langflower 0.0.1

package/vendor/server/skeleton/skills/langflower-helper/architecture.md

@@ -0,0 +1,96 @@
+# Langflower architecture — trust and transparency
+
+High-level facts for “how it works / why trust it”. Not a package DAG. Prefer
+this file over inventing infrastructure.
+
+## What it is
+
+- Local, **project-scoped** coding agent.
+- The **workflow graph** is the pipeline law (hard harness): stages and edges
+  are authored; the model does not freely skip QA/review/gates outside the
+  graph.
+
+## Layers (user-facing)
+
+1. CLI — `langflower start` (default port **4010**)
+2. Local **server** — owns runtime, project I/O under `.langflower/`
+3. **WebSocket bus** — control plane between server and editor
+4. Angular **editor** — canvas, feed, inspector, Settings, composer
+
+Project product data for Langflower lives under **`.langflower/`**.
+
+## WS bus (shared contracts)
+
+- UI and server speak the **same typed event contracts** (shared
+  `langflowerWsConfig` / co-versioned protocol).
+- Client sends **intents** such as `runner.*.requested`,
+  `workflow.*.requested`.
+- Server emits authoritative **snapshots and facts** (e.g.
+  `workflow.current.snapshot`, `runner.*`, `executionFeed`, …).
+- No parallel “UI-only” DTO protocol for the main control plane; not a hidden
+  REST RPC for runner/workflow orchestration.
+- Shared contracts keep editor and runtime aligned and inspectable.
+
+## Server-first state / thin UI
+
+- Authoritative workflow, runner, and feed state live on the **server**.
+- The editor is a **stateless thin client**: on connect it is hydrated with
+  **initial snapshots**, then stays in sync by **event sourcing** over the WS
+  bus (folding server facts into local projections — not the source of truth).
+- **Closing or reopening a browser tab does not stop a running workflow** while
+  `langflower start` / the server process stays up. Reconnect → fresh snapshots
+    - continued facts.
+- Soft **Pause**, Hard **Stop**, and checkpoint discard are **explicit** runner
+  controls — do **not** equate “tab closed” with Stop. Process kill / reboot is
+  different again (see Knowledge base §7).
+
+## Execution model
+
+- Nodes and edges run on a **reactive runtime**: ports are live streams. A node
+  may receive and emit **at any time** while the run is alive — not a single
+  batch “call node → return once”.
+- **HITL loops** (Review Gate, feedback ports, composer Start/Send) are normal
+  **internal waiting state of nodes**, not a separate product “interrupt the
+  engine” mechanism.
+- LLM provider idle/HTTP 5xx/network failures retry from the last committed
+  round, then keep the node alive for **Steer / Resume**. Partial draft text is
+  telemetry, not committed history. Authentication/configuration failures are
+  terminal. Generic reload of any failed non-LLM node is not yet shipped.
+- Also on the bus: runner start / resume / interrupt intents, `permission.ask`,
+  and related facts.
+
+## Agents on the canvas
+
+- LLM / agent nodes use **skills** and harness tools as wired in the graph.
+- **Sub-Agent** is an **explicit canvas node** for control and observability
+  (own toolLog/response, visible spawn) — not hidden work inside one LLM turn.
+- **Graph order** stages the pipeline; the model does not invent the stage
+  sequence outside the graph.
+- Project memory / wiki create is **queue-driven**: index with harness `glob`,
+  persist work units under `.langflower/memory/history/work-queue.md`, then
+  serial Sub-Agent survey/write (`kb-create` pattern). Memory tools are wired
+  packs (no permission ask). Do not equate “wiki” with embedding search.
+
+## Extensibility
+
+- Custom nodes use the same **`@langflower/node-sdk`** path as built-ins.
+- **TypeScript** is first-class; `tsc` / IDE types act as compile-time gates
+  before Custom → **Update**.
+- Plain JS / Go / Python are not the authoring path. Sandboxed arbitrary
+  user-node execution and a custom-pack compiler are **not** claimed as
+  shipped Implementable here.
+
+## What stays local
+
+- No required cloud account for Langflower itself.
+- Secrets via config / `{env:VAR}`; Settings does not reveal a saved API key.
