Loaders & Vector Store

LarkDocLoader

Load Feishu documents and Wiki pages into Synaptic [Document]s for RAG pipelines.

use synaptic::lark::{LarkConfig, LarkDocLoader};
use synaptic::core::Loader;

let config = LarkConfig::new("cli_xxx", "secret_xxx");

// Load specific document tokens
let loader = LarkDocLoader::new(config.clone())
    .with_doc_tokens(vec!["doxcnAbcXxx".to_string()]);

// Or traverse an entire Wiki space
let loader = LarkDocLoader::new(config)
    .with_wiki_space_id("spcXxx");

let docs = loader.load().await?;
for doc in &docs {
    println!("Title: {}", doc.metadata["title"]);
    println!("URL:   {}", doc.metadata["url"]);
    println!("Length: {} chars", doc.content.len());
}

Document Metadata

Each document includes:

Field	Description
`doc_id`	The Feishu document token
`title`	Document title
`source`	`lark:doc:<token>`
`url`	Direct Feishu document URL
`doc_type`	Always `"docx"`

Builder Options

Method	Description
`with_doc_tokens(tokens)`	Load specific document tokens
`with_wiki_space_id(id)`	Traverse all docs in a Wiki space

LarkWikiLoader

Recursively load all pages from a Feishu Wiki space as Documents. The with_space_id and with_max_depth builder methods control which space is traversed and how deep to recurse.

use synaptic::lark::{LarkConfig, LarkWikiLoader};
use synaptic::core::Loader;

let config = LarkConfig::new("cli_xxx", "secret_xxx");
let loader = LarkWikiLoader::new(config)
    .with_space_id("spcXxx")
    .with_max_depth(3);

let docs = loader.load().await?;
println!("Loaded {} Wiki pages", docs.len());

LarkDriveLoader

Load files from a Feishu Drive folder, automatically dispatching to the appropriate sub-loader (doc, spreadsheet, etc.) based on file type.

use synaptic::lark::{LarkConfig, LarkDriveLoader};
use synaptic::core::Loader;

let config = LarkConfig::new("cli_xxx", "secret_xxx");
let loader = LarkDriveLoader::new(config, "fldcnXxx");

let docs = loader.load().await?;
for doc in &docs {
    println!("{}: {} chars", doc.metadata["file_name"], doc.content.len());
}

LarkSpreadsheetLoader

Load rows from a Feishu spreadsheet as Synaptic [Document]s. Each row becomes one document; column headers are stored as metadata keys.

use synaptic::lark::{LarkConfig, LarkSpreadsheetLoader};
use synaptic::core::Loader;

let config = LarkConfig::new("cli_xxx", "secret_xxx");
let loader = LarkSpreadsheetLoader::new(config, "shtcnXxx", "0");

let docs = loader.load().await?;
for doc in &docs {
    println!("Row: {}", doc.content);
    println!("Sheet: {}", doc.metadata["sheet_id"]);
}

LarkVectorStore

Store and search vectors using the Feishu Search API as the backend. Feishu handles embedding on the server side; your documents are indexed in Lark and retrieved via semantic search.

use synaptic::lark::{LarkConfig, LarkVectorStore};
use synaptic::core::VectorStore;

let config = LarkConfig::new("cli_xxx", "secret_xxx");
let store = LarkVectorStore::new(config, "data_source_id_xxx");

// Index documents
store.add_documents(docs).await?;

// Semantic search — embedding is handled by the Feishu platform
let results = store.similarity_search("quarterly earnings", 5).await?;
for doc in &results {
    println!("{}", doc.content);
}

Keyboard shortcuts

Synaptic