storme/nexusAI
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

refinement fixes

Browse Source
This commit is contained in:
Storme-bit
2026-08-17 00:37:17 -07:00
parent 8f8177615c
commit 7757de76b7
5 changed files with 324 additions and 34 deletions
Download Patch File Download Diff File Expand all files Collapse all files
packages/orchestration-service/src/chat/index.js
+20 -13
View File
@@ -97,14 +97,14 @@ async function autoNameSession(externalId, userMessage, aiResponse) {
}
async function getSemanticEpisodes(
userMessage,
vector,
sessionId,
recentIds,
projectSessionIds = null,
{ semanticLimit, scoreThreshold } = {},
) {
if (!vector) return [];
try {
const vector = await embedding.embed(userMessage);
const results = await qdrant.searchEpisodes(vector, {
limit: semanticLimit,
scoreThreshold: scoreThreshold,
@@ -127,9 +127,9 @@ async function getSemanticEpisodes(
}
}
async function getRelevantEntities(userMessage, projectId = null) {
async function getRelevantEntities(vector, projectId = null) {
if (!vector) return [];
try {
const vector = await embedding.embed(userMessage);
const results = await qdrant.searchEntities(vector, { projectId });
logger.info(
'[orchestration] Entity search results:',
@@ -233,7 +233,7 @@ function selectWithinBudget(scoredPool, contextBudget, minRecentEpisodes, recent
}
async function getFusedEpisodes(userMessage, session, recentIds, projectSessionIds, settings) {
async function getFusedEpisodes(userMessage, session, recentIds, projectSessionIds, vector, settings) {
const { semanticLimit, scoreThreshold, semanticWeight, keywordWeight } = settings;
const ftsSessionIds = projectSessionIds ?? [session.id];
@@ -243,7 +243,7 @@ async function getFusedEpisodes(userMessage, session, recentIds, projectSessionI
: Promise.resolve([]);
const [semanticEps, rawKeywordEps] = await Promise.all([
getSemanticEpisodes(userMessage, session.id, recentIds, projectSessionIds, { semanticLimit, scoreThreshold }),
getSemanticEpisodes(vector, session.id, recentIds, projectSessionIds, { semanticLimit, scoreThreshold }),
ftsPromise,
]);
@@ -283,10 +283,19 @@ async function assembleContext(externalId, userMessage) {
const isFirstMessage = recentEpisodes.length === 0;
const recentIds = new Set(recentEpisodes.map(e => e.id));
// 4. Fused retrieval + entity search in parallel (both are independent)
// 4. Embed the query once — the vector is shared by semantic episode search
// and entity search, so embedding it twice was a wasted round-trip + Ollama call.
let queryVector = null;
try {
queryVector = await embedding.embed(userMessage);
} catch (err) {
logger.warn('[orchestration] Query embedding failed; semantic + entity search disabled this turn:', err.message);
}
// 4b. Fused retrieval + entity search in parallel (both are independent)
const [fusedWithScores, entityResults] = await Promise.all([
getFusedEpisodes(userMessage, session, recentIds, projectSessionIds, { semanticLimit, scoreThreshold, semanticWeight, keywordWeight }),
getRelevantEntities(userMessage, session.project_id ?? null),
getFusedEpisodes(userMessage, session, recentIds, projectSessionIds, queryVector, { semanticLimit, scoreThreshold, semanticWeight, keywordWeight }),
getRelevantEntities(queryVector, session.project_id ?? null),
]);
// 5. Entity-linked episode IDs for scoring bonus
@@ -342,8 +351,7 @@ async function chat(externalId, userMessage, options = {}) {
logger.error('[orchestration] Failed to save episode:', err.message);
}
const allEpisodes = await memory.getRecentEpisodes(session.id, 9999);
triggerSummary(session, allEpisodes);
triggerSummary(session);
if (isFirstMessage && !session.name) {
autoNameSession(externalId, userMessage, result.text).catch(() => {});
@@ -393,8 +401,7 @@ async function chatStream(externalId, userMessage, onChunk, options = {}) {
if (fullText.trim()) {
await memory.createEpisode(session.id, userMessage, fullText, tokenCount, session.project_id ?? null);
const allEpisodes = await memory.getRecentEpisodes(session.id, 9999);
triggerSummary(session, allEpisodes);
triggerSummary(session);
} else {
logger.warn('[orchestration] Stream finished with no assistant text; episode not saved');
}
packages/orchestration-service/src/services/summarization.js
+21 -21
View File
@@ -73,9 +73,12 @@ async function generateSummary(episodes, existingSummary = null) {
return content;
}
async function maybeSummarize(session, allEpisodes) {
// 1. Sum total tokens for this session
const totalTokens = allEpisodes.reduce((sum, ep) => sum + (ep.token_count || 0), 0);
async function maybeSummarize(session) {
// 1. Cheap aggregate — is this session even over the token threshold?