+- Provider calls leave the machine when the user configures a remote provider —
+  that is expected LLM traffic, not a Langflower cloud login wall.
+
+## Honesty boundary
+
+- Product Status can be **Partial** / Missing / Draft for some use-cases.
+- This file is a **trust-oriented summary**, not a promise that every use-case
+  is Implementable with a real LLM.
+- If a detail is not here and not in the helper Knowledge base → **not in my
+  facts**.

package/vendor/server/skeleton/skills/langflower-helper/layout.md

@@ -0,0 +1,51 @@
+# Langflower editor layout — where can I…?
+
+Facts for UI chrome and on-disk paths. Prefer this file over guessing.
+
+## Editor chrome map
+
+- **Topbar** — workflow list (switch / create / rename as shipped), gear →
+  **Settings**.
+- **Left palette** — built-in nodes; **Custom** packs; **Update** reloads
+  custom packs after disk changes.
+- **Center** — canvas (nodes, edges, selection).
+- **Right aside** — mutually exclusive modes: **feed** | **inspector** |
+  **Settings**. Opening Settings (gear) swaps out feed/inspector; canvas stays.
+- **Composer (bottom)** — **Start** / Hard **Stop** / soft **Pause**; HITL
+  input and actions; `permission.ask` Allow/Deny while a run needs them.
+
+## Where can I…?
+
+| Goal                                   | Where                                                                                                                     |
+| -------------------------------------- | ------------------------------------------------------------------------------------------------------------------------- |
+| Change provider / API key              | Gear → **Settings** (right aside), or edit `.langflower/langflower.jsonc` (prefer `{env:VAR}`)                            |
+| Start a Chat Input graph               | Composer **Start** — not plain **Run** (Run stays disabled for those graphs)                                              |
+| See run output / stream                | Right aside **feed**                                                                                                      |
+| Answer HITL / send into a waiting node | Composer + feed context                                                                                                   |
+| Allow or Deny `permission.ask`         | Composer permission controls                                                                                              |
+| Soft-pause last feed agent / continue  | Composer **Pause** (per-node) / Send or Resume (soft pause ≠ Hard Stop)                                                   |
+| Hard-stop a run                        | Composer **Stop**                                                                                                         |
+| Edit selected node params / ports      | Select node → right aside **inspector** (not Settings)                                                                    |
+| Add or reload custom nodes             | Put pack under `.langflower/nodes/<pack>/`, then Custom → **Update**                                                      |
+| Switch workflow                        | Topbar workflow list                                                                                                      |
+| Open Settings without leaving canvas   | Gear — Settings in right aside                                                                                            |
+| Find sample coding graphs              | Skeleton / demo-project workflow files — **manual copy** into project, then load → Start. No Sample workflows catalog UI. |
+| Project instructions for agents        | `.langflower/instructions.md`                                                                                             |
+
+## On-disk `.langflower/` map
+
+| Path               | Role                                     |
+| ------------------ | ---------------------------------------- |
+| `langflower.jsonc` | Providers, secrets refs, project config  |
+| `workflows/`       | Saved workflow graphs                    |
+| `skills/`          | Agent skills (e.g. `langflower-helper/`) |
+| `nodes/`           | Custom node packs (e.g. `my-nodes/`)     |
+| `instructions.md`  | Project-level agent instructions         |
+| `runs/`            | Run artifacts / checkpoints when created |
+
+## Out of chrome
+
+- Source and other files **outside** `.langflower/` are the user’s project —
+  bootstrap does not rewrite them.
+- Closing the browser tab is not “where is Stop” — see companion
+  `architecture.md` (server-first runs) and Knowledge base §7.

package/vendor/server/skeleton/skills/langflower-node-writer/SKILL.md

@@ -0,0 +1,56 @@
+---
+name: Langflower node writer
+description: >-
+    Guides authors writing project custom nodes under .langflower/nodes packs
+    (defineNode / defineToolRegistrations / defineReactiveNode, ADR-030 layout,
+    reload via Custom Update).
+---
+
+# Langflower node writer
+
+You help the user write **custom nodes** for Langflower.
+
+## Honesty (do not invent)
+
+- Custom nodes are **TypeScript** via `@langflower/node-sdk` only — first-class
+  language; `tsc` / IDE types are the compile-time validators. **Not** plain JS
+  as the authoring path, Go, Python, Rust, or other runtimes.
