[docs]: highlight Dreamverse deployment paths + add Server B200 (SSH) guide - #1409
Conversation
Merge ProtectionsYour pull request matches the following merge protections and will not be merged until they are valid. 🟠 PR merge requirementsWaiting for
Waiting checks:
|
There was a problem hiding this comment.
Code Review
This pull request updates the documentation in the main README and the Dreamverse-specific README to include comprehensive deployment instructions for the Dreamverse realtime video generation and editing platform. The feedback suggests improving the server installation instructions by using the official standalone installer for uv to avoid PEP 668 errors on modern Linux distributions, and clarifying the frontend development workflow when working with a remote GPU backend to ensure the web UI is accessible.
| # 1. Clone (or pull) FastVideo and install the Dreamverse extra | ||
| git clone https://github.com/hao-ai-lab/FastVideo.git | ||
| cd FastVideo | ||
| pip install --upgrade pip && pip install uv |
There was a problem hiding this comment.
On modern Linux server distributions (such as Ubuntu 24.04), running pip install uv directly on the system Python will fail with a PEP 668 externally-managed-environment error. Using the official standalone installer is more robust for server environments.
| pip install --upgrade pip && pip install uv | |
| curl -fsSL https://astral.sh/uv/install.sh | sh && source $HOME/.local/bin/env |
| Serve the web UI from the same host (or point an external frontend at the | ||
| backend): | ||
|
|
||
| ```bash | ||
| cd apps/dreamverse/web && npm ci | ||
| BACKEND_HOST=<b200-host> BACKEND_PORT=8009 npm run dev | ||
| ``` |
There was a problem hiding this comment.
When running the frontend on the remote B200 server via npm run dev, it typically binds to localhost on that server, making it inaccessible from your local browser. Additionally, the frontend needs to be accessed by the user.\n\nA much cleaner and common workflow is to run the frontend locally on your development machine while pointing to the remote GPU backend. For example:\n\nbash\n# Run this on your local machine\ncd apps/dreamverse/web && npm ci\nBACKEND_HOST=<b200-host> BACKEND_PORT=8009 npm run dev\n\n\nAlternatively, if you do run the frontend on the remote server, you will need to set up SSH port forwarding (e.g., ssh -L 5299:localhost:5299 <user>@<b200-host>) to access it locally at http://localhost:5299.
4f6a30f to
9aacee9
Compare
…ight Prep for the Dreamverse release: - apps/dreamverse/README.md: add a "Server B200 deployment (SSH)" section with native (source install) and Docker options, linking to the existing Install / FFmpeg / Quick Start sections for the shared steps and documenting only the server-specific deltas (bind 0.0.0.0, remote BACKEND_HOST, process supervision, port exposure). Add a general first-boot warmup expectation note under "Launch Dreamverse" (torch.compile warmup can take tens of minutes on a cold cache regardless of deploy method; FASTVIDEO_ENABLE_STARTUP_WARMUP=0 to skip while testing). - README.md: add a Dreamverse highlight section and a Key Features bullet linking to the app and the live demo.
9aacee9 to
e9a10c1
Compare
Add a one-line "Deploy on:" navigation line under the intro linking to the local / Server B200 (SSH) / Docker / Modal deploy guides, so the README highlights the deployment options up front without duplicating the Install section.
Summary
Prep for the Dreamverse release by making the deployment story discoverable from both the repo root and the app README.
apps/dreamverse/README.mdssh→uv pip install -e ".[dreamverse]"→ native FFmpeg → export keys →dreamverse-server(+ frontend).docker_build.sh→docker_run.sh.torch.compile+ startup warmup can take tens of minutes on a cold cache;/readyzstays503until it finishes).README.md(root)All commands mirror the existing
scripts/launch,docker/, andinstall_native_ffmpeg.shflows — no new tooling introduced.Notes for reviewers
A few choices called out so contributors can refine:
npm run dev(dev server), matching the existing local quick start. A production server may prefer a built/static UI or the Docker UI image.8009publicly (firewall / reverse proxy / auth) — worth adding for internet-facing servers.tmux/systemd/nohup); nosystemdunit sample yet.Test evidence
pre-commit run --files README.md apps/dreamverse/README.md→ codespell + PyMarkdown Passed.docker_run.shrequiresCEREBRAS_API_KEY/GROQ_API_KEY, maps host8009, mounts the HF cache + outputs;install_native_ffmpeg.shemitsffmpeg-env.sh).