nexent
Route Nexent SDK, backend, frontend, knowledge/memory, and deployment tasks across the zero-code AI agent platform monorepo.
- License 许可
- MIT
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- nexent
- license
- MIT
- metadata
- { "disco-role": "operating" }
- description
- Route Nexent SDK, backend, frontend, knowledge/memory, and deployment tasks across the zero-code AI agent platform monorepo.
- disable-model-invocation
- true
02 资源文件 6
- 参考文档references/package-overview.md2.3 KB
- 参考文档references/repo-provenance.md2.2 KB
- 参考文档references/repo-routing-metadata.json489 B
- 参考文档references/testing-and-verification.md2.4 KB
- 参考文档references/troubleshooting.md2.2 KB
- 脚本scripts/nexent_static_probe.py2.4 KB
03 SKILL.md 正文
Nexent Repo Skill
Use this repo skill when a task involves Nexent, a zero-code platform and SDK for generating and operating AI agents. It routes future agents to the right operating context for SDK runtime, backend APIs, document/knowledge/memory workflows, frontend integration, and deployment operations.
Read first
references/repo-provenance.mdbefore deciding whether this skill is current for a checkout.references/package-overview.mdfor the monorepo layout, package roles, install surfaces, and cross-layer rules.references/testing-and-verification.mdbefore selecting native tests, static checks, or optional live-service verification.references/troubleshooting.mdfor cross-cutting install/import, dependency, env, service, and routing failures.
Route by task
| User task | Use |
|---|---|
Build, test, or debug SDK agents, AgentConfig, ModelConfig, streaming agent_run, model/tool/MCP/A2A setup, sandbox, scheduler, monitoring, or SDK skill manager behavior |
sub-skills/sdk-agent-runtime/SKILL.md |
| Change or diagnose FastAPI routes, services, database helpers, backend env constants, errors, prompt templates, auth/tenant permissions, model/agent/skill APIs, or backend tests | sub-skills/backend-services-api/SKILL.md |
| Work on document ingestion, file splitting/conversion, vector DB, knowledge-base search, MinIO/storage, memory records/retrieval/dreaming, or data-process worker boundaries | sub-skills/knowledge-data-memory/SKILL.md |
Update Next.js pages/components, frontend services/types, chat streaming UI, i18n, stores/hooks, or npm run check-all failures |
sub-skills/frontend-integration/SKILL.md |
| Operate or modify Docker/Kubernetes/offline deployment, env examples, image builds, SQL migrations/init sync, monitoring deployment, uninstall/upgrade behavior | sub-skills/deployment-operations/SKILL.md |
Common setup surfaces
- Python work targets Python 3.11. The SDK package is distributed as
nexent; the backend has its own project metadata and depends on the SDK source for application behavior. - Backend development normally installs backend dependencies and the local SDK package; SDK-only work can install the SDK package alone.
- Frontend work uses the frontend package scripts: development server, type check, lint, format check, and build.
- Full-stack runtime needs deployment-managed services such as PostgreSQL/Supabase, Redis, Elasticsearch, MinIO, data-process workers, and optional monitoring/provider integrations.
Bundled helper
Run scripts/nexent_static_probe.py for a safe high-level checkout probe. It prints package metadata, important directories, and whether sub-skill static helpers can see their expected files. It does not start services, call providers, or run native tests.
Global guardrails
- Keep backend environment-variable reads centralized in backend constants; SDK code receives explicit configuration objects and parameters.
- Do not run live model, MCP, provider, Docker, Kubernetes, Redis, Elasticsearch, MinIO, Ray, Celery, or database checks unless the task explicitly asks and the user supplies the needed environment/credentials.
- Prefer focused pytest/static checks to the full suite unless validating a release or broad refactor.
- For schema changes, coordinate backend models/helpers, migration SQL, fresh-deploy init SQL, and frontend types where public payloads change.
- Keep generated skill usage self-contained: use bundled references and scripts here for orientation, then inspect the task checkout only for the files you are actively changing.
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).