+- Do **not** claim sandboxed execution of arbitrary user-node code is shipped.
+- Do **not** claim a custom-pack compiler epic is product-Implementable unless
+  the user’s docs say so — packs reload via Custom → **Update** /
+  `COMPILATION_ERRORS.md` today.
+- Langflower does **not** auto-`npm install` pack dependencies.
+
+## Contract (do not invent forks)
+
+- Default pack: `.langflower/nodes/my-nodes/` (one folder = one `package.json`).
+- Prefer **`defineNode`** for sync/Promise nodes. Use
+  **`defineToolRegistrations`** for LLM-callable `ToolHandle` packs on a
+  `tools` port. Use **`defineReactiveNode`** only when exclusive multi-output
+  branches, streams, or advanced bind wiring are required.
+- Import from **`@langflower/node-sdk`** — never from a generated
+  `nodes/types.ts` or a required `index.ts` barrel.
+- Each `*.ts` / `*.tsx` may `export default` a definition or an array.
+  Skip `*.test.ts`, `*.d.ts`, `dist/`, `node_modules/`.
+- Peer deps on the host SDK / RxJS are supplied by Langflower; author libs go
+  in pack `dependencies`. The user runs `npm install` in the pack.
+- After file changes: Custom section → **Update**. Failures land in
+  `COMPILATION_ERRORS.md` in that pack.
+
+## When drafting a node
+
+1. Pick a stable `type` string (pack-unique) and `displayName`.
+2. Declare `inputs` / `outputs` with `wireType`s; keep `uiSchema` honest
+   (`defineToolRegistrations` owns a `tools` output — no custom inputs).
+3. Implement `execute` (`defineNode`), `tools` handlers
+   (`defineToolRegistrations`), or `bind` (`defineReactiveNode`).
+4. Show the full file the user can paste into `my-nodes/`.
+5. Remind them to Update Custom and fix compile errors from
+   `COMPILATION_ERRORS.md` if present.
+
+## References in the project
+
+- `.langflower/instructions.md`
+- `.langflower/nodes/my-nodes/README.md`
+- Seed demos: `git-diff.ts` (`defineNode`), `git-diff-tool.ts`
+  (`defineToolRegistrations`), `review-gate.ts` (`defineReactiveNode`)

package/vendor/server/skeleton/workflows/advanced-coder.json

@@ -0,0 +1,343 @@
+{
+	"metadata": {
+		"name": "Advanced coder",
+		"description": "Plan → refine → coder → review → QA with internal loops and HITL after QA (successor intent to coding-agent). Partial sample graph; configure a provider before running.",
+		"createdAt": "2026-07-24T00:00:00.000Z",
+		"updatedAt": "2026-07-24T00:00:00.000Z"
+	},
+	"graph": {
+		"viewport": {
+			"x": 0,
+			"y": 0,
+			"scale": 0.55,
+			"width": 1600,
+			"height": 900
+		},
+		"nodes": [
+			{
+				"id": "goal",
+				"type": "common-chat-input",
+				"params": {},
+				"inputs": {},
+				"ui": {
+					"position": {
+						"x": 40,
+						"y": 280
+					},
+					"label": "Goal"
+				}
+			},
+			{
+				"id": "planner",
+				"type": "common-openai-llm",
+				"params": {
+					"rolePreset": "plan",
+					"providerId": "",
+					"model": "",
+					"maxIterations": 8,
+					"maxFeedbackTurns": 4
+				},
+				"inputs": {
+					"systemPrompt": "You are the Planner for a coding change.\nExplore the repo with harness tools (read/glob/grep; limited .md writes).\nIf requirements are missing, ask one focused clarifying question in your response (Clarify Review Gate will reply on feedback).\nWhen ready, emit a concise draft plan on response for red-team critique and the plan Review Gate.\nWhen feedback arrives (clarify replies, red-team attack, plan-gate notes, or result-gate improve notes), revise the plan — do not ignore it."
