Overview / StrayCat

StrayCat

The stray cat goes around tools and hook, making troubles

Constructors

new StrayCat()

new StrayCat(userId, ws?): StrayCat

Parameters

Parameter	Type
userId	string
ws?	ElysiaWS

Returns

StrayCat

Properties

Property	Modifier	Type	Default value
activeForm?	public	string	undefined
userId	public	string	undefined
workingMemory	public	WorkingMemory	undefined
wsQueue	public	WSMessage[]	[]

Accessors

agentManager

Get Signature

get agentManager(): AgentManager

Returns

AgentManager

---

currentEmbedder

Get Signature

get currentEmbedder(): Embeddings

Returns

Embeddings

---

currentLLM

Get Signature

get currentLLM(): BaseChatModel

Returns

BaseChatModel

---

lastUserMessage

Get Signature

get lastUserMessage(): Message

Returns

Message

---

plugins

Get Signature

get plugins(): object[]

Returns

object[]

---

rabbitHole

Get Signature

get rabbitHole(): RabbitHole

Returns

RabbitHole

---

vectorMemory

Get Signature

get vectorMemory(): VectorMemory

Returns

VectorMemory

---

whiteRabbit

Get Signature

get whiteRabbit(): WhiteRabbit

Returns

WhiteRabbit

Methods

addHistory()

addHistory(message): void

Adds messages to the chat history.

Parameters

Parameter	Type	Description
message	MemoryMessage[]	the messages to add

Returns

void

---

addInteraction()

addInteraction(interaction): void

Adds an interaction to the working memory.

Parameters

Parameter	Type	Description
interaction	ModelInteraction	the interaction to add

Returns

void

---

addWebSocket()

addWebSocket(value): void

This property is used to establish a new WebSocket connection.

Parameters

Parameter	Type	Description
value	undefined | ElysiaWS	The WebSocket instance.

Returns

void

---

classify()

classify<S, T>(sentence, labels, examples?): Promise<null | S>

Experimental

Classifies the given sentence into one of the provided labels.

Type Parameters

Type Parameter
S extends string
T extends [S, ...S[]]

Parameters

Parameter	Type	Description
sentence	string	The sentence to classify.
labels	T	The labels to classify the sentence into.
examples?	{ [key in string]: S[] }	Optional examples to help the LLM classify the sentence.

Returns

Promise<null | S>

The label of the sentence or null if it could not be classified.

---

clearHistory()

clearHistory(): void

Clears the chat history.

Returns

void

---

getHistory()

getHistory(k?): MemoryMessage[]

If passed a number k, retrieves the last k messages in the chat history. Otherwise, retrieves all messages in the chat history.

Parameters

Parameter	Type	Description
k?	number	the number of messages to retrieve

Returns

MemoryMessage[]

the messages present in the chat history

---

getInteraction()

getInteraction(k?): ModelInteraction[]

If passed a number k, retrieves the last k interactions in the working memory. Otherwise, retrieves all interactions in the working memory.

Parameters

Parameter	Type	Description
k?	number	the number of interactions to retrieve

Returns

ModelInteraction[]

the interactions present in the working memory

---

getPluginInfo()

getPluginInfo(id): undefined | { active: boolean; manifest: { authorName: string; authorUrl: string; description: string; name: string; pluginUrl: string; tags: string[]; thumb: string; version: string; }; settings: {}; }

Retrieves information about a plugin.

Parameters

Parameter	Type	Description
id	string	The ID of the plugin.

Returns

undefined | { active: boolean; manifest: { authorName: string; authorUrl: string; description: string; name: string; pluginUrl: string; tags: string[]; thumb: string; version: string; }; settings: {}; }

An object containing the plugin's active status, manifest, and settings.

Returns undefined if the plugin is not found.

---

llm()

Call Signature

llm(prompt, stream?): Promise<AIMessageChunk>

Executes the LLM with the given prompt and returns the response.

Parameters

Parameter	Type	Description
prompt	BaseLanguageModelInput	The prompt or messages to be passed to the LLM.
stream?	false	Optional parameter to enable streaming mode.

Returns

Promise<AIMessageChunk>

Call Signature

llm(prompt, stream?): Promise<IterableReadableStream<AIMessageChunk>>

Executes the LLM with the given prompt and returns the response.

Parameters

Parameter	Type	Description
prompt	BaseLanguageModelInput	The prompt or messages to be passed to the LLM.
stream?	true	Optional parameter to enable streaming mode.

Returns

Promise<IterableReadableStream<AIMessageChunk>>

---

queryDb()

queryDb<T>(question, type, source): Promise<string>

Experimental

Executes a SQL query based on a natural language question.

Type Parameters

Type Parameter
T extends "oracle" | "postgres" | "sqlite" | "mysql" | "mssql"

Parameters

Parameter	Type	Description
question	string	The user question.
type	T	The SQL dialect to use.
source	Omit<Extract<MysqlConnectionOptions, { type: T; }> | Extract<PostgresConnectionOptions, { type: T; }> | Extract<CockroachConnectionOptions, { type: T; }> | Extract<SqliteConnectionOptions, { type: T; }> | Extract<SqlServerConnectionOptions, { type: T; }> | Extract<SapConnectionOptions, { type: T; }> | Extract<OracleConnectionOptions, { type: T; }> | Extract<CordovaConnectionOptions, { type: T; }> | Extract<NativescriptConnectionOptions, { type: T; }> | Extract<ReactNativeConnectionOptions, { type: T; }> | Extract<SqljsConnectionOptions, { type: T; }> | Extract<MongoConnectionOptions, { type: T; }> | Extract<AuroraMysqlConnectionOptions, { type: T; }> | Extract<AuroraPostgresConnectionOptions, { type: T; }> | Extract<ExpoConnectionOptions, { type: T; }> | Extract<BetterSqlite3ConnectionOptions, { type: T; }> | Extract<CapacitorConnectionOptions, { type: T; }> | Extract<SpannerConnectionOptions, { type: T; }>, "type">	The data source to execute the query on.

Returns

Promise<string>

The result of the SQL query in natural language.

---

recallRelevantMemories()

recallRelevantMemories(query?): Promise<void>

Recalls relevant memories based on the given query. If no query is provided, it uses the last user's message text as the query.

Parameters

Parameter	Type	Description
query?	string	The query string to search for relevant memories.

Returns

Promise<void>

---

run()

run(msg, save, returnWhy): Promise<WSMessage>

Processes the user message and returns the response.

Parameters

Parameter	Type	Default value	Description
msg	Message	undefined	The message to send.
save	boolean	true	Whether to save the message or not in the chat history (default: true).
returnWhy	boolean	true	Whether to return the 'why' field in the response (default: true).

Returns

Promise<WSMessage>

The response message.

---

send()

send(msg): void

Sends a message through the websocket connection.

If the websocket connection is not open, the message is queued.

If the message is of type 'chat', it is also stored in the chat history.

Parameters

Parameter	Type	Description
msg	WSMessage	The message to send.

Returns

void