Spaces:
Sleeping
Sleeping
Commit History
-
Merge main: /seed endpoint
667ac4a -
feat(api): add idempotent POST /seed for hosted demo data
Cloud IPs can't scrape YouTube/Instagram, so a deployed instance starts empty. /seed loads the deterministic demo pair (transcripts embedded with BGE) into the DB. Only fills an empty DB — returns existing videos untouched otherwise, so it can't overwrite real ingests. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>e9f8268 -
chore(frontend): patch Next.js (CVE-2025-66478) and postcss advisories
- next 15.1.3 -> 15.5.19 (clears CVE-2025-66478; stays on Next 15, React 19) - postcss devDep -> 8.5.10 and an overrides pin so Next's bundled postcss is also patched (GHSA-qx2v-qp2m-jg93) without downgrading Next npm audit now reports 0 vulnerabilities; production build verified. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>e1dc5fc -
Merge main: fix gitignored app/models package
1fc782d -
fix: app/models package was silently gitignored and never committed
.gitignore had an unanchored "models/" (intended for a root model cache), which git matches at every depth — so it also ignored app/models/, the ORM package (Video, Chunk). The app ran locally only because the files sat untracked on disk; a fresh clone or Docker build (which builds from the git tree) crashed with ModuleNotFoundError: No module named 'app.models'. Anchor the pattern to "/models/" and commit the package. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>b7a4e12 -
chore: make backend deployable on Hugging Face Spaces (free, no card)
HF Spaces Docker SDK gives 16GB RAM free with no payment method, unlike Render's free tier. Adapt the image for it without breaking other hosts: - Dockerfile: run as UID 1000 with a writable HOME/HF_HOME, default port 7860 (HF's expected port; $PORT still overrides for Render et al.) - README: add HF Space frontmatter (sdk: docker, app_port) on this branch only, so main's README stays clean Push this branch to a HF Space's git remote to deploy the backend. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>6cd5ce0 -
Merge pull request #1 from xpushkal/deploy/render-vercel
Unverified signatureAdd Render + Vercel deploy configd095888 -
chore: add Render + Vercel deploy config
Backend (FastAPI + PyTorch/BGE/Whisper) can't run on Vercel serverless (size, timeouts, in-process memory), so it ships as a Docker web service on Render with Neon Postgres+pgvector; the Next.js frontend goes on Vercel. - Dockerfile: CPU-only torch wheel to stay under free-tier image limits - render.yaml: Blueprint with prefilled env + secret placeholders - .dockerignore: keep the backend image lean (no frontend/venv/data) - DEPLOY.md: Neon -> Render -> Vercel -> CORS walkthrough Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>92d98bb
-
feat:test
c97ea54 -
fix:scoring percentage, routing and cold start
5bfb4dd
-
fix:rag pipeline
3341e3b -
fix:youtube metadata format
2fafda7 -
fix:wesbite UI
a829449 -
fix:website ui update
3d67361 -
testing and bug fix
e90c3c4 -
fis:api routing
43a4b46
-
feat:db setup
8d72767 -
feat: implement creator-rag MVP — router-based RAG over two short-form videos
a8b0be6 -
feat: implement creator-rag MVP — router-based RAG over two short-form videos
5718c1e -
feat:Docx
fb97d73 -
feat:Docx
930fb79 -
feat:Docx
df20f56 -
first commit
6a45d8f