Import — manual test checklist

Use self-hosted Hub (npm run hub) or CLI for local file paths. Hosted (gateway): import is bridge-backed when BRIDGE_URL is set on the gateway; see IMPORT-URL-AND-DOCUMENTS-PHASES.md §4V and hub/gateway/README.md. Canonical source_type values: lib/import-source-types.mjs.

Prerequisites

• Configured vault (config/local.yaml or KNOWTATION_VAULT_PATH).
• After large imports: Re-index (Hub) or knowtation index.

Per source type (one happy path each)

Audio / video

1. Hub UI: Audio (transcribe) is available on self-hosted Hub with OPENAI_API_KEY. Video in the import dialog is coming soon; use CLI knowtation import video or transcribe elsewhere and import Markdown.
2. Set OPENAI_API_KEY in the environment of the process running Hub or CLI.
3. Use a Whisper-supported extension: .mp3, .mp4, .mpeg, .mpga, .m4a, .wav, .webm (see lib/transcribe.mjs).
4. File size: OpenAI’s transcription API enforces a 25MB maximum per file (you may see 413 Payload Too Large when the limit is exceeded). Prefer compressed audio (M4A/MP3) for longer recordings; split or downsample before import.
5. Import, then index and search for a phrase from the transcript.

Agents (MCP)

1. MCP import accepts the same source_type strings as the CLI (see import-source-types.mjs).
2. input must be a path visible to the MCP server process (working directory, mounts, container volumes).
3. After import: get_note / search / list_notes to verify; use tiered retrieval per RETRIEVAL-AND-CLI-REFERENCE.md.

Hosted (production)

• If BRIDGE_URL is not set on the gateway, POST /api/v1/import returns 501 with code: NOT_AVAILABLE and an error string (see hub/gateway/server.mjs). The Hub should surface that JSON clearly.
• When BRIDGE_URL is set, the gateway proxies to the bridge — same POST /api/v1/import as self-hosted Hub, including 4A₂, 4B, and 4C (4C is client-only; no extra route). See PARITY-MATRIX-HOSTED.md and IMPORT-URL-AND-DOCUMENTS-PHASES.md §4V.

Hub — Phase 4A₂ (in-browser ZIP), 4B (sequential multi), and 4C (drop zone)

When: npm run hub (or hosted with bridge), signed in, Import modal.

1. 4A₂ — Markdown, two local .md (multi-select): source Markdown → expect one POST /api/v1/import with a hub-bulk.zip (DevTools → Network) and >0 notes.
2. 4A₂ — Choose folder: source Markdown → Choose folder (ZIP in browser) on a small tree of .md → one POST, Re-index if needed, notes present.
3. 4A₂ — ChatGPT: a folder that includes conversations.json (or only that file) → source ChatGPT → one POST with client zip if not already a .zip upload; >0 notes.
4. 4B — Two PDFs: source PDF → multi-select two small PDFs → two sequential POST / import and a Batch: 2 of 2 success line (or failure text per file).
5. 4B — Stop batch: start a 3+ PDF batch, Stop batch before the end; message indicates partial/bail.
6. 4C — Drop folder (Chromium): source Markdown → drag a small folder of .md onto the dashed Drop files or a folder area → status shows file count, Import → same as Choose folder (one hub-bulk.zip POST for tree mode). In Safari / Firefox, directory drag may be flatter; if paths look wrong, use Choose folder or a desktop ZIP instead.
7. Cap: if you have no large fixtures, trust unit tests: node --test test/hub-client-import-zip.test.mjs.

Hosted 4C + bridge (Chromium)

Run after a deploy to production or a Netlify deploy preview (same checks as 4C locally, but against the hosted gateway URL). Prereq: BRIDGE_URL set on the gateway, signed in, small fixture folder of .md files.

Automated regression

• node --test test/import-importers-golden.test.mjs test/import-source-types.test.mjs test/import-markdown.test.mjs test/embedding-usage.test.mjs test/hub-client-import-zip.test.mjs