Merge pull request #3749 from omnivore-app/fix/youtube-model

Some fixes for youtube transcripts
This commit is contained in:
Jackson Harper 2024-03-29 20:34:40 +08:00 committed by GitHub
commit bfc1be6546
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
3 changed files with 10 additions and 60 deletions

View file

@ -111,8 +111,7 @@
"voca": "^1.4.0",
"winston": "^3.3.3",
"word-counting": "^1.1.4",
"youtubei": "^1.3.4",
"youtubei.js": "^9.1.0"
"youtubei": "1.3.7"
},
"devDependencies": {
"@babel/register": "^7.14.5",

View file

@ -96,7 +96,7 @@ export const createTranscriptHTML = async (
}
const llm = new OpenAI({
modelName: 'gpt-4',
modelName: 'gpt-4-0125-preview',
configuration: {
apiKey: process.env.OPENAI_API_KEY,
},
@ -107,32 +107,11 @@ export const createTranscriptHTML = async (
{transcriptData}`
)
const chain = promptTemplate.pipe(llm)
const result = await chain.invoke({
transcriptData: transcript.map((item) => item.text).join(' '),
})
let transcriptChunkLength = 0
let transcriptChunk: TranscriptProperties[] = []
for (const item of transcript) {
if (transcriptChunkLength + item.text.length > 8000) {
const result = await chain.invoke({
transcriptData: transcriptChunk.map((item) => item.text).join(' '),
})
transcriptMarkdown += result
transcriptChunk = []
transcriptChunkLength = 0
}
transcriptChunk.push(item)
transcriptChunkLength += item.text.length
}
if (transcriptChunk.length > 0) {
const result = await chain.invoke({
transcriptData: transcriptChunk.map((item) => item.text).join(' '),
})
transcriptMarkdown += result
}
transcriptMarkdown = result
}
// If the LLM didn't give us enough data fallback to the raw template

View file

@ -2644,11 +2644,6 @@
dependencies:
text-decoding "^1.0.0"
"@fastify/busboy@^2.0.0":
version "2.1.1"
resolved "https://registry.yarnpkg.com/@fastify/busboy/-/busboy-2.1.1.tgz#b9da6a878a371829a0502c9b6c1c143ef6663f4d"
integrity sha512-vBZP4NlzfOlerQTnba4aqZoMhE/a9HY7HRqoOPaETQcSQuWEIyZMHGfVu6w9wGtGK5fED5qRs2DteVCjOH60sA==
"@ffmpeg-installer/darwin-arm64@4.1.5":
version "4.1.5"
resolved "https://registry.yarnpkg.com/@ffmpeg-installer/darwin-arm64/-/darwin-arm64-4.1.5.tgz#b7b5c262dd96d1aea4807514e1cdcf6e11f82743"
@ -19876,13 +19871,6 @@ jest@^27.4.5:
import-local "^3.0.2"
jest-cli "^27.5.1"
jintr@^1.1.0:
version "1.1.0"
resolved "https://registry.yarnpkg.com/jintr/-/jintr-1.1.0.tgz#223a3b07f5e03d410cec6e715c537c8ad1e714c3"
integrity sha512-Tu9wk3BpN2v+kb8yT6YBtue+/nbjeLFv4vvVC4PJ7oCidHKbifWhvORrAbQfxVIQZG+67am/mDagpiGSVtvrZg==
dependencies:
acorn "^8.8.0"
jose@^2.0.5:
version "2.0.7"
resolved "https://registry.yarnpkg.com/jose/-/jose-2.0.7.tgz#3aabbaec70bff313c108b9406498a163737b16ba"
@ -30667,13 +30655,6 @@ undici@^4.9.3:
resolved "https://registry.yarnpkg.com/undici/-/undici-4.14.1.tgz#7633b143a8a10d6d63335e00511d071e8d52a1d9"
integrity sha512-WJ+g+XqiZcATcBaUeluCajqy4pEDcQfK1vy+Fo+bC4/mqXI9IIQD/XWHLS70fkGUT6P52Drm7IFslO651OdLPQ==
undici@^5.19.1:
version "5.28.3"
resolved "https://registry.yarnpkg.com/undici/-/undici-5.28.3.tgz#a731e0eff2c3fcfd41c1169a869062be222d1e5b"
integrity sha512-3ItfzbrhDlINjaP0duwnNsKpDQk3acHI3gVJ1z4fmwMK31k5G9OVIAMLSIaP6w4FaGkaAkN6zaQO9LUvZ1t7VA==
dependencies:
"@fastify/busboy" "^2.0.0"
unfetch@^4.2.0:
version "4.2.0"
resolved "https://registry.yarnpkg.com/unfetch/-/unfetch-4.2.0.tgz#7e21b0ef7d363d8d9af0fb929a5555f6ef97a3be"
@ -32339,19 +32320,10 @@ yocto-queue@^1.0.0:
resolved "https://registry.yarnpkg.com/yocto-queue/-/yocto-queue-1.0.0.tgz#7f816433fb2cbc511ec8bf7d263c3b58a1a3c251"
integrity sha512-9bnSc/HEW2uRy67wc+T8UwauLuPJVn28jb+GtJY16iiKWyvmYJRXVT4UamsAEGQfPohgr2q4Tq0sQbQlxTfi1g==
youtubei.js@^9.1.0:
version "9.1.0"
resolved "https://registry.yarnpkg.com/youtubei.js/-/youtubei.js-9.1.0.tgz#bcf154c9fa21d3c8c1d00a5e10360d0a065c660e"
integrity sha512-C5GBJ4LgnS6vGAUkdIdQNOFFb5EZ1p3xBvUELNXmIG3Idr6vxWrKNBNy8ClZT3SuDVXaAJqDgF9b5jvY8lNKcg==
dependencies:
jintr "^1.1.0"
tslib "^2.5.0"
undici "^5.19.1"
youtubei@^1.3.4:
version "1.3.4"
resolved "https://registry.yarnpkg.com/youtubei/-/youtubei-1.3.4.tgz#b9761e33dcc6e0a9569e6628ba1fc48c729636f0"
integrity sha512-xN6p2oddcTpreF/ojU2mChwdiUlV+TwwUL6xgP6lXRuxeGS5MokM1tzRdXCgIpxkzYYNNAWpt7xvPuAUQM0PCg==
youtubei@1.3.7:
version "1.3.7"
resolved "https://registry.yarnpkg.com/youtubei/-/youtubei-1.3.7.tgz#9cb3cf67b506e09bcf6d6a9cfbdf31ad68948dff"
integrity sha512-DBtwjxqDccpDdSzihXltuqXp4NlG4PZoGfilUbbyrfXvRHhMLDEZO91dzTcJdLXwBBv+wovtGr2VA/OikPnw/Q==
dependencies:
node-fetch "2.6.7"
protobufjs "7.2.4"