Choosing an API

Start with the smallest API that matches your task.

Decision table

Your task	Start with
Ask a model a question	akasha.ask()
Ask questions about an image	asker.vision()
Generate a new image	akasha.gen_image()
Edit an existing image	akasha.edit_image()
Answer questions about local documents	akasha.RAG()
Let the model call tools	akasha.agents()
Summarize a file or URL	akasha.summary()
Remember facts across conversations	MemoryManager

Example: the same question with and without retrieval

import akasha

# General model knowledge
qa = akasha.ask(model="gemini:gemini-2.5-flash")
print(qa("What is a vector store?"))

# Your documents as the source of context
rag = akasha.RAG(
    model="gemini:gemini-2.5-flash",
    embeddings="gemini:gemini-embedding-001",
)
print(rag("./docs", "What does our project say about vector stores?"))

If the answer depends on your files, use RAG instead of placing the entire file into a prompt yourself.