+				},
+				"ui": {
+					"position": {
+						"x": 300,
+						"y": 200
+					},
+					"label": "Planner"
+				}
+			},
+			{
+				"id": "ask",
+				"type": "common-hitl-review-gate",
+				"params": {},
+				"inputs": {},
+				"ui": {
+					"position": {
+						"x": 560,
+						"y": 40
+					},
+					"label": "Clarify"
+				}
+			},
+			{
+				"id": "red-team",
+				"type": "common-openai-llm",
+				"params": {
+					"rolePreset": "custom",
+					"providerId": "",
+					"model": "",
+					"enabledToolIds": ["read", "glob", "grep"],
+					"maxIterations": 6,
+					"maxFeedbackTurns": 0
+				},
+				"inputs": {
+					"systemPrompt": "You are an independent Planner red-team — not the Planner's co-author.\nYour userPrompt is the Planner's latest draft. Attack scope, missing tests, unsafe assumptions, and overclaim.\nBe hostile but specific. End with 2–4 bullet findings. Do not rubber-stamp."
+				},
+				"ui": {
+					"position": {
+						"x": 560,
+						"y": 280
+					},
+					"label": "Planner Red Team"
+				}
+			},
+			{
+				"id": "plan-gate",
+				"type": "common-hitl-review-gate",
+				"params": {},
+				"inputs": {},
+				"ui": {
+					"position": {
+						"x": 560,
+						"y": 520
+					},
+					"label": "Plan Accept"
+				}
+			},
+			{
+				"id": "planner-feedback",
+				"type": "common-merge",
+				"params": {},
+				"inputs": {},
+				"ui": {
+					"position": {
+						"x": 300,
+						"y": 480
+					},
+					"label": "Planner feedback"
+				}
+			},
+			{
+				"id": "coder",
+				"type": "common-openai-llm",
+				"params": {
+					"rolePreset": "coder",
+					"providerId": "",
+					"model": "",
+					"maxIterations": 10,
+					"maxFeedbackTurns": 3
+				},
+				"inputs": {
+					"systemPrompt": "You receive an accepted plan as the user prompt. Implement with harness tools. Prefer minimal edits. Summarize files changed in your final response.\nWhen QA or Review feedback arrives, revise — do not ignore it."
+				},
+				"ui": {
+					"position": {
+						"x": 860,
+						"y": 200
+					},
+					"label": "Coder"
+				}
+			},
+			{
+				"id": "qa",
+				"type": "common-openai-llm",
+				"params": {
+					"rolePreset": "custom",
+					"providerId": "",
+					"model": "",
+					"enabledToolIds": ["read", "glob", "grep", "bash"],
+					"maxIterations": 8,
+					"maxFeedbackTurns": 0
+				},
+				"inputs": {
+					"systemPrompt": "You are the QA agent. Your userPrompt is the Coder's latest summary/change context.\nRun project tests via bash when appropriate (respect permission.ask).\nReport pass/fail clearly. Be specific about failures so Coder can revise via feedback."
+				},
+				"ui": {
+					"position": {
+						"x": 1120,
+						"y": 80
+					},
+					"label": "QA Agent"
+				}
+			},
+			{
+				"id": "review",
+				"type": "common-review",
+				"params": {
+					"providerId": "",
+					"model": "",
+					"maxIterations": 5
+				},
+				"inputs": {
+					"systemPrompt": "You are principles / plan-fit Review after QA.\nCall accept only if the change matches the accepted plan criteria and is non-blocking.\nCall feedback with concrete revision notes otherwise."
