Anvia

Add context ​

Wrap a prepared VectorStore and its embedding model with createVectorContext() and add it to the agent's context array.

1. Prepare the index before requests ​

Load, chunk, embed, and store documents in an ingestion job:

ts
import { embedDocuments } from '@anvia/core/embeddings';
import { InMemoryVectorStore } from '@anvia/core/vector-store';
const { documents: embedded } = await embedDocuments({
    model: embeddingModel,
    documents: articles,
    id: (article) => article.slug,
    content: (article) => `${article.title}\n\n${article.body}`,
    metadata: (article) => ({
        product: article.product,
        published: article.published,
    })
});
const docsStore = InMemoryVectorStore.fromDocuments({ documents: embedded });

The in-memory store is useful for tests and small process-local corpora. Use a persistent vector-store adapter when the collection must survive restarts or scale independently.

Do not rebuild or re-embed the corpus for every message.

2. Create the retrieval entry ​

ts
import { Agent, createVectorContext } from '@anvia/core';
const docsContext = createVectorContext({
    store: docsStore,
    model: embeddingModel,
    topK: 4,
    minScore: 0.74
});
const agent = new Agent({
    id: 'docs-support',
    model,
    instructions: [
        'Answer from retrieved documentation.',
        'Say when the documentation does not contain the answer.',
    ].join('\n'),
    context: [docsContext],
});

topK is required and must be a positive safe integer. minScore is optional and must be finite. Tune it against real results from that store and embedding model.

3. Run the agent normally ​

ts
const result = await agent.generate({
    prompt: 'How long does a password reset link remain valid?'
})

if (result.type === 'response') {
  console.log(result.output)
}

The route does not make a separate retrieval call. Anvia searches the context index before each model turn with non-empty retrieval text.

4. Start with a small context budget ​

Begin with three to five focused chunks. Increase topK only when answers consistently need evidence from more passages.

Raise minScore when unrelated documents appear. Revisit chunking when the right facts exist but are split across poor boundaries or bundled with unrelated content.

Next, control the exact model document with formatting.

Built for Anvia.