Added GPT=produced KSA when not found.
This commit is contained in:
parent
08c064a869
commit
3cf752e8e3
@ -15,6 +15,7 @@ import { v4 as uuidv4 } from 'uuid';
|
||||
import pkg from 'pdfjs-dist';
|
||||
import mysql from 'mysql2/promise'; // <-- MySQL instead of SQLite
|
||||
import OpenAI from 'openai';
|
||||
import Fuse from 'fuse.js';
|
||||
|
||||
// Basic file init
|
||||
const __filename = fileURLToPath(import.meta.url);
|
||||
@ -2052,6 +2053,247 @@ app.delete('/api/premium/milestone-impacts/:impactId', authenticatePremiumUser,
|
||||
}
|
||||
});
|
||||
|
||||
let onetKsaData = []; // entire array from ksa_data.json
|
||||
let allKsaNames = []; // an array of unique KSA names (for fuzzy matching)
|
||||
|
||||
(async function loadKsaJson() {
|
||||
try {
|
||||
const filePath = path.join(__dirname, '..', 'public', 'ksa_data.json');
|
||||
const raw = await fs.readFile(filePath, 'utf8');
|
||||
onetKsaData = JSON.parse(raw);
|
||||
|
||||
// Build a set of unique KSA names for fuzzy search
|
||||
const nameSet = new Set();
|
||||
for (const row of onetKsaData) {
|
||||
nameSet.add(row.elementName);
|
||||
}
|
||||
allKsaNames = Array.from(nameSet);
|
||||
console.log(`Loaded ksa_data.json with ${onetKsaData.length} rows; ${allKsaNames.length} unique KSA names.`);
|
||||
} catch (err) {
|
||||
console.error('Error loading ksa_data.json:', err);
|
||||
}
|
||||
})();
|
||||
|
||||
// 2) Create fuzzy search index
|
||||
let fuse = null;
|
||||
function initFuzzySearch() {
|
||||
if (!fuse) {
|
||||
fuse = new Fuse(allKsaNames, {
|
||||
includeScore: true,
|
||||
threshold: 0.3, // adjust to your preference
|
||||
});
|
||||
}
|
||||
}
|
||||
|
||||
function fuzzyMatchKsaName(name) {
|
||||
if (!fuse) initFuzzySearch();
|
||||
const results = fuse.search(name);
|
||||
if (!results.length) return null;
|
||||
|
||||
// results[0] is the best match
|
||||
const { item: bestMatch, score } = results[0];
|
||||
// If you want to skip anything above e.g. 0.5 score, do:
|
||||
if (score > 0.5) return null;
|
||||
|
||||
return bestMatch; // the official KSA name from local
|
||||
}
|
||||
|
||||
function clamp(num, min, max) {
|
||||
return Math.max(min, Math.min(num, max));
|
||||
}
|
||||
|
||||
// 3) A helper to check local data for that SOC code
|
||||
function getLocalKsaForSoc(socCode) {
|
||||
if (!onetKsaData.length) return [];
|
||||
return onetKsaData.filter((r) => r.onetSocCode === socCode);
|
||||
}
|
||||
|
||||
// 4) ChatGPT call
|
||||
async function fetchKsaFromOpenAI(socCode, careerTitle) {
|
||||
const openai = new OpenAI({ apiKey: process.env.OPENAI_API_KEY });
|
||||
|
||||
// 1. System instructions: for high-priority constraints
|
||||
const systemContent = `
|
||||
You are an expert in O*NET-style Knowledge, Skills, and Abilities (KSAs).
|
||||
Always produce a thorough KSA list for the career described.
|
||||
Carefully follow instructions about minimum counts per category.
|
||||
No additional commentary or disclaimers.
|
||||
`;
|
||||
|
||||
// 2. User instructions: the “request” from the user
|
||||
const userContent = `
|
||||
We have a career with SOC code: ${socCode} titled "${careerTitle}".
|
||||
We need 3 arrays in JSON: "knowledge", "skills", "abilities".
|
||||
|
||||
**Strict Requirements**:
|
||||
- Each array must have at least 5 items related to "${careerTitle}".
|
||||
- Each item: { "elementName": "...", "importanceValue": (1–5), "levelValue": (0–7) }.
