required)	`catalog-id`	string	Catalog ID that you input to search files in the Catalog

Output	ID	Type	Description
Files	`files`	array[object]	Files metadata in catalog

#Get Chunks Metadata

get the metadata of the chunks from a file in the catalog

Input	ID	Type	Description
Task ID (required)	`task`	string	`TASK_GET_CHUNKS_METADATA`
Catalog ID (required)	`catalog-id`	string	Catalog ID that you input to search files in the Catalog
Namespace (required)	`namespace`	string	Fill in your namespace, you can get namespace through the tab of switching namespace
File UID (required)	`file-uid`	string	The unique identifier of the file

Output	ID	Type	Description
Chunks	`chunks`	array[object]	Chunks metadata of the file in catalog

#Get File In Markdown

get the file content in markdown format

Input	ID	Type	Description
Task ID (required)	`task`	string	`TASK_GET_FILE_IN_MARKDOWN`
Catalog ID (required)	`catalog-id`	string	Catalog ID that you input to search files in the Catalog
Namespace (required)	`namespace`	string	Fill in your namespace, you can get namespace through the tab of switching namespace
File UID (required)	`file-uid`	string	The unique identifier of the file

Output	ID	Type	Description
File UID	`original-file-uid`	string	The unique identifier of the file
Content	`content`	string	The content of the file in markdown format
Create Time	`create-time`	string	TThe creation time of the source file in ISO 8601 format
Update Time	`update-time`	string	The update time of the source file in ISO 8601 format

#Match File Status

Check if the specified file's processing status is done

Input	ID	Type	Description
Task ID (required)	`task`	string	`TASK_MATCH_FILE_STATUS`
Catalog ID (required)	`catalog-id`	string	Catalog ID that you input to check files' processing status in the Catalog
Namespace (required)	`namespace`	string	Fill in your namespace, you can get namespace through the tab of switching namespace
File UID (required)	`file-uid`	string	The unique identifier of the file

Output	ID	Type	Description
Status	`succeeded`	boolean	The status of the file processing, if succeeded, return true

#Retrieve

search the chunks in the catalog

Input	ID	Type	Description
Task ID (required)	`task`	string	`TASK_RETRIEVE`
Catalog ID (required)	`catalog-id`	string	Catalog ID that you input to search files in the Catalog
Namespace (required)	`namespace`	string	Fill in your namespace, you can get namespace through the tab of switching namespace
Text Prompt (required)	`text-prompt`	string	The prompt string to search the chunks
Top K	`top-k`	integer	The number of top chunks to return. The range is from 1~20, and default is 5

Output	ID	Type	Description
Chunks	`chunks`	array[object]	Chunks data from smart search

#Ask

Reply the questions based on the files in the catalog

Input	ID	Type	Description
Task ID (required)	`task`	string	`TASK_ASK`
Catalog ID (required)	`catalog-id`	string	Catalog ID that you input to search files in the Catalog
Namespace (required)	`namespace`	string	Fill in your namespace, you can get namespace through the tab of switching namespace
Question (required)	`question`	string	The question to reply
Top K	`top-k`	integer	The number of top answers to return. The range is from 1~20, and default is 5

Output	ID	Type	Description
Answer	`answer`	string	Answers data from smart search
Chunks (optional)	`chunks`	array[object]	Chunks data to answer question

#Example Recipes

Recipe for the Ask your Catalog pipeline.

version: v1beta
component:
    artifact-0:
        type: artifact
        task: TASK_ASK
        input:
            catalog-id: ${variable.catalog_name}
            namespace: ${variable.namespace}
            question: ${variable.question}
            top-k: 5
variable:
    catalog_name:
        title: catalog-name
        description: The name of your catalog i.e. "instill-ai"
        instill-format: string
    namespace:
        title: namespace
        description: The namespace of your catalog i.e. "instill-ai"
        instill-format: string
    question:
        title: question
        description: The question to ask your catalog i.e. "What is Instill AI doing?", "What is Artifact?"
        instill-format: string
output:
    answer:
        title: answer
        value: ${artifact-0.output.answer}