Workflow Builder

LenserFight's workflow builder is a visual node-graph editor for automating AI pipelines. Each workflow is a directed acyclic graph (DAG): nodes perform discrete operations, and edges carry data downstream from one node to the next.

Canvas Basics

[Trigger] ──► [Prompt Template] ──► [Summarizer] ──► [HTTP Request]
                                         │
                                    [IF Condition] ──true──► [Notify]
                                         │
                                      false──► [Stop]

Open a workflow from the sidebar or create one with New Workflow. The canvas renders your graph. Click a node to open its configuration panel on the right. Drag from a node's output port to another node's input port to draw an edge. Edges are typed — a node that produces text cannot connect directly to a node that expects a list without an intermediate Extract Field or Split in Batches node.

Zoom and pan with the scroll wheel and middle-click drag. Use the minimap (bottom-right) to navigate large graphs. Press Ctrl+Z / Cmd+Z to undo.

Node Types

Category	Node	What it does
Triggers	Manual	Starts a run when you click Run in the toolbar
Triggers	Schedule	Fires on a cron expression; supports retry and concurrency policy
Triggers	Webhook	Starts a run when an external system POSTs JSON to the trigger URL
Triggers	Event	Reacts to a platform event (e.g. battle.closed, lens.run)
AI	Prompt Template	Resolves a lens template with supplied parameters and sends it to a configured model
AI	Summarizer	Condenses a block of text to a target length using a model
AI	Classifier	Routes input to one of N labelled output paths based on model classification
AI	Translator	Translates text between languages using a model
Data	Filter	Passes only items matching a condition
Data	Sort	Orders a list by a field
Data	Aggregate	Reduces a list to a count, sum, min, max, or concatenation
Data	Extract Field	Pulls a single field out of a JSON object using a dot-path
Data	Deduplicate	Removes duplicate items from a list by a key field
Data	Text Splitter	Breaks a long string into overlapping or non-overlapping chunks
Data	Split in Batches	Splits a list into sub-lists of a fixed size for downstream processing
Flow	IF Condition	Evaluates a boolean expression and routes to the true or false branch
Flow	Switch	Routes to one of N branches based on an expression value
Flow	Merge	Waits for all upstream branches to complete, then combines their outputs
Flow	Try-Catch	Wraps a sub-graph; the catch branch runs if any node inside throws
Flow	Stop / Return	Terminates the run immediately and optionally emits a return value
I/O	HTTP Request	Sends a GET, POST, PUT, PATCH, or DELETE request and returns the response
I/O	GraphQL	Runs a GraphQL query or mutation against a configured endpoint
I/O	Object Storage	Reads or writes a file to the connected storage bucket
Integrations	Supabase Query	Runs a parameterised SQL or RPC call against your Supabase project
Integrations	GitHub Issue	Creates or updates a GitHub issue via the GitHub REST API
Integrations	Linear Issue	Creates or updates a Linear issue via the Linear API
Integrations	Notion Write	Appends a block to a Notion page via the Notion API

Running Workflows

Manual

Click Run in the toolbar. If the workflow's root trigger has defined inputs, a dialog prompts you to supply them before the run starts. The canvas activates and edges light up as the execution progresses.

Scheduled

Add a Schedule trigger node and enter a cron expression (e.g. 0 9 * * 1 for every Monday at 09:00 UTC). Open the trigger panel to configure:

• Retry policy — maximum retries and backoff strategy on failure.
• Concurrency policy — whether a new run may start while a previous run is still executing (allow, skip, or queue).

Schedules activate when the workflow is saved and published. Pause or delete the schedule trigger to stop future runs without deleting the workflow.

Webhook

Set the trigger type to Webhook. Copy the URL shown in the trigger panel. External systems POST a JSON body to that URL to start a run. The JSON body is available as trigger.body inside the graph.

curl -X POST https://api.lenserfight.com/webhooks/<token> \
  -H "Content-Type: application/json" \
  -d '{"repo": "acme/backend", "event": "push"}'

Single-Node Test

Right-click any node and choose Run This Node. The node executes in isolation using pinned data (if set) or an empty input. Use this to debug a single step without running the full graph.

To pin data on a node, click the node, open the Inputs tab, and click Pin next to any field. Pinned data persists across runs until cleared.

Reading Run Results

After a run completes:

• Green edges show the execution path taken.
• Dashed grey edges show branches that were not taken (e.g. the false branch of an IF Condition).
• Red node border marks a node that threw an error.

Click any node to inspect:

• Inputs tab — the data that entered the node for this run.
• Outputs tab — the data produced by the node.

The Event Log (bottom panel) shows the full event stream with timestamps, node names, and token counts for AI nodes. Filter by node or event type to narrow the view.

Export a run summary by clicking Export in the event log — the download is a JSON file containing all node inputs, outputs, and timings.

Cost

Each AI node call (Prompt Template, Summarizer, Classifier, Translator) consumes tokens and bills against the API key attached to the node's provider configuration. Set a Budget in Settings → Budget to cap total spend per run or per month. When a run would exceed the budget, it is cancelled and the event log records BUDGET_EXCEEDED.

To review cost for a completed run, open the Event Log and look at the Cost column, or use the summarize_workflow MCP tool for a concise post-run report.