|
||||
- Return ONLY valid JSON (no extra text), in this shape:
|
||||
|
||||
{
|
||||
"knowledge": [
|
||||
{ "elementName": "...", "importanceValue": 3, "levelValue": 5 },
|
||||
...
|
||||
],
|
||||
"skills": [...],
|
||||
"abilities": [...]
|
||||
}
|
||||
|
||||
No extra commentary. Exactly 3 arrays, each with at least 5 items.
|
||||
Make sure to include relevant domain-specific knowledge (e.g. “Programming,” “Computer Systems,” etc.).
|
||||
`;
|
||||
|
||||
// 3. Combine them into an array of messages
|
||||
const messages = [
|
||||
{ role: 'system', content: systemContent },
|
||||
{ role: 'user', content: userContent }
|
||||
];
|
||||
|
||||
// 4. Make the GPT-4 call
|
||||
const completion = await openai.chat.completions.create({
|
||||
model: 'gpt-4',
|
||||
messages: messages,
|
||||
temperature: 0.2,
|
||||
max_tokens: 600
|
||||
});
|
||||
|
||||
// 5. Attempt to parse the JSON
|
||||
const rawText = completion?.choices?.[0]?.message?.content?.trim() || '';
|
||||
let parsed = { knowledge: [], skills: [], abilities: [] };
|
||||
try {
|
||||
parsed = JSON.parse(rawText);
|
||||
} catch (err) {
|
||||
console.error('Error parsing GPT-4 JSON:', err, rawText);
|
||||
}
|
||||
|
||||
return parsed; // e.g. { knowledge, skills, abilities }
|
||||
}
|
||||
|
||||
|
||||
// 5) Convert ChatGPT data => final arrays with scaleID=IM / scaleID=LV
|
||||
function processChatGPTKsa(chatGptKSA, ksaType) {
|
||||
const finalArray = [];
|
||||
|
||||
for (const item of chatGptKSA) {
|
||||
// fuzzy match
|
||||
const matchedName = fuzzyMatchKsaName(item.elementName);
|
||||
if (!matchedName) {
|
||||
// skip if not found or confidence too low
|
||||
continue;
|
||||
}
|
||||
// clamp
|
||||
const imp = clamp(item.importanceValue, 1, 5);
|
||||
const lvl = clamp(item.levelValue, 0, 7);
|
||||
|
||||
// produce 2 records: IM + LV
|
||||
finalArray.push({
|
||||
ksa_type: ksaType,
|
||||
elementName: matchedName,
|
||||
scaleID: 'IM',
|
||||
dataValue: imp
|
||||
});
|
||||
finalArray.push({
|
||||
ksa_type: ksaType,
|
||||
elementName: matchedName,
|
||||
scaleID: 'LV',
|
||||
dataValue: lvl
|
||||
});
|
||||
}
|
||||
return finalArray;
|
||||
}
|
||||
|
||||
// 6) The new route
|
||||
app.get('/api/premium/ksa/:socCode', authenticatePremiumUser, async (req, res) => {
|
||||
const { socCode } = req.params;
|
||||
const { careerTitle = '' } = req.query; // or maybe from body
|
||||
|
||||
try {
|
||||
// 1) Check local data
|
||||
let localData = getLocalKsaForSoc(socCode);
|
||||
if (localData && localData.length > 0) {
|
||||
return res.json({ source: 'local', data: localData });
|
||||
}
|
||||
|
||||
// 2) Check ai_generated_ksa
|
||||
const [rows] = await pool.query(
|
||||
'SELECT * FROM ai_generated_ksa WHERE soc_code = ? LIMIT 1',
|
||||
[socCode]
|
||||
);
|
||||
if (rows && rows.length > 0) {
|
||||
const row = rows[0];
|
||||
const knowledge = JSON.parse(row.knowledge_json || '[]');
|
||||
const skills = JSON.parse(row.skills_json || '[]');
|
||||
const abilities = JSON.parse(row.abilities_json || '[]');
|
||||
|
||||
// Check if they are truly empty
|
||||
const isAllEmpty = !knowledge.length && !skills.length && !abilities.length;
|
||||
if (!isAllEmpty) {
|
||||
// We have real data
|
||||
return res.json({
|
||||
source: 'db',
|
||||
data: { knowledge, skills, abilities }
|
||||
});
|
||||
}
|
||||
console.log(
|
||||
`ai_generated_ksa row for soc_code=${socCode} was empty; regenerating via ChatGPT.`
|
||||
);
|
||||
}
|
||||
|
||||
// 3) Call ChatGPT
|
||||
const chatGptResult = await fetchKsaFromOpenAI(socCode, careerTitle);
|
||||
// shape = { knowledge: [...], skills: [...], abilities: [...] }
|
||||
|
||||
// 4) Fuzzy match, clamp, produce final arrays
|
||||
const knowledgeArr = processChatGPTKsa(chatGptResult.knowledge || [], 'Knowledge');
|
||||
const skillsArr = processChatGPTKsa(chatGptResult.skills || [], 'Skill');
|
||||
const abilitiesArr = processChatGPTKsa(chatGptResult.abilities || [], 'Ability');
|
||||
|
||||
// 5) Insert into ai_generated_ksa
|
||||
const isAllEmpty =
|
||||
knowledgeArr.length === 0 &&
|
||||
skillsArr.length === 0 &&
|
||||
abilitiesArr.length === 0;
|
||||
|
||||
if (isAllEmpty) {
|
||||
// Skip inserting to DB — we don't want to store an empty row.
