serving
Guides FastVideo HTTP and WebSocket serving, typed request contracts, health checks, session state, and server configuration.
- License 许可
- Apache-2.0
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- serving
- license
- Apache 2.0
- metadata
- { "disco-role": "operating" }
- description
- Guides FastVideo HTTP and WebSocket serving, typed request contracts, health checks, session state, and server configuration.
- disable-model-invocation
- true
02 资源文件 4
- 参考文档references/http-contract.md1.4 KB
- 参考文档references/streaming-contract.md1.5 KB
- 参考文档references/troubleshooting.md1.9 KB
- 脚本scripts/make_serve_config.py984 B
03 SKILL.md 正文
Serving
Use for fastvideo serve, OpenAI-compatible HTTP generation, streaming WebSocket
sessions, liveness, continuation state, and operator defaults.
Choose the transport
- No
streamingblock: stateless HTTP server with video/image endpoints. - A
streamingblock: WebSocket/v1/streamsessions with JSON control frames and binary fMP4 media. router-serve: separate routing/load-balancing configuration; treat it as deployment infrastructure, not the generation request API.
Launch with a nested config:
fastvideo serve --config serve.yaml
fastvideo serve --config serve.yaml --server.port 9000
Read HTTP contract, streaming contract, and server troubleshooting. Use the bundled config helper to create a safe template; review model, host, port, output, and credential settings before starting a listener.
Operational rules
Health checks should be cheap and should not require a full generation unless an
operator explicitly wants that. HTTP request fields override explicit
default_request fields, which override hardcoded fallbacks; omitted schema
defaults do not become operator intent. Streaming sessions have timeout and
segment caps and terminal states; clients must send the correct opening frame.
Do not expose a server publicly without authentication/network controls supplied by the deployment environment. Prompt enhancement and safety integrations are optional and may require credentials or model files.
04 同仓库更多 Skills
passnet-triton-opt
Make a single PassNet Triton kernel fast AND numerically faithful: performance model (when a replacement can win at all), block/grid/warp tuning, autotune policy, launch overhead, and per-op numeric recipes to pass the dtype baseline tolerances. Use when a pass matches and is correct but speedup below expected, or when correctness fails by small numeric margins.
passnet-solve
END-TO-END playbook for solving one PassNet sample: analyze the computation graphs, decide the optimization strategy (what to fuse, what to replace, what to leave alone), drive the iteration loop, and maximize the sample score. This is the ENTRY skill — invoke it first for any PassNet optimization task; it tells you when to use passnet-pattern-fusion, passnet-triton-opt and passnet-feedback.
passnet-skill
PassNet GPU kernel optimization via compiler passes. Design and implement Triton-based optimization passes, create pass files under ./pass_dir/, self-evaluate with pass_evaluator, and iterate to maximize GPU speedup.
passnet-pattern-fusion
Author PassNet pass files: write patterns that actually MATCH the FX graph, pick fusion regions (which ops to absorb into one Triton kernel), and structure multi-pass submissions with the mandatory shared-dispatch architecture. Use when creating or fixing pattern()/replacement_args()/replacement_func() files, when a pass "failed to match", or when deciding how to fuse multiple kernels/ops.
passnet-orchestrate
Round/batch planning and decision support for PassNet graph optimization. Use for multi-sample triage, worker allocation, eval-budget policy, and post-eval keep/revert decisions. For solving one concrete sample end to end, use passnet-solve as the entry skill; consult this skill only when you need broader planning or a specific decision gate.
passnet-feedback
Fast, deterministic feedback for PassNet work: pre-flight pattern/match verification WITHOUT burning a GPU evaluation, per-node bottleneck analysis of a sample's graphs, and eval-log parsing into per-variant status + estimated score + failure classification. Use BEFORE every GPU evaluation (check_pattern), at the START of a sample (analyze_graph), and AFTER every evaluation (parse_eval_log).