+				},
+				"ui": {
+					"position": {
+						"x": 1120,
+						"y": 320
+					},
+					"label": "Principles Review"
+				}
+			},
+			{
+				"id": "coder-feedback",
+				"type": "common-merge",
+				"params": {},
+				"inputs": {},
+				"ui": {
+					"position": {
+						"x": 860,
+						"y": 400
+					},
+					"label": "Coder feedback"
+				}
+			},
+			{
+				"id": "result-gate",
+				"type": "common-hitl-review-gate",
+				"params": {},
+				"inputs": {},
+				"ui": {
+					"position": {
+						"x": 1380,
+						"y": 280
+					},
+					"label": "Result HITL"
+				}
+			},
+			{
+				"id": "done",
+				"type": "common-finish",
+				"params": {},
+				"inputs": {},
+				"ui": {
+					"position": {
+						"x": 1620,
+						"y": 280
+					},
+					"label": "Finish"
+				}
+			}
+		],
+		"edges": [
+			{
+				"edgeId": "e-goal-planner",
+				"fromNodeId": "goal",
+				"fromPort": ["message", 0],
+				"toNodeId": "planner",
+				"toPort": ["userPrompt", 0]
+			},
+			{
+				"edgeId": "e-planner-ask",
+				"fromNodeId": "planner",
+				"fromPort": ["response", 0],
+				"toNodeId": "ask",
+				"toPort": ["result", 0]
+			},
+			{
+				"edgeId": "e-ask-planner-merge",
+				"fromNodeId": "ask",
+				"fromPort": ["feedback", 0],
+				"toNodeId": "planner-feedback",
+				"toPort": ["value", 0]
+			},
+			{
+				"edgeId": "e-planner-redteam",
+				"fromNodeId": "planner",
+				"fromPort": ["response", 0],
+				"toNodeId": "red-team",
+				"toPort": ["userPrompt", 0]
+			},
+			{
+				"edgeId": "e-redteam-planner-merge",
+				"fromNodeId": "red-team",
+				"fromPort": ["response", 0],
+				"toNodeId": "planner-feedback",
+				"toPort": ["value", 1]
+			},
+			{
+				"edgeId": "e-planner-plangate",
+				"fromNodeId": "planner",
+				"fromPort": ["response", 0],
+				"toNodeId": "plan-gate",
+				"toPort": ["result", 0]
+			},
+			{
+				"edgeId": "e-plangate-planner-merge",
+				"fromNodeId": "plan-gate",
+				"fromPort": ["feedback", 0],
+				"toNodeId": "planner-feedback",
+				"toPort": ["value", 2]
+			},
+			{
+				"edgeId": "e-resultgate-planner-merge",
+				"fromNodeId": "result-gate",
+				"fromPort": ["feedback", 0],
+				"toNodeId": "planner-feedback",
+				"toPort": ["value", 3]
+			},
+			{
+				"edgeId": "e-planner-merge-feedback",
+				"fromNodeId": "planner-feedback",
+				"fromPort": ["value", 0],
+				"toNodeId": "planner",
+				"toPort": ["feedback", 0]
+			},
+			{
+				"edgeId": "e-plangate-coder",
+				"fromNodeId": "plan-gate",
+				"fromPort": ["response", 0],
+				"toNodeId": "coder",
+				"toPort": ["userPrompt", 0]
+			},
+			{
+				"edgeId": "e-plangate-review-task",
+				"fromNodeId": "plan-gate",
+				"fromPort": ["response", 0],
+				"toNodeId": "review",
+				"toPort": ["task", 0]
+			},
+			{
+				"edgeId": "e-coder-qa",
+				"fromNodeId": "coder",
+				"fromPort": ["response", 0],
+				"toNodeId": "qa",
+				"toPort": ["userPrompt", 0]
+			},
+			{
+				"edgeId": "e-qa-coder-merge",
+				"fromNodeId": "qa",
+				"fromPort": ["response", 0],
+				"toNodeId": "coder-feedback",
+				"toPort": ["value", 0]
+			},
+			{
+				"edgeId": "e-qa-review",
+				"fromNodeId": "qa",
+				"fromPort": ["response", 0],
+				"toNodeId": "review",
+				"toPort": ["result", 0]
+			},
+			{
+				"edgeId": "e-review-coder-merge",
+				"fromNodeId": "review",
+				"fromPort": ["feedback", 0],
+				"toNodeId": "coder-feedback",
+				"toPort": ["value", 1]
+			},
+			{
+				"edgeId": "e-coder-merge-feedback",
+				"fromNodeId": "coder-feedback",
+				"fromPort": ["value", 0],
+				"toNodeId": "coder",
+				"toPort": ["feedback", 0]
+			},
+			{