|
||||
return res.status(500).json({
|
||||
error: 'ChatGPT returned no KSA data. Please try again later.',
|
||||
data: { knowledge: [], skills: [], abilities: [] }
|
||||
});
|
||||
}
|
||||
|
||||
// Otherwise, insert into DB as normal:
|
||||
await pool.query(`
|
||||
INSERT INTO ai_generated_ksa (
|
||||
soc_code,
|
||||
career_title,
|
||||
knowledge_json,
|
||||
skills_json,
|
||||
abilities_json
|
||||
)
|
||||
VALUES (?, ?, ?, ?, ?)
|
||||
`, [
|
||||
socCode,
|
||||
careerTitle,
|
||||
JSON.stringify(knowledgeArr),
|
||||
JSON.stringify(skillsArr),
|
||||
JSON.stringify(abilitiesArr)
|
||||
]);
|
||||
|
||||
return res.json({
|
||||
source: 'chatgpt',
|
||||
data: {
|
||||
knowledge: knowledgeArr,
|
||||
skills: skillsArr,
|
||||
abilities: abilitiesArr
|
||||
}
|
||||
});
|
||||
} catch (err) {
|
||||
console.error('Error retrieving KSA fallback data:', err);
|
||||
return res.status(500).json({ error: err.message || 'Failed to fetch KSA data.' });
|
||||
}
|
||||
});
|
||||
|
||||
/* ------------------------------------------------------------------
|
||||
RESUME OPTIMIZATION ENDPOINT
|
||||
------------------------------------------------------------------ */
|
||||
|
||||
10
package-lock.json
generated
10
package-lock.json
generated
@ -25,6 +25,7 @@
|
||||
"docx": "^9.5.0",
|
||||
"dotenv": "^16.4.7",
|
||||
"file-saver": "^2.0.5",
|
||||
"fuse.js": "^7.1.0",
|
||||
"helmet": "^8.0.0",
|
||||
"jsonwebtoken": "^9.0.2",
|
||||
"jwt-decode": "^4.0.0",
|
||||
@ -9685,6 +9686,15 @@
|
||||
"url": "https://github.com/sponsors/ljharb"
|
||||
}
|
||||
},
|
||||
"node_modules/fuse.js": {
|
||||
"version": "7.1.0",
|
||||
"resolved": "https://registry.npmjs.org/fuse.js/-/fuse.js-7.1.0.tgz",
|
||||
"integrity": "sha512-trLf4SzuuUxfusZADLINj+dE8clK1frKdmqiJNb1Es75fmI5oY6X2mxLVUciLLjxqw/xr72Dhy+lER6dGd02FQ==",
|
||||
"license": "Apache-2.0",
|
||||
"engines": {
|
||||
"node": ">=10"
|
||||
}
|
||||
},
|
||||
"node_modules/gauge": {
|
||||
"version": "3.0.2",
|
||||
"resolved": "https://registry.npmjs.org/gauge/-/gauge-3.0.2.tgz",
|
||||
|
||||
@ -20,6 +20,7 @@
|
||||
"docx": "^9.5.0",
|
||||
"dotenv": "^16.4.7",
|
||||
"file-saver": "^2.0.5",
|
||||
"fuse.js": "^7.1.0",
|
||||
"helmet": "^8.0.0",
|
||||
"jsonwebtoken": "^9.0.2",
|
||||
"jwt-decode": "^4.0.0",
|
||||
|
||||
@ -37,7 +37,7 @@ function ensureHttp(urlString) {
|
||||
if (/^https?:\/\//i.test(urlString)) {
|
||||
return urlString;
|
||||
}
|
||||
// Otherwise prepend 'https://' (or 'http://').
