AI Flows
AI Flows enable developers to design distributed DAGs(Directed Acycling Graphs) to address complex requirements. This block helps in orchestrating multiple AI services and components to create comprehensive AI-driven workflows.
When do you need flows?
Flows are required when you have multiple steps to be executed one after the other in a consistent manner. Typical examples are RAG pipelines, which extract information from Vector databases and then forward it to LLMs and multi-step LLM execution like “Tree of thoughts” used in resolving complex prompts into simper prompts.
AI Flow Tasks
AI Flows are a workflows with a string of specific AI and Machine Learning tasks which are executed in sequence to achieve a specific outcome. Below are list of tasks that available on kis.ai platform out of the box.
List of AI Flow Tasks
| Name | Description |
|---|---|
| llm-chat | chat with llms local or remote, using standard Open AI chat completion API |
| text-to-image | convert text prompts to image through multi-modal LLMs and services |
| text-to-video | convert text prompts to video through multi-model LLMs and services |
| voice-to-text | convert voice to text, multi-voice and with diarization |
| text-to-voice | convert text to voice |
| scrape | submits a request to scraping service, waits for the response and passes the response to next task |
| ocr | converts and images or pdf to text |
| api | make an api call and return the response as json, yaml, text, csv or xlsx |
| embed-text | converts the text into vector embeddings |
| embed-doc | converts the document into vector embeddings |
| embed-image | converts images into vector embeddings |
| embed-video | converts videos into vector embeddings |
| semantic-embed | add embeddings to vector database |
| semantic-add-text | add text data to vector database after converting to embeddings |
| semantic-add-image | add image to vector database after converting to embeddings |
| semantic-add-video | add video to vector database after converting to embeddings |
| semantic-search | search on vector databases |
| lexical-add-text | add text to search index |
| lexical-add-doc | add document to search index |
| validate-json | validate json based on jsonschema |
| validate-yaml | validate yaml based on yaml schema |
| validate-code | validate code with parsers and compilers |
| validate-kis.ai | validate kis.ai formats |
| rag-add-topic | add content to a rag topic (lexical, semantic and knowedgegraph databases) |
| rag-update-topic | update content of a rag topic (lexical, semantic and knowledgraph databases) |
| rag-delete-topic | delete content of a rag topic |
| rag-search-topic | simple or hybrid search over a RAG topic |
| kg-text-ontology | generates ontology from given text and domain |
| kg-text-graph | generates a knowledge graph from the text |
| kg-validate-text | validates a text against a graph or ontology |
| kg-validate-graph | validates a given knowledge graph against parent graph or ontology |
| log | logs the message to observability logs. Useful to debug your flows |
| echo | echoes back the message you send, typically used to test the flow |