---

name: langflow description: "Use Langflow's visual AI workflow builder, FastAPI backend, LFX executor, SDK, frontend, deployment, and monorepo maintenance workflows." disable-model-invocation: true metadata: disco-role: operating license: MIT

Langflow

Use this repo skill when working with Langflow, the visual workflow builder for AI agents and applications. It covers flow authoring, Python components and bundles, the FastAPI backend, the lightweight lfx executor, REST/SDK clients, the React frontend, deployment/operations, and safe monorepo maintenance.

First Checks

1. Confirm the task is about Langflow specifically, or about a visual AI workflow builder with flows, components, lfx, Langflow REST APIs, Langflow SDK, MCP, Docker deployment, or Langflow repository maintenance.
2. If you are checking whether this skill is current for a checkout, read references/repo-provenance.md.
3. For package topology and distribution/import relationships, read references/package-topology.md.
4. For cross-cutting install/import/runtime pitfalls, read references/troubleshooting.md.

Route by Task

User task	Read
Create, inspect, validate, tweak, import/export, or smoke-run Langflow flow JSON	flow-authoring
Add or update Python components, inputs/outputs, Message/Data returns, extension bundles, manifests, or component index behavior	component-development
Change FastAPI routes, graph execution, services, settings, auth/authz, database models, Alembic migrations, storage, telemetry, or backend runtime errors	backend-runtime
Use or maintain lfx run, lfx serve, lfx-mcp, Flow DevOps commands, stateless execution, or executor validation	executor-cli
Build REST/API clients, use langflow-sdk, stream run results, push/pull flows, normalize flow JSON, or inspect OpenAPI contracts	sdk-and-api-clients
Modify the React/Vite UI, graph workspace, frontend controllers/hooks/stores, icons, or frontend tests	frontend-development
Install, run, configure, Dockerize, upgrade, operate, or troubleshoot a Langflow deployment	deployment-and-operations
Maintain the monorepo: setup, focused tests, formatting, generated files, package versions, migrations, releases, docs, or CI-style checks	repo-maintenance

Common Decisions

• Flow run path: Use flow-authoring for the flow document itself, executor-cli for local lfx run/serve behavior, sdk-and-api-clients for calling a running Langflow server, and backend-runtime for implementation bugs in run endpoints or graph execution.
• Component/UI split: Use component-development for Python component identity, inputs/outputs, bundles, and component indexes. Use frontend-development for palette display, icons, graph node UI, React stores, and frontend contract failures.
• Operations/backend split: Use deployment-and-operations for environment variables, Docker, databases, volumes, secrets, and server startup. Use backend-runtime when editing source routes/services/settings/migrations.
• Maintenance boundary: Use repo-maintenance to choose focused validation and obey monorepo rules, then route into a feature sub-skill for implementation details.

Install And Verify

For normal use, install the public package in a supported Python environment:

uv pip install -U langflow
uv run langflow --help
uv run lfx --help
uv run python -c "import langflow, lfx, langflow_sdk; print('imports ok')"

For source checkout work, follow the repo-maintenance setup route first because this monorepo has multiple Python packages plus a frontend package. Install only provider, vector-store, document, local-model, or deployment extras needed for the selected workflow.

Minimal Public Facts

• Public distributions include langflow, langflow-base, lfx, langflow-sdk, and extracted lfx-* extension bundles.
• Public imports include langflow, lfx, langflow_sdk, and extension bundle modules such as lfx_arxiv, lfx_docling, lfx_duckduckgo, and lfx_ibm.
• langflow CLI commands include run, superuser, copy-db, migration, api-key, and lfx.
• lfx CLI commands include setup (init, login), authoring (create, validate, requirements, upgrade, extension), running (run, serve), and remote (status, push, pull, export) workflows.
• The Python SDK exposes sync and async clients, flow/project CRUD models, run/stream helpers, flow file normalization, typed exceptions, and test helpers.

Safe Defaults

• Prefer focused validation before broad test or service runs.
• Do not execute untrusted flow JSON, custom component code, or Python graph scripts merely to inspect them.
• Do not run credentialed providers, cloud deployments, Docker mutations, destructive migrations, package publishing, or model/GPU workloads unless the user explicitly asks and the environment is safe.
• Do not paste real API keys into flow JSON, docs, examples, .env, .lfx config, shell history, or committed files.

Bundled Helpers

Each sub-skill owns its helper scripts. Run helpers from the nearest sub-skill directory so relative scripts/... paths resolve after import.

• Flow JSON preflight: sub-skills/flow-authoring/scripts/validate_flow_json.py.
• Component skeleton checks: sub-skills/component-development/scripts/check_component_skeleton.py.
• FastAPI route inspection: sub-skills/backend-runtime/scripts/inspect_routes.py.
• LFX flow command template checks: sub-skills/executor-cli/scripts/validate_lfx_flow.py.
• Flow JSON normalization: sub-skills/sdk-and-api-clients/scripts/normalize_flow_file.py.
• Frontend package script checks: sub-skills/frontend-development/scripts/check_frontend_package.py.
• Langflow environment variable checks: sub-skills/deployment-and-operations/scripts/check_env_vars.py.
• Deprecated import checks for repo maintenance: sub-skills/repo-maintenance/scripts/check_deprecated_imports.py.