@toolcog/core
Toolcog core
Interfaces
Embedder()
A function that returns an EmbeddingVector for each provided string.
If the embed argument is a string, then an embedding vector is returned.
If the embed argument is an array of strings, then an array of embedding
vectors is returned.
Embedder<
T>(embed,options?):Promise<Embedded<T>>
A function that returns an EmbeddingVector for each provided string.
If the embed argument is a string, then an embedding vector is returned.
If the embed argument is an array of strings, then an array of embedding
vectors is returned.
Type Parameters
• T extends string | readonly string[]
Parameters
• embed: T
• options?: EmbedderOptions
Returns
Promise<Embedded<T>>
Defined in
EmbedderConfig
Options for configuring an Embedder function.
Note that embedder plugins may augment this type with additional options.
Extended by
Properties
model?
optionalmodel:string&object
The default model the embedder should use.
Defined in
EmbedderOptions
Options for controlling an Embedder call.
Note that embedder plugins may augment this type with additional options.
Extended by
Properties
model?
optionalmodel:string&object
The model the embedder should use to generate embedding vectors.
Defined in
signal?
optionalsignal:AbortSignal
An abort signal that can be used to cancel the embedder call.
Defined in
EmbeddingModelNames
Each key of this type represents the name of a known embedding model. Embedder plugins augment this type to add supported model names.
Use the EmbeddingModel type for strings that should represent
embedding model names. The EmbeddingModel type extracts the keys of
this type. The indirection through this type is necessary because type
aliases cannot be augmented.
Embeddings<V>
A set of embeddings keyed by the embedded text.
Type Parameters
• V = EmbeddingVector
Indexable
[text: string]: Embedding<V>
GenerativeConfig
Options for configuring a Generator function.
Note that generator plugins may augment this type with additional options.
Extends
Properties
defaults?
optionaldefaults:Record<string,unknown>
Defined in
model?
optionalmodel:string&object
The default model the generator should use.
Inherited from
Defined in
stream?
optionalstream:boolean
Whether or not the generator should stream responses by default.
Inherited from
Defined in
system?
optionalsystem:string
The default system prompt the generator should use when generating the response.
Inherited from
Defined in
tools?
optionaltools:null| readonlyToolSource[]
The default set of tools the generator should use.
Overrides
Defined in
GenerativeFunction()<F>
Type Parameters
• F extends (…args) => unknown = (…args) => unknown
GenerativeFunction(…
args):GenerativeReturnType<F>
Parameters
• …args: GenerativeParameters<F>
Returns
Defined in
Properties
description
readonlydescription:undefined|string
Defined in
id
readonlyid:string
Defined in
instructions
readonlyinstructions:InstructionsSource
Defined in
name
readonlyname:string
Defined in
parameters
readonlyparameters:undefined|Schema
Defined in
returns
readonlyreturns:undefined|Schema
Defined in
tools
readonlytools: readonlyToolSource[]
Defined in
GenerativeModelNames
Each key of this type represents the name of a known generative model. Generator plugins augment this type to add supported model names.
Use the GenerativeModel type for strings that should represent
generative model names. The GenerativeModel type extracts the keys of
this type. The indirection through this type is necessary because type
aliases cannot be augmented.
GenerativeOptions
Options for controlling a Generator call.
Note that generator plugins may augment this type with additional options.
Extends
Properties
id?
optionalid:string
Inherited from
Defined in
instructions?
optionalinstructions:InstructionsSource
Instructions the generator should follow when generating the response.
Overrides
Defined in
model?
optionalmodel:string&object
The model the generator should use to generate the response.
Inherited from
Defined in
parameters?
optionalparameters:Schema
A schema that describes the arguments to the generator call.
Inherited from
Defined in
returns?
optionalreturns:Schema
A schema that describes the value the generator must generate.
Inherited from
Defined in
signal?
optionalsignal:null|AbortSignal
An abort signal that can be used to cancel the generator call.
Inherited from
Defined in
stream?
optionalstream:boolean
Whether or not the generator should stream responses.
Inherited from
Defined in
system?
optionalsystem:string
The system prompt the generator should use when generating the response.
Inherited from
Defined in
tools?
optionaltools:null| readonlyToolSource[]
The tools the generator should use when generating the response.
Overrides
Defined in
Generator()
A function that uses a generative model to generate its return value.
options.model specifies the generative model to use, if defined.
