Documentation
¶
Index ¶
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
func FormatDocsAsContext ¶
FormatDocsAsContext is a utility to format a JSON list of documents into a readable context string. This is typically used to inject retrieved knowledge into an LLM prompt.
Parameters:
- docsJSON: The JSON string returned by RetrieverAction.Execute.
Returns:
- A formatted string (e.g., "Document 1...\nDocument 2...").
func NewGenkitRetrieverAction ¶
func NewGenkitRetrieverAction(name string, retriever ai.Retriever, embedder ai.Embedder) core.Action
NewGenkitRetrieverAction creates a protected core.Action from a Genkit ai.Retriever.
Parameters:
- name: The unique name for this action.
- retriever: The Genkit retriever instance.
- embedder: Optional embedder (if required by the retriever).
Returns:
- A core.Action that can be used with `client.Protect()`.
Types ¶
type Document ¶
type Document struct {
// Content is the textual body of the document.
Content string `json:"content"`
// Source indicates the origin of the document (e.g., filename, URL).
Source string `json:"source"`
}
Document represents a single unit of retrieved knowledge.
type DocumentRetriever ¶
type DocumentRetriever interface {
// Retrieve finds semantically similar documents for a given query.
//
// Parameters:
// - ctx: The execution context.
// - query: The search string.
//
// Returns:
// - A slice of Document matches, or an error.
Retrieve(ctx context.Context, query string) ([]Document, error)
}
DocumentRetriever defines the interface for a vector search backend. It abstracts the underlying storage mechanism (Pinecone, Weaviate, Local).
type GenkitRetriever ¶
type GenkitRetriever struct {
// contains filtered or unexported fields
}
GenkitRetriever adapts the Genkit ai.Retriever interface to the Manglekit DocumentRetriever interface. It acts as a bridge, allowing Manglekit to use any retriever plugin compatible with Genkit (e.g., Pinecone, Chroma).
func NewGenkitRetriever ¶
func NewGenkitRetriever(retriever ai.Retriever, embedder ai.Embedder) *GenkitRetriever
NewGenkitRetriever initializes a new adapter for a Genkit retriever.
Parameters:
- retriever: The Genkit retriever instance.
- embedder: An optional embedder. If the retriever handles embeddings internally, this can be nil.
Returns:
- A pointer to the initialized GenkitRetriever.
func (*GenkitRetriever) Retrieve ¶
Retrieve executes a search query against the underlying Genkit retriever. It converts the results into Manglekit's standard Document format.
Parameters:
- ctx: The execution context.
- query: The search query string.
Returns:
- A slice of Document structs found by the retriever.
type RetrieverAction ¶
type RetrieverAction struct {
// contains filtered or unexported fields
}
RetrieverAction wraps a DocumentRetriever into a core.Action. This allows retrieval operations to be governed by policies and traced.
func NewRetrieverAction ¶
func NewRetrieverAction(name string, retriever DocumentRetriever) *RetrieverAction
NewRetrieverAction creates a new RetrieverAction.
Parameters:
- name: The unique name for this action.
- retriever: The retrieval backend implementation.
Returns:
- A pointer to the initialized RetrieverAction.
func (*RetrieverAction) Execute ¶
Execute performs the vector search. It expects a string payload (query) and returns a JSON string payload (list of documents).
Parameters:
- ctx: The execution context.
- input: The input Envelope (Payload must be string).
Returns:
- A result Envelope containing a JSON string of retrieved documents.
func (*RetrieverAction) Metadata ¶
func (r *RetrieverAction) Metadata() core.ActionMetadata
Metadata returns the action's metadata.
Source Files