caching original question as answer, limiting acceptable answer score

Author: wjohnsto

services/video-search/src/api/search.ts

@@ -41,47 +41,57 @@ export default function initialize({
         >);
     }
 
-    async function searchVideos(question: string) {
-        log.debug(`Original question: ${question}`, {
-            location: `${prefix}.search.search`,
+    async function checkAnswerCache(question: string) {
+        const haveAnswers = await answerVectorStore.checkIndexExists();
+
+        if (!(haveAnswers && config.searches.answerCache)) {
+            return;
+        }
+
+        log.debug(`Searching for closest answer to question: ${question}`, {
+            location: `${prefix}.search.getAnswer`,
+            question,
         });
-        const semanticQuestion = await prompt.getSemanticQuestion(question);
 
-        const haveAnswers = await answerVectorStore.checkIndexExists();
+        const [result] = await answerVectorStore.similaritySearchWithScore(
+            question,
+            1,
+        );
 
-        if (haveAnswers && config.searches.answerCache) {
-            log.debug(
-                `Searching for closest answer to question: ${semanticQuestion}`,
-                {
-                    location: `${prefix}.search.getAnswer`,
-                    question,
-                },
-            );
+        if (Array.isArray(result) && result.length > 0) {
+            log.debug(`Found closest answer with score: ${String(result[1])}`, {
+                location: `${prefix}.search.getAnswer`,
+                score: result[1],
+            });
 
-            const [result] = await answerVectorStore.similaritySearchWithScore(
-                semanticQuestion,
-                1,
-            );
+            if (result[1] < 0.2) {
+                log.debug(`Found answer to question ${question}`, {
+                    location: `${prefix}.search.getAnswer`,
+                });
 
-            if (Array.isArray(result) && result.length > 0) {
-                log.debug(
-                    `Found closest answer with score: ${String(result[1])}`,
-                    {
-                        location: `${prefix}.search.getAnswer`,
-                        score: result[1],
-                    },
-                );
-
-                if (Array.isArray(result) && result.length > 0) {
-                    log.debug(`Found answer to question ${semanticQuestion}`, {
-                        location: `${prefix}.search.getAnswer`,
-                    });
-
-                    return result[0].metadata;
-                }
+                return result[0].metadata;
             }
+
+            log.debug(`Score too low for question ${question}`, {
+                location: `${prefix}.search.getAnswer`,
+                score: result[1],
+            });
+        }
+    }
+
+    async function searchVideos(question: string) {
+        log.debug(`Original question: ${question}`, {
+            location: `${prefix}.search.search`,
+        });
+
+        const existingAnswer = await checkAnswerCache(question);
+
+        if (typeof existingAnswer !== 'undefined') {
+            return existingAnswer;
         }
 
+        const semanticQuestion = await prompt.getSemanticQuestion(question);
+
         log.debug(`Semantic question: ${semanticQuestion}`, {
             location: `${prefix}.search.search`,
         });
@@ -102,11 +112,12 @@ export default function initialize({
             location: `${prefix}.search.search`,
         });
 
-        const answerDocument = await prompt.answerQuestion(
-            semanticQuestion,
-            videos,
-        );
-        await answerVectorStore.addDocuments([answerDocument]);
+        // TODO: modify the prompt to ask both questions
+        const answerDocument = await prompt.answerQuestion(question, videos);
+
+        if (config.searches.answerCache) {
+            await answerVectorStore.addDocuments([answerDocument]);
+        }
 
         return answerDocument.metadata;
     }

services/video-search/src/api/store.ts

@@ -18,23 +18,31 @@ export default function initialize({
     prefix,
     embeddings,
 }: Pick<ApiConfig, 'prefix' | 'embeddings'>): Store {
+    /**
+     * This vector store will match a short question (e.g. "Tell me about streams") with
+     * a bank of longer transcript summaries + questions.
+     */
     const vectorStore = new RedisVectorStore(embeddings, {
         redisClient: client,
         indexName: `${prefix}-${config.redis.VIDEO_INDEX_NAME}`,
         keyPrefix: `${prefix}-${config.redis.VIDEO_PREFIX}`,
         indexOptions: {
             ALGORITHM: VectorAlgorithms.HNSW,
-            DISTANCE_METRIC: 'COSINE',
+            DISTANCE_METRIC: 'IP',
         },
     });
 
+    /**
+     * This vector store will match a short question (e.g. "Tell me about streams") with previously
+     * answered questions (e.g. "What is a stream?")
+     */
     const answerVectorStore = new RedisVectorStore(embeddings, {
         redisClient: client,
         indexName: `${prefix}-${config.redis.ANSWER_INDEX_NAME}`,
         keyPrefix: `${prefix}-${config.redis.ANSWER_PREFIX}`,
         indexOptions: {
-            ALGORITHM: VectorAlgorithms.HNSW,
-            DISTANCE_METRIC: 'COSINE',
+            ALGORITHM: VectorAlgorithms.FLAT,
+            DISTANCE_METRIC: 'L2',
         },
     });
 

services/video-search/src/config.ts

@@ -52,7 +52,7 @@ export default {
     },
     searches: {
         KNN: 3,
-        answerCache: false,
+        answerCache: true,
     },
     log: {
         LEVEL: LOG_LEVEL ?? 'info',