mcp-agent-workflows
Use and maintain Vizro-MCP FastMCP server workflows for dashboard creation, validation, schemas, and PyCafe sharing.
- License 许可
- Apache-2.0
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- mcp-agent-workflows
- license
- Apache 2.0
- metadata
- { "disco-role": "operating" }
- description
- Use and maintain Vizro-MCP FastMCP server workflows for dashboard creation, validation, schemas, and PyCafe sharing.
- disable-model-invocation
- true
03 SKILL.md 正文
MCP Agent Workflows
Use this sub-skill when the task involves vizro-mcp, MCP server setup, agent-assisted dashboard generation, dashboard validation, schema extraction, PyCafe links, or repository changes under vizro-mcp/.
Route elsewhere when the task is mainly about:
- Hand-writing a dashboard without MCP:
../core-dashboard-build/SKILL.md. - Chart semantics inside a generated dashboard:
../charts-and-figures/SKILL.md. - Deprecated
vizro-ai:../legacy-ai-chart-agent/SKILL.md.
Package facts
- Import package:
vizro_mcp. - Main evidence file:
vizro-mcp/src/vizro_mcp/server.py. - The installed
vizro_mcp.serverexposesmcp,create_dashboard,validate_dashboard_config, schema/result models, and PyCafe helper functions in this snapshot. vizro-mcpdepends onmcp,httpx,click,pandas[html,parquet,excel],pydantic-settings, andvizro.
Safe first moves
For a user asking to use MCP with Vizro:
-
Explain that Vizro-MCP is the maintained agent-assistance route in this monorepo.
-
Check installed package importability before launching a server:
python - <<'PY' import vizro_mcp.server as s print(hasattr(s, 'mcp'), hasattr(s, 'create_dashboard'), hasattr(s, 'validate_dashboard_config')) PY -
Inspect
vizro-mcp/README.mdfor current server/CLI setup expected by the repository. -
Use validation helpers/schema output before generating or sharing external artifacts.
Dashboard generation/validation principles
- Treat generated dashboard code/config as untrusted until validated by Vizro models and, when feasible, built with
Vizro().build(). - Validate data paths and external URLs before embedding them in dashboards.
- Avoid writing files outside a user-approved output directory.
- Prefer deterministic local validation over agent/model calls when debugging generated code.
- If PyCafe sharing is requested, confirm whether network upload/link creation is allowed.
Schema work
When the task asks for schema/model JSON or LLM-facing dashboard instructions:
- Use
vizro_mcp.serverschema helpers and schema models as the source of truth. - Cross-check with
vizro-coremodel fields usingscripts/inspect_vizro_schema.pywhen changes involve core model definitions. - Keep schema output free of local absolute paths and secrets.
Repository edits
For changes under vizro-mcp/:
cd vizro-mcp
hatch run lint
hatch run test-unit tests/unit/vizro_mcp/test_server.py
If Hatch is unavailable, use the prepared installed-package environment and run focused pytest from vizro-mcp/, but preserve the package working directory.
Common failures
validate_dashboard_configerrors usually mean the generated dashboard does not match currentvizro.modelsschemas. Inspect the model fields and fix the config, not the validator.- Missing Excel/parquet/html dependencies: install/verify
pandas[html,parquet,excel]extras. - MCP server launch issues: separate import errors from transport/server configuration errors.
- PyCafe URL issues: check network permission and the helper in
tools/pycafe/create_pycafe_links_comments.py.
Evidence anchors
vizro-mcp/README.mdvizro-mcp/src/vizro_mcp/server.pyvizro-mcp/src/vizro_mcp/_schemas/schemas.pyvizro-mcp/tests/unit/vizro_mcp/test_server.pytools/pycafe/create_pycafe_links_comments.py
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).