llm_weaver/architecture.rs
1/// The following diagram shows a very high level slimmed down overview of the
2/// architecture of the library and how an application might use it.
3///
4/// Only the traits [`Loom`](crate::Loom) and [`Config`](crate::Config) are expanded to show some of
5/// their main associated types.
6#[cfg_attr(doc, aquamarine::aquamarine)]
7/// ```mermaid
8/// graph TB
9/// subgraph Chat Application
10/// App
11/// chat_gpt[Chat GPT]
12/// bard[Bard]
13/// end
14/// App-. impl .- Loom
15/// App-. impl .- Config
16/// chat_gpt[Chat GPT]-. impl .- llm
17/// bard[Bard]-. impl .- llm
18/// subgraph LLM Weaver
19/// llm>Llm]
20/// subgraph Config
21/// prompt_model[PromptModel]-- prompt --> chat_gpt
22/// summary_model[SummaryModel]-- prompt --> bard
23/// tapestry_chest_type[Chest]
24/// end
25/// subgraph Loom
26/// weave-- save prompt and response --> tapestry_chest_type
27/// weave-- generate summary --> summary_model
28/// weave-- generate response --> prompt_model
29/// end
30/// tapestry_chest_handler>TapestryChestHandler]
31/// tapestry_chest[TapestryChest]-. default impl .- tapestry_chest_handler
32/// tapestry_chest_type --> tapestry_chest
33/// tapestry_chest --> redis
34/// redis[Redis]
35/// end
36/// ```
37///
38/// The application must implement the [`Loom`](crate::Loom) and [`Config`](crate::Config) traits in
39/// order to utilize the library. This includes but is not limited to providing the types that
40/// implement the [`Llm`](crate::Llm) trait which defines the LLMs which will be used to
41/// prompt and generate summaries.
42///
43/// The [`Config`](crate::Config) trait also requires the application to supply an implementation
44/// for [`Chest`](crate::Config::Chest) which is responsible for storing and retrieving the
45/// [`TapestryFragment`](crate::TapestryFragment)s, but is not required since llm_weaver provides a
46/// default implementation that uses Redis as the storage backend.
47pub struct Diagram;