Description

# banks [](https://pypi.org/project/banks) [](https://pypi.org/project/banks) [](https://coveralls.io/github/masci/banks?branch=main) [](https://github.com/masci/banks/actions/workflows/release.yml) [](https://github.com/masci/banks/actions/workflows/test.yml) [](https://github.com/masci/banks/actions/workflows/docs.yml) [](https://github.com/pypa/hatch) [](https://github.com/astral-sh/ruff) [](https://mypy-lang.org/) [](https://spdx.org/licenses/) [Banks](https://en.wikipedia.org/wiki/Arrival_(film)) is the linguist professor who will help you generate meaningful LLM prompts using a template language that makes sense. If you're still using `f-strings` for the job, keep reading. Docs are available [here](https://masci.github.io/banks/).  ----- **Table of Contents** - [banks](#banks) - [Installation](#installation) - [Features](#features) - [Cookbook](#cookbook) - [Examples](#examples) - [:point\_right: Render a prompt template as chat messages](#point_right-render-a-prompt-template-as-chat-messages) - [:point\_right: Add images to the prompt for vision models](#point_right-add-images-to-the-prompt-for-vision-models) - [:point\_right: Use a LLM to generate a text while rendering a prompt](#point_right-use-a-llm-to-generate-a-text-while-rendering-a-prompt) - [:point\_right: Function calling directly from the prompt](#point_right-function-calling-directly-from-the-prompt) - [:point\_right: Use prompt caching from Anthropic](#point_right-use-prompt-caching-from-anthropic) - [Reuse templates from registries](#reuse-templates-from-registries) - [Async support](#async-support) - [Contributing](#contributing) - [License](#license) ## Installation ```console pip install banks # install optional deps; litellm, redis pip install "banks[all]" ``` ## Features Prompts are instrumental for the success of any LLM application, and Banks focuses around specific areas of their lifecycle: - :orange_book: **Templating**: Banks provides tools and functions to build prompts text and chat messages from generic blueprints. - :tickets: **Versioning and metadata**: Banks supports attaching metadata to prompts to ease their management, and versioning is first-class citizen. - :file_cabinet: **Management**: Banks provides ways to store prompts on disk along with their metadata. ## Cookbook - :blue_book: [In-prompt chat completion](./cookbook/in_prompt_completion.ipynb) - :blue_book: [Prompt caching with Anthropic](./cookbook/Prompt_Caching_with_Anthropic.ipynb) - :blue_book: [Prompt versioning](./cookbook/Prompt_Versioning.ipynb) ## Examples For a more extensive set of code examples, [see the documentation page](https://masci.github.io/banks/examples/). ### :point_right: Render a prompt template as chat messages You'll find yourself feeding an LLM a list of chat messages instead of plain text more often than not. Banks will help you remove the boilerplate by defining the messages already at the prompt level. ```py from banks import Prompt prompt_template = """ {% chat role="system" %} You are a {{ persona }}. {% endchat %} {% chat role="user" %} Hello, how are you? {% endchat %} """ p = Prompt(prompt_template) print(p.chat_messages({"persona": "helpful assistant"})) # Output: # [ # ChatMessage(role='system', content=[ # ContentBlock(type=<ContentBlockType.text: 'text'>, cache_control=None, text='You are a helpful assistant.', # image_url=None, input_audio=None, input_video=None, input_document=None) # ], tool_call_id=None, name=None), # ChatMessage(role='user', content=[ # ContentBlock(type=<ContentBlockType.text: 'text'>, cache_control=None, text='Hello, how are you?', # image_url=None, input_audio=None, input_video=None, input_document=None) # ], tool_call_id=None, name=None) # ] ``` ### :point_right: Add images to the prompt for vision models If you're working with a multimodal model, you can include images directly in the prompt, and Banks will do the job needed to upload them when rendering the chat messages: ```py import litellm from banks import Prompt prompt_template = """ {% chat role="user" %} Gues