deployment-and-configuration
Boot, configure, and operate Yuxi Docker, Lite, production, and service-topology workflows safely.
- License 许可
- MIT
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- deployment-and-configuration
- license
- MIT
- metadata
- { "disco-role": "operating" }
- description
- Boot, configure, and operate Yuxi Docker, Lite, production, and service-topology workflows safely.
- disable-model-invocation
- true
02 资源文件 4
- 参考文档references/configuration-and-secrets.md8.9 KB
- 参考文档references/runtime-topology.md6.6 KB
- 参考文档references/troubleshooting.md6.0 KB
- 脚本scripts/check-runtime-health.sh6.4 KB
03 SKILL.md 正文
Deployment and Configuration
Load this sub-skill when the task is about starting, diagnosing, or safely operating a Yuxi deployment rather than changing agent, knowledge-base, CLI, or frontend code.
Owns
- Docker Compose development, Lite, production, and optional OCR profile topology.
.env/.env.prod, startup environment variables, administrator system options, model-provider credentials, CORS, reverse proxy, MinIO public assets, sandbox-provisioner, and OCR service configuration.- Read-only runtime checks for an already started stack.
- Deployment troubleshooting, including logs, health endpoints, image pulls, service dependencies, and service-required verification candidates.
Route elsewhere
- Agent run lifecycle, Skills/MCP/SubAgent behavior, streaming internals, and
sandbox tool semantics: use the
agent-runtimesub-skill after confirming the stack is healthy. - Knowledge-base ingestion, parser behavior, OCR engine output, graph retrieval,
or chunking details: use
knowledge-and-ocrafter confirming required services and credentials. yuxiCLI usage, external API/SSE client examples, and Langfuse evaluation commands: usecli-and-external-integrationafter confirming base URL and authentication assumptions.- Repository editing, tests, lint, release, or changelog workflow: use
repo-development.
Safety gates
- Do not print, paste, commit, or summarize real
.env,.env.prod, JWT, sandbox, database, MinIO, model-provider, API-key, OCR, or Langfuse secrets. Report only variable names and whether values are missing when necessary. - Treat initialization, image-pull, reset, seed, and service restart commands as side-effectful. Run them only after explicit user approval and only in the intended deployment directory.
- Use
scripts/check-runtime-health.shfor read-only checks. It does not start, stop, rebuild, seed, or edit configuration. - Production API-key traffic must use HTTPS at the external boundary. Do not
recommend sending
yxkey_...tokens over plain HTTP except for local-only development. - External model-provider connectivity, Langfuse tracing/evaluation, cloud OCR, and web search checks require credentials, network access, and explicit opt-in.
- GPU-backed OCR services (
mineru-api,paddlex) are optional service profiles; do not represent them as core verified CPU capabilities.
Fast operating workflow
-
Identify the deployment flavor:
- Development:
docker-compose.yml, web onhttp://localhost:5173, API onhttp://localhost:5050. - Lite:
make up-lite, intended to skip heavy knowledge/graph services. - Production:
docker-compose.prod.yml, web and API behind Nginx on port 80. - Optional OCR profile:
--profile alladds MinerU and PaddleX services.
- Development:
-
Check configuration readiness without exposing values. Use
references/configuration-and-secrets.mdfor required variables and where each layer is read. -
If a stack is already running, run the bundled read-only probe from a Yuxi deployment directory:
# development stack scripts/check-runtime-health.sh --project-dir . --dev # production stack scripts/check-runtime-health.sh --project-dir . --prodIf the script is copied outside the skill tree, keep its relative path local to the copied location and pass
--project-dirto the deployment directory. -
For failures, gather only bounded diagnostics:
docker compose psdocker compose logs --tail=100 api worker web sandbox-provisioner- API health endpoint response from
/api/system/health - Browser-visible web entrypoint status
-
Escalate to the routed sub-skill only after the relevant service health and credential gate is understood.
References
references/runtime-topology.md— service maps, Compose flavors, ports, Makefile entries, optional profiles, and native service-required candidates.references/configuration-and-secrets.md— startup env, administrator system options, model providers, CORS/proxy, MinIO, sandbox, OCR, API key, and Langfuse safety rules.references/troubleshooting.md— symptom-driven diagnosis and safe next commands.scripts/check-runtime-health.sh— read-only health/log probe for a running deployment.
Native verification candidates
Use these only as verification candidates in an active Yuxi checkout or service environment; this sub-skill remains self-contained for operating context.
| Candidate | Requirement | Proves |
|---|---|---|
docker-dev-runtime |
Docker Compose development stack; service-required | Compose health and container logs support dev deployment guidance. |
docker-prod-runtime |
Production Compose plus .env.prod secrets; service-required/reference-checked |
Production topology, Nginx proxy path, and secret gates are understood. |
model-provider-connectivity |
Network and explicit provider credentials; optional/external | Real provider calls are only tested when the user enables them. |
ocr-config-center-e2e |
Running API/Postgres and authenticated user; optional/service-required | OCR configuration endpoints reflect runtime env/admin settings. |
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).