|
||||
// Otherwise prepend 'https://'.
|
||||
return `https://${urlString}`;
|
||||
}
|
||||
|
||||
@ -53,11 +53,13 @@ function renderImportance(val) {
|
||||
function renderLevel(val) {
|
||||
const max = 7;
|
||||
const rounded = Math.round(val);
|
||||
const filled = '■'.repeat(rounded);
|
||||
const filled = '■'.repeat(rounded);
|
||||
const empty = '□'.repeat(max - rounded);
|
||||
return `${filled}${empty}`;
|
||||
}
|
||||
|
||||
|
||||
|
||||
function EducationalProgramsPage() {
|
||||
const location = useLocation();
|
||||
const navigate = useNavigate();
|
||||
@ -103,7 +105,7 @@ function EducationalProgramsPage() {
|
||||
);
|
||||
if (proceed) {
|
||||
navigate('/milestone-tracker', { state: { selectedSchool: school } });
|
||||
};
|
||||
}
|
||||
};
|
||||
|
||||
function getSearchLinks(ksaName, careerTitle) {
|
||||
@ -145,24 +147,86 @@ function EducationalProgramsPage() {
|
||||
loadKsaData();
|
||||
}, []);
|
||||
|
||||
// Filter: only IM >=3, then combine IM+LV
|
||||
useEffect(() => {
|
||||
if (!socCode || !allKsaData.length) {
|
||||
setKsaForCareer([]);
|
||||
return;
|
||||
async function fetchAiKsaFallback(socCode, careerTitle) {
|
||||
// Optionally show a “loading” indicator
|
||||
setLoadingKsa(true);
|
||||
setKsaError(null);
|
||||
|
||||
try {
|
||||
const token = localStorage.getItem('token');
|
||||
if (!token) {
|
||||
throw new Error('No auth token found; cannot fetch AI-based KSAs.');
|
||||
}
|
||||
let filtered = allKsaData.filter((r) => r.onetSocCode === socCode);
|
||||
filtered = filtered.filter((r) => r.recommendSuppress !== 'Y');
|
||||
filtered = filtered.filter((r) => ['IM', 'LV'].includes(r.scaleID));
|
||||
|
||||
let combined = combineIMandLV(filtered);
|
||||
combined = combined.filter((item) => {
|
||||
return item.importanceValue !== null && item.importanceValue >= 3;
|
||||
});
|
||||
combined.sort((a, b) => (b.importanceValue || 0) - (a.importanceValue || 0));
|
||||
// Call the new endpoint in server3.js
|
||||
const resp = await fetch(
|
||||
`/api/premium/ksa/${socCode}?careerTitle=${encodeURIComponent(careerTitle)}`,
|
||||
{
|
||||
headers: {
|
||||
Authorization: `Bearer ${token}`
|
||||
}
|
||||
}
|
||||
);
|
||||
|
||||
if (!resp.ok) {
|
||||
throw new Error(`AI KSA endpoint returned status ${resp.status}`);
|
||||
}
|
||||
|
||||
const json = await resp.json();
|
||||
// Expect shape: { source: 'chatgpt' | 'db' | 'local', data: { knowledge, skills, abilities } }
|
||||
|
||||
// The arrays from server may already be in the “IM/LV” format
|
||||
// so we can combine them into one array for display:
|
||||
const finalKsa = [...json.data.knowledge, ...json.data.skills, ...json.data.abilities];
|
||||
finalKsa.forEach(item => {
|
||||
item.onetSocCode = socCode;
|
||||
});
|
||||
const combined = combineIMandLV(finalKsa);
|
||||
setKsaForCareer(combined);
|
||||
}, [socCode, allKsaData]);
|
||||
} catch (err) {
|
||||
console.error('Error fetching AI-based KSAs:', err);
|
||||
setKsaError('Could not load AI-based KSAs. Please try again later.');
|
||||
setKsaForCareer([]);
|
||||
} finally {
|
||||
setLoadingKsa(false);
|
||||
}
|
||||
}
|
||||
|
||||
// Filter: only IM >=3, then combine IM+LV
|
||||
useEffect(() => {
|
||||
if (!socCode) {
|
||||
// no career => no KSA
|
||||
setKsaForCareer([]);
|
||||
return;
|
||||
}
|
||||
|
||||
if (!allKsaData.length) {
|
||||
// We haven't loaded local data yet (or it failed to load).
