# Capslane HTTP workflows

These inactive exports use built-in n8n nodes and native transcripts. No community package or catalog approval is required. Configure your own credentials after import. A transcript call uses one request, including a cached result; a failed extraction may also consume the reserved request. Neither workflow silently starts audio generation.

## Google Sheets URL to saved transcript

Import `n8n-sheets-transcript-http.json`. In a test spreadsheet, create a header row with `id`, `url`, `transcript`, `language`, `status`. Give each input row a unique non-empty text `id` and a public YouTube URL. Keep the other cells empty.

1. Select the same document and sheet in New URL row and Save transcript.
2. Select a Google Sheets trigger credential and a Google Sheets OAuth2 credential with access to that document. Re-select the column mapping if n8n asks to refresh the sheet schema; match on `id`.
3. Create a Header Auth credential with Name `x-api-key` and Value your Capslane key. Select it on Get native transcript. Do not paste the key into the workflow JSON.
4. Test with one new row. Verify that the corresponding row receives complete text, its language and `done`. Writing uses RAW cell values so transcript text is not evaluated as a spreadsheet formula.
5. Activate only after this check. The trigger watches added rows, so writing a result back does not retrigger extraction. A manual replay still costs a request. Do not reuse IDs or sort/change IDs during an execution.

The template rejects text longer than 45,000 characters rather than truncating it. Use the Markdown template for longer transcripts. Errors stop the execution before the result is written. Review failed executions and retry only after fixing credentials, URL, quota or upstream availability. Test initially with one row; use bounded batches for imports.

## YouTube to knowledge-base Markdown

Import `n8n-knowledge-base-http.json`, choose the Header Auth credential on Get native transcript and replace the URL in Video URL. Execute manually. Download the binary `data` file from Download Markdown. It contains complete text, source URL, language and links to each timestamped chunk. The JSON before that node is also suitable for mapping to your existing document store.

The sample `dQw4w9WgXcQ` is a public video already used in the Capslane SDK and integration fixtures. Its availability can change. Tests use synthetic transcript responses, never committed copies of video text.

Capslane prepares a source document. It does not produce a summary or write to a vector database. In a retrieval pipeline, retain the source URL and timestamps; treat transcript content as untrusted source material, not instructions to the model. Replace the manual input only after a successful individual execution.

## Missing captions and generated jobs

The supplied HTTP templates use `mode=native`. A missing track is an explicit error. Do not merely change this to `auto`: a 202 response contains a job ID, not a transcript. With the community node, use Get Transcript with Wait for Generated Transcript enabled before the same formatting steps. In a built-in HTTP workflow, persist the accepted job ID, delay, call `GET /v1/transcript/{jobId}` with the same credential and branch on `status`. Continue only on `completed` with `content`; stop on `failed` or `cancelled`; bound waiting to twenty minutes. Check the same job after a temporary polling failure instead of creating another generation.

Free workspaces have 50 monthly transcript requests, 15 generated minutes and one active generation. Polling does not consume another transcript request.

## Validation and references

Schemas are checked against the installed n8n 2.37.7 node implementations: HTTP Request 4.4, Google Sheets 4.7, Sheets Trigger 1, Code 2 and Convert to File 1.1. Both flows executed successfully in that runtime with a synthetic HTTP service on September 7, 2026. The Sheets check substitutes its Google trigger and write with fixture boundaries, verifies two paired rows, and preserves the published HTTP and formatting nodes. The Markdown check also executes Convert to File and verifies the binary output. Repository tests cover invalid input, missing content and the cell-size guard. Live Google account access must be checked in your own test sheet after import.

- https://capslane.com/integrations/n8n
- https://docs.n8n.io/integrations/builtin/core-nodes/n8n-nodes-base.httprequest/
- https://docs.n8n.io/integrations/builtin/app-nodes/n8n-nodes-base.googlesheets/sheet-operations/