The model will be prompted to follow any instructions provided in
options.instructions. And it will be given access to any tools
provided in options.tools.
If options.function defines a function schema, the model will be prompted
to interpret the args according to its parameter schema, and to generate
a response that conform the its return schema. If no function schema is
provided, args will be used as the raw prompt, and the model’s response
will be returned as a string.
Generator(
args,options?):Promise<unknown>
A function that uses a generative model to generate its return value.
options.model specifies the generative model to use, if defined.
The model will be prompted to follow any instructions provided in
options.instructions. And it will be given access to any tools
provided in options.tools.
If options.function defines a function schema, the model will be prompted
to interpret the args according to its parameter schema, and to generate
a response that conform the its return schema. If no function schema is
provided, args will be used as the raw prompt, and the model’s response
will be returned as a string.
Parameters
• args: unknown
• options?: GeneratorOptions
Returns
Promise<unknown>
Defined in
GeneratorConfig
Options for configuring a Generator function.
Note that generator plugins may augment this type with additional options.
Extended by
Properties
model?
optionalmodel:string&object
The default model the generator should use.
Defined in
stream?
optionalstream:boolean
Whether or not the generator should stream responses by default.
Defined in
system?
optionalsystem:string
The default system prompt the generator should use when generating the response.
Defined in
tools?
optionaltools:null| readonlyToolSource[]
The default set of tools the generator should use.
Defined in
GeneratorOptions
Options for controlling a Generator call.
Note that generator plugins may augment this type with additional options.
Extended by
Properties
id?
optionalid:string
Defined in
instructions?
optionalinstructions:InstructionsSource
Instructions the generator should follow when generating the response.
Defined in
model?
optionalmodel:string&object
The model the generator should use to generate the response.
Defined in
parameters?
optionalparameters:Schema
A schema that describes the arguments to the generator call.
Defined in
returns?
optionalreturns:Schema
A schema that describes the value the generator must generate.
Defined in
signal?
optionalsignal:null|AbortSignal
An abort signal that can be used to cancel the generator call.
Defined in
stream?
optionalstream:boolean
Whether or not the generator should stream responses.
Defined in
system?
optionalsystem:string
The system prompt the generator should use when generating the response.
Defined in
tools?
optionaltools:null| readonlyToolSource[]
The tools the generator should use when generating the response.
Defined in
Idiom()<T>
A collection of embeddings associated with a value.
Type Parameters
• T
Idiom():
Embeddings<Float32Array>
A collection of embeddings associated with a value.
Returns
Embeddings<Float32Array>
Defined in
Properties
id
readonlyid:string
Defined in
value
readonlyvalue:T
Defined in
IdiomResolver()
IdiomResolver(
id,value):undefined|Embeddings<Float32Array>
Parameters
• id: string
• value: unknown
Returns
undefined | Embeddings<Float32Array>
Defined in
Index()<T>
A similarity search index.
Type Parameters
• T extends readonly unknown[]
Index(
query?,options?):Promise<T[number][]>
A similarity search index.
Parameters
• query?: unknown
• options?: IndexOptions
Returns
Promise<T[number][]>
Defined in
Properties
embedder
readonlyembedder:undefined|Embedder
Defined in
id
readonlyid:undefined|string
Defined in
idioms
readonlyidioms:Idioms<T>
Defined in
IndexConfig
Options for configuring an Index function.
Extends
Extended by
Properties
limit?
optionallimit:number
Defined in
model?
optionalmodel:string&object
The default model the embedder should use.
Inherited from
Defined in
Indexer()
A function that returns a similarity Index for a set of idioms.
Indexer<
T>(idioms,options?):Promise<Index<T>>
A function that returns a similarity Index for a set of idioms.
Type Parameters
• T extends readonly unknown[]
Parameters
• idioms: Idioms<T>
• options?: IndexerOptions
Returns
Promise<Index<T>>
Defined in
IndexerConfig
Options for configuring an Indexer function.
Extends
Properties
distance?
optionaldistance:EmbeddingDistance
Defined in
embedder?
optionalembedder:Embedder
Defined in
limit?
optionallimit:number
Inherited from
Defined in
model?
optionalmodel:string&object
The default model the embedder should use.
Inherited from
Defined in
IndexerOptions
Options for controlling an Indexer call.
