tessl/npm-llamaindex
Data framework for your LLM application
—
Pending
Quality
Pending
Does it follow best practices?
Impact
Pending
No eval scenarios have been run
Securityby
Pending
The risk profile of this skill
response-synthesis.mddocs/
Response Synthesis
Response generation and synthesis strategies for combining retrieved information into coherent answers in LlamaIndex.TS.
Import
import { ResponseSynthesizer } from "llamaindex";Overview
Response synthesis combines retrieved information from multiple sources into coherent, comprehensive answers using various strategies optimized for different use cases.
Base Synthesizer Interface
interface BaseSynthesizer {
synthesize(query: string, nodes: BaseNode[]): Promise<EngineResponse>;
}Response Synthesizer
class ResponseSynthesizer implements BaseSynthesizer {
constructor(options?: {
responseMode?: ResponseMode;
serviceContext?: ServiceContext;
});
synthesize(query: string, nodes: BaseNode[]): Promise<EngineResponse>;
responseMode: ResponseMode;
}
type ResponseMode = "refine" | "compact" | "tree_summarize" | "simple_summarize";Response Modes
Tree Summarize
Best for comprehensive answers from multiple sources.
const treeSynthesizer = new ResponseSynthesizer({
responseMode: "tree_summarize",
});Refine Mode
Iteratively refines answers with each chunk.
const refineSynthesizer = new ResponseSynthesizer({
responseMode: "refine",
});Compact Mode
Combines chunks to maximize context usage.
const compactSynthesizer = new ResponseSynthesizer({
responseMode: "compact",
});Usage with Query Engines
import { RetrieverQueryEngine, ResponseSynthesizer } from "llamaindex";
const queryEngine = new RetrieverQueryEngine({
retriever: index.asRetriever(),
responseSynthesizer: new ResponseSynthesizer({
responseMode: "tree_summarize",
}),
});Factory Function
function createResponseSynthesizer(mode: ResponseMode): BaseSynthesizer;Best Practices
// Choose mode based on use case
const synthesizer = createResponseSynthesizer(
documents.length > 10 ? "tree_summarize" : "refine"
);