|
||||
// We can either wait, or directly try fallback now.
|
||||
// For example:
|
||||
fetchAiKsaFallback(socCode, careerTitle);
|
||||
return;
|
||||
}
|
||||
|
||||
// Otherwise, we have local data loaded:
|
||||
let filtered = allKsaData.filter((r) => r.onetSocCode === socCode);
|
||||
filtered = filtered.filter((r) => r.recommendSuppress !== 'Y');
|
||||
filtered = filtered.filter((r) => ['IM', 'LV'].includes(r.scaleID));
|
||||
|
||||
let combined = combineIMandLV(filtered);
|
||||
combined = combined.filter((item) => {
|
||||
return item.importanceValue !== null && item.importanceValue >= 3;
|
||||
});
|
||||
combined.sort((a, b) => (b.importanceValue || 0) - (a.importanceValue || 0));
|
||||
|
||||
if (combined.length === 0) {
|
||||
// We found ZERO local KSA records for this socCode => fallback
|
||||
fetchAiKsaFallback(socCode, careerTitle);
|
||||
} else {
|
||||
// We found local KSA data => just use it
|
||||
setKsaForCareer(combined);
|
||||
}
|
||||
}, [socCode, allKsaData, careerTitle]);
|
||||
|
||||
// Load user profile
|
||||
useEffect(() => {
|
||||
@ -236,7 +300,6 @@ function EducationalProgramsPage() {
|
||||
|
||||
// Sort schools in useMemo
|
||||
const filteredAndSortedSchools = useMemo(() => {
|
||||
|
||||
if (!schools) return [];
|
||||
let result = [...schools];
|
||||
|
||||
@ -290,7 +353,6 @@ function EducationalProgramsPage() {
|
||||
const isAbility = k.ksa_type === 'Ability';
|
||||
const links = !isAbility ? getSearchLinks(elementName, careerTitle) : null;
|
||||
const definition = ONET_DEFINITIONS[elementName] || 'No definition available';
|
||||
|
||||
|
||||
return (
|
||||
<tr key={idx} className="border-b text-sm">
|
||||
@ -427,7 +489,7 @@ function EducationalProgramsPage() {
|
||||
</div>
|
||||
</div>
|
||||
);
|
||||
} // <-- MAKE SURE WE DO NOT HAVE EXTRA BRACKETS HERE
|
||||
}
|
||||
|
||||
// If no CIP codes => fallback
|
||||
if (!cipCodes.length) {
|
||||
@ -513,8 +575,8 @@ function EducationalProgramsPage() {
|
||||
)}
|
||||
</div>
|
||||
|
||||
<div className="grid gap-4 grid-cols-[repeat(auto-fit,minmax(250px,1fr))]">
|
||||
{filteredAndSortedSchools.map((school, idx) => {
|
||||
<div className="grid gap-4 grid-cols-[repeat(auto-fit,minmax(250px,1fr))]">
|
||||
{filteredAndSortedSchools.map((school, idx) => {
|
||||
// 1) Ensure the website has a protocol:
|
||||
const displayWebsite = ensureHttp(school['Website']);
|
||||
|
||||
@ -548,8 +610,8 @@ function EducationalProgramsPage() {
|
||||
</div>
|
||||
);
|
||||
})}
|
||||
</div>
|
||||
</div>
|
||||
</div>
|
||||
);
|
||||
}
|
||||
|
||||
|
||||
BIN
user_profile.db
BIN
user_profile.db
Binary file not shown.
Loading…
Reference in New Issue
Block a user