Extends
Properties
distance?
optionaldistance:EmbeddingDistance
Defined in
embedder?
optionalembedder:Embedder
Defined in
id?
optionalid:string
Defined in
limit?
optionallimit:number
Inherited from
Defined in
model?
optionalmodel:string&object
The model the embedder should use to generate embedding vectors.
Inherited from
Defined in
signal?
optionalsignal:AbortSignal
An abort signal that can be used to cancel the embedder call.
Inherited from
Defined in
IndexOptions
Options for controlling an Index call.
Extends
Extended by
Properties
limit?
optionallimit:number
Defined in
model?
optionalmodel:string&object
The model the embedder should use to generate embedding vectors.
Inherited from
Defined in
signal?
optionalsignal:AbortSignal
An abort signal that can be used to cancel the embedder call.
Inherited from
Defined in
Tool()<F>
Type Parameters
• F extends (…args) => unknown = (…args) => unknown
Tool(…
args):ReturnType<F>
Parameters
• …args: Parameters<F>
Returns
ReturnType<F>
Defined in
Properties
description
readonlydescription:undefined|string
Defined in
id
readonlyid:string
Defined in
name
readonlyname:string
Defined in
parameters
readonlyparameters:undefined|Schema
Defined in
returns
readonlyreturns:undefined|Schema
Defined in
Type Aliases
Embedded<T>
Embedded<
T>:Textendsstring?EmbeddingVector:Textends readonlystring[] ?EmbeddingVector[] :Textendsstring| readonlystring[] ?EmbeddingVector|EmbeddingVector[] :never
The return type of an Embedder call. If the argument is a string, an embedding vector is returned. If the argument is an array of strings, an array of embedding vectors is returned.
Type Parameters
• T extends string | readonly string[] = string | readonly string[]
Defined in
Embedding<V>
Embedding<
V>:{ [Model in EmbeddingModel]?: V }
A set of embedding vectors keyed by the model that generated each vector.
Type Parameters
• V = EmbeddingVector
Defined in
EmbeddingDistance()
EmbeddingDistance: (
a,b) =>number
A distance metric in an embedding vector space.
Parameters
• a: EmbeddingVector
• b: EmbeddingVector
Returns
number
Defined in
EmbeddingModel
EmbeddingModel: keyof
EmbeddingModelNames|string&object
The identifying name of an embedding model.
Defined in
EmbeddingVector
EmbeddingVector:
Float32Array
The type of an embedding vector. Embedding vectors are stored as float arrays for memory efficiency.
Defined in
GenerativeModel
GenerativeModel: keyof
GenerativeModelNames|string&object
The identifying name of a generative model.
Defined in
GenerativeParameters<F>
GenerativeParameters<
F>:IsVariadic<Parameters<F>> extendstrue?Parameters<F> : [...Parameters<F>,GenerativeOptions]
Type Parameters
• F extends (…args) => unknown
Defined in
GenerativeReturnType<F>
GenerativeReturnType<
F>:Promise<Awaited<ReturnType<F>>>
Type Parameters
• F extends (…args) => unknown
Defined in
Idioms<T>
Idioms<
T>:Textends readonly [] ? readonly [] :Textends readonly [infer Elem] ? readonly [Idiom<Elem>] :Textends readonly [infer Head,...(infer Tail)] ? readonly [Idiom<Head>,...Idioms<Tail>] :Textends readonly [...(infer Body), infer Foot] ? readonly [...Idioms<Body>,Idiom<Foot>] :Textends readonly infer Elem[] ? readonlyIdiom<Elem>[] :never
Type Parameters
• T extends readonly unknown[]
Defined in
InstructionsSource
InstructionsSource: (
args) =>Promise<string|undefined> |string|undefined|Promise<string|undefined> |string|undefined
Defined in
Tools<F>
Tools<