// Avoids pulling every episode row on messages that won't summarize.
const statsRes = await fetch(`${MEMORY_URL}/sessions/${session.id}/episode-stats`);
if (!statsRes.ok) return;
const { totalTokens } = await statsRes.json();
if (totalTokens < THRESHOLD_TOKENS) return; // under threshold — nothing to do
// 2. Fetch existing summaries for session
@@ -84,26 +87,23 @@ async function maybeSummarize(session, allEpisodes) {
const summaries = await summariesRes.json();
const latest = summaries.at(-1) ?? null;
const lastCoveredId = latest
? parseInt(latest.episode_range?.split('-').at(-1)) || 0
const lastCoveredId = latest
? parseInt(latest.episode_range?.split('-').at(-1)) || 0
: 0;
// 3. Guard — don't re-summarize until MIN_EPISODES_SINCE new episodes have accumulated
if (latest) {
const newEpisodes = allEpisodes.filter(ep => ep.id > lastCoveredId);
if (newEpisodes.length < MIN_EPISODES_SINCE) return;
}
// 4. Determine episodes to summarize
const episodesToSummarize = latest
? allEpisodes.filter(ep => ep.id > lastCoveredId)
: allEpisodes;
// 3. Fetch only the un-summarized tail (full text), not the whole session.
// With no prior summary, lastCoveredId is 0 → this returns all episodes.
const episodesToSummarize = await fetch(`${MEMORY_URL}/sessions/${session.id}/episodes/since/${lastCoveredId}`)
.then(r => r.ok ? r.json() : []);
// 4. Guard — don't re-summarize until MIN_EPISODES_SINCE new episodes have accumulated
if (latest && episodesToSummarize.length < MIN_EPISODES_SINCE) return;
if (episodesToSummarize.length === 0) return;
// 5. Determine episode range from the episodes actually being summarized
const summarizedIds = episodesToSummarize.map(ep => ep.id).sort((a,b) => a - b);
const summarizedIds = episodesToSummarize.map(ep => ep.id).sort((a, b) => a - b);
const episodeRange = `${summarizedIds.at(0)}-${summarizedIds.at(-1)}`;
const totalEpisodeTokens = allEpisodes.reduce((sum, ep) => sum + (ep.token_count || 0), 0);
// add temporarily before the generateSummary call
logger.debug('[summarization] episodes to summarize:', episodesToSummarize.length);
const content = await generateSummary(
@@ -122,7 +122,7 @@ async function maybeSummarize(session, allEpisodes) {
body: JSON.stringify({
sessionId: session.id,
content,
tokenCount: totalEpisodeTokens,
tokenCount: totalTokens,
episodeRange,
}),
});
@@ -133,7 +133,7 @@ async function maybeSummarize(session, allEpisodes) {
headers: { 'Content-Type': 'application/json' },
body: JSON.stringify({
content,
tokenCount: totalEpisodeTokens,
tokenCount: totalTokens,
episodeRange,
}),
});
@@ -141,9 +141,9 @@ async function maybeSummarize(session, allEpisodes) {
}
}
async function triggerSummary(session, allEpisodes) {
async function triggerSummary(session) {
// Intentionally fire-and-forget — caller doesn't await this
maybeSummarize(session, allEpisodes).catch(err =>
maybeSummarize(session).catch(err =>
logger.warn('[summarization] Summary failed (non-critical):', err.message)
);
}