tri.ai.text.chunks.process.TextDocEmbeddings.kt Maven / Gradle / Ivy

Go to download
Show more of this group Show more artifacts with this name
Show all versions of promptkt Show documentation
LLM and prompt engineering.
The newest version!
package tri.ai.text.chunks.process

/*-
 * #%L
 * tri.promptfx:promptkt
 * %%
 * Copyright (C) 2023 - 2024 Johns Hopkins University Applied Physics Laboratory
 * %%
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at
 *
 *      http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 * #L%
 */

import kotlinx.coroutines.runBlocking
import tri.ai.embedding.EmbeddingPrecision
import tri.ai.embedding.EmbeddingService
import tri.ai.text.chunks.TextChunk
import tri.ai.text.chunks.TextDoc
import tri.util.info
import java.net.URI

/**
 * Utilities for adding embedding information to [TextDoc]s.
 * Enforces a maximum number of embeddings to calculate in one query.
 */
object TextDocEmbeddings {

    /** Max number of embeddings to calculate in one query. */
    const val MAX_EMBEDDING_BATCH_SIZE = 20

    /** Calculate an embedding for a single chunk. */
    suspend fun EmbeddingService.calculate(doc: TextDoc, chunk: TextChunk) =
        calculateEmbedding(chunk.text(doc.all))

    /** Calculate embeddings for a single document. */
    suspend fun EmbeddingService.calculate(doc: TextDoc): List> =
        doc.chunks.map { it.text(doc.all) }.chunked(MAX_EMBEDDING_BATCH_SIZE)
            .flatMap { calculateEmbedding(it) }

    /** Add embedding info for all chunks in a document. */
    fun EmbeddingService.addEmbeddingInfo(doc: TextDoc) {
        val embeddings = runBlocking { calculate(doc) }
        doc.chunks.forEachIndexed { i, chunk ->
            chunk.putEmbeddingInfo(modelId, embeddings[i], precision)
        }
    }

    /** Save embedding info with a chunk. */
    fun TextChunk.putEmbeddingInfo(modelId: String, embedding: List, precision: EmbeddingPrecision) {
        attributes.putIfAbsent("embeddings", mutableMapOf>())
        (attributes["embeddings"] as EmbeddingInfo)[modelId] = embedding.map { precision.op(it) }
    }

    /** Get embedding info object. */
    fun TextChunk.getEmbeddingInfo(): EmbeddingInfo? =
        attributes["embeddings"] as? EmbeddingInfo

    /** Get embedding info for a specific model. */
    fun TextChunk.getEmbeddingInfo(modelId: String): List? =
        (attributes["embeddings"] as? EmbeddingInfo)?.get(modelId)

    /** Chunks a text into sections and calculates the embedding for each section. */
    suspend fun EmbeddingService.chunkedEmbedding(path: URI, text: String, maxChunkSize: Int): TextDoc {
        info("Calculating embeddings for $path...")
        val doc = TextDoc(path.toString(), text).apply {
            metadata.path = path
        }
        doc.chunks.addAll(chunkTextBySections(text, maxChunkSize))
        doc.calculateMissingEmbeddings(this)
        return doc
    }

    /** Calculates embedding info for all chunks in a document where it is missing. */
    suspend fun TextDoc.calculateMissingEmbeddings(embeddingService: EmbeddingService) {
        val id = embeddingService.modelId
        val chunksToCalculate = chunks.filter { it.getEmbeddingInfo(id) == null }
        if (chunksToCalculate.isNotEmpty()) {
            chunksToCalculate.chunked(MAX_EMBEDDING_BATCH_SIZE).forEach { batch ->
                embeddingService.calculateEmbedding(batch.map { it.text(all) }).forEachIndexed { i, embedding ->
                    batch[i].putEmbeddingInfo(id, embedding, embeddingService.precision)
                }
            }
        }
    }

}

typealias EmbeddingInfo = MutableMap>