F>:Fextends readonly [] ? readonly [] :Fextends readonly [infer Elem] ?Elemextends (…args) =>unknown? readonly [Tool<Elem>] :never:Fextends readonly [infer Head,...(infer Tail)] ? readonly [Headextends (…args) =>unknown?Tool<Head> :never,...(Tail extends readonly (...)[] ? Tools<Tail> : never)] :Fextends readonly [...(infer Body), infer Foot] ? readonly [...(Body extends readonly (...)[] ? Tools<Body> : never),Footextends (…args) =>unknown?Tool<Foot> :never] :Fextends readonly infer Elem[] ?Elemextends (…args) =>unknown? readonlyTool<Elem>[] :never:Fextends (…args) =>unknown?Tool<F> :never
Type Parameters
• F extends readonly (…args) => unknown[]
Defined in
ToolSource
ToolSource: (
args) =>Promise<readonlyTool[] |Tool|undefined> | readonlyTool[] |Tool|undefined|Promise<readonlyTool[] |Tool|undefined> | readonlyTool[] |Tool|undefined
Defined in
Functions
decodeEmbedding()
decodeEmbedding(
embedding):Embedding<Float32Array>
Parameters
• embedding: Embedding<Buffer>
Returns
Embedding<Float32Array>
Defined in
decodeEmbeddings()
decodeEmbeddings(
embeddings):Embeddings<Float32Array>
Parameters
• embeddings: Embeddings<Buffer>
Returns
Embeddings<Float32Array>
Defined in
decodeEmbeddingVector()
decodeEmbeddingVector(
vector):Float32Array
Parameters
• vector: Buffer
Returns
Float32Array
Defined in
defineFunction()
defineFunction<
F>(config?):GenerativeFunction<F>
Type Parameters
• F extends (…args) => unknown
Parameters
• config?: GenerativeConfig
Returns
Defined in
defineIdiom()
defineIdiom<
T>(value):Idiom<T>
Type Parameters
• T
Parameters
• value: T
Returns
Idiom<T>
Defined in
defineIdioms()
defineIdioms<
T>(values):Idioms<T>
Type Parameters
• T extends readonly unknown[]
Parameters
• values: readonly [T]
Returns
Idioms<T>
Defined in
defineIndex()
defineIndex<
T>(values,config?):Index<T>
Type Parameters
• T extends readonly unknown[]
Parameters
• values: readonly [T] | Idioms<T>
• config?: IndexConfig
Returns
Index<T>
Defined in
defineTool()
defineTool<
F>(func):Tool<F>
Generates an LLM tool for a TypeScript function.
Type Parameters
• F extends (…args) => unknown
Parameters
• func: F
Returns
Tool<F>
Defined in
defineTools()
defineTools<
F>(funcs):Tools<F>
Type Parameters
• F extends readonly (…args) => unknown[]
Parameters
• funcs: readonly [F]
Returns
Tools<F>
Defined in
encodeEmbedding()
encodeEmbedding(
embedding):Embedding<Buffer>
Parameters
• embedding: Embedding<Float32Array>
Returns
Embedding<Buffer>
Defined in
encodeEmbeddings()
encodeEmbeddings(
embeddings):Embeddings<Buffer>
Parameters
• embeddings: Embeddings<Float32Array>
Returns
Embeddings<Buffer>
Defined in
encodeEmbeddingVector()
encodeEmbeddingVector(
vector):Buffer
Parameters
• vector: Float32Array
Returns
Buffer
Defined in
prompt()
prompt(instructions, args, options)
prompt<
T>(instructions,args?,options?):Promise<T>
Type Parameters
• T = string
Parameters
• instructions: undefined | string
• args?: Record<string, unknown>
• options?: GenerativeOptions
Returns
Promise<T>
Defined in
prompt(args, options)
prompt<
T>(args?,options?):Promise<T>
Type Parameters
• T = string
Parameters
• args?: Record<string, unknown>
• options?: GenerativeOptions
Returns
Promise<T>
Defined in
resolveInstructions()
resolveInstructions(
instructions,args):Promise<undefined|string>
Parameters
• instructions: InstructionsSource
• args: unknown
Returns
Promise<undefined | string>
Defined in
resolveTool()
resolveTool(
tool,args):Promise<undefined|Tool<(…args) =>unknown> | readonlyTool<(…args) =>unknown>[]>
Parameters
• tool: ToolSource
• args: unknown
Returns
Promise<undefined | Tool<(…args) => unknown> | readonly Tool<(…args) => unknown>[]>
Defined in
resolveTools()
resolveTools(
tools,args):Promise<undefined|Tool<(…args) =>unknown>[]>
Parameters
• tools: undefined | null | readonly ToolSource[]
• args: unknown
Returns
Promise<undefined | Tool<(…args) => unknown>[]>