+				"edgeId": "e-review-resultgate",
+				"fromNodeId": "review",
+				"fromPort": ["response", 0],
+				"toNodeId": "result-gate",
+				"toPort": ["result", 0]
+			},
+			{
+				"edgeId": "e-resultgate-done",
+				"fromNodeId": "result-gate",
+				"fromPort": ["response", 0],
+				"toNodeId": "done",
+				"toPort": ["value", 0]
+			}
+		]
+	}
+}

package/vendor/server/skeleton/workflows/agents-dialog.json

@@ -0,0 +1,134 @@
+{
+	"metadata": {
+		"name": "Agents dialog",
+		"description": "Chat loop with two agents — proposer and critique — then HITL review before finish.",
+		"createdAt": "2026-07-24T00:00:00.000Z",
+		"updatedAt": "2026-07-24T00:00:00.000Z"
+	},
+	"graph": {
+		"viewport": {
+			"x": 0,
+			"y": 0,
+			"scale": 0.85,
+			"width": 1100,
+			"height": 800
+		},
+		"nodes": [
+			{
+				"id": "chat",
+				"type": "common-chat-input",
+				"params": {},
+				"inputs": {},
+				"ui": {
+					"position": { "x": 40, "y": 240 },
+					"label": "Topic"
+				}
+			},
+			{
+				"id": "proposer",
+				"type": "common-openai-llm",
+				"params": {
+					"rolePreset": "custom",
+					"providerId": "",
+					"model": "",
+					"maxIterations": 6,
+					"maxFeedbackTurns": 3
+				},
+				"inputs": {
+					"systemPrompt": "You are the Proposer. Draft a clear short claim or answer for the user's topic. When feedback arrives from Critique or Review, revise — do not ignore it."
+				},
+				"ui": {
+					"position": { "x": 300, "y": 160 },
+					"label": "Proposer"
+				}
+			},
+			{
+				"id": "critique",
+				"type": "common-critique",
+				"params": {
+					"providerId": "",
+					"model": "",
+					"maxIterations": 6
+				},
+				"inputs": {
+					"systemPrompt": "Call tool `feedback` or `accept` — never reply in plain text.\nYou are an independent Critique. Attack contradictions, overclaim, and missing evidence. Call accept only when the packet is defensible."
+				},
+				"ui": {
+					"position": { "x": 560, "y": 360 },
+					"label": "Critique"
+				}
+			},
+			{
+				"id": "review",
+				"type": "common-hitl-review-gate",
+				"params": {},
+				"inputs": {},
+				"ui": {
+					"position": { "x": 820, "y": 200 },
+					"label": "Human review"
+				}
+			},
+			{
+				"id": "done",
+				"type": "common-finish",
+				"params": {},
+				"inputs": {},
+				"ui": {
+					"position": { "x": 1080, "y": 200 },
+					"label": "Done"
+				}
+			}
+		],
+		"edges": [
+			{
+				"edgeId": "e-chat-proposer",
+				"fromNodeId": "chat",
+				"fromPort": ["message", 0],
+				"toNodeId": "proposer",
+				"toPort": ["userPrompt", 0]
+			},
+			{
+				"edgeId": "e-proposer-critique",
+				"fromNodeId": "proposer",
+				"fromPort": ["response", 0],
+				"toNodeId": "critique",
+				"toPort": ["packet", 0]
+			},
+			{
+				"edgeId": "e-chat-critique-assignment",
+				"fromNodeId": "chat",
+				"fromPort": ["message", 0],
+				"toNodeId": "critique",
+				"toPort": ["assignment", 0]
+			},
+			{
+				"edgeId": "e-critique-feedback",
+				"fromNodeId": "critique",
+				"fromPort": ["feedback", 0],
+				"toNodeId": "proposer",
+				"toPort": ["feedback", 0]
+			},
+			{
+				"edgeId": "e-critique-accept",
+				"fromNodeId": "critique",
+				"fromPort": ["response", 0],
+				"toNodeId": "review",
+				"toPort": ["result", 0]
+			},
+			{
+				"edgeId": "e-review-feedback",
+				"fromNodeId": "review",
+				"fromPort": ["feedback", 0],
+				"toNodeId": "proposer",
+				"toPort": ["feedback", 0]
+			},
+			{
+				"edgeId": "e-review-done",
+				"fromNodeId": "review",
+				"fromPort": ["response", 0],
+				"toNodeId": "done",
+				"toPort": ["value", 0]
+			}
+		]
+	}
+}