docs: embed instructional video clips, and fix the integration install instructions - #44
Open
nellins wants to merge 44 commits into
Open
docs: embed instructional video clips, and fix the integration install instructions#44nellins wants to merge 44 commits into
nellins wants to merge 44 commits into
Conversation
Eight muted looping MP4 demos (Screen Studio takes cut in the bm-video Remotion pipeline), embedded via a new VideoLoop content component that mirrors ThemeImage. Purely additive: no existing content modified. - quickstart-cloud: workspace intro (2:53) - note-editor: editing modes, frontmatter, command palette - comments-and-suggestions: suggest edits + AI review - mcp-app: Pocketbook tour - themes: theme picker + dark mode - web-app: import flow (Import data section) Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…deo (v30) Requested by Drew. Same :video embed pattern, now served from the docs repo; audio track kept since the player has controls. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Author
|
Added in 72ac0b6 at Drew's request: the what-is-basic-memory page's explainer embed now serves the current hero video (v30) from this repo ( |
Per Drew's review: no autoplay, no loop — controls, scrub, and fullscreen instead. VideoLoop renamed DocsVideo to match. Every embed gets a poster of its title card (frame 0 is the blank cream before the card springs in); the hero gets a mid-video poster frame. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Author
|
Follow-up per Drew's review of the embeds: autoplay and looping are gone — videos now present as their title-card poster with full controls (play, scrub, fullscreen). Component renamed VideoLoop → DocsVideo accordingly; poster frames added under public/videos/posters/ (544KB total). |
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
… page Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Product naming change per Drew. 33 references across 11 pages, including both page titles (nav labels follow). URLs/slugs intentionally unchanged (no redirect infra) — /cloud/mcp-app and /whats-new/interactive-mcp-app keep their paths. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Author
|
Heads-up for review: this branch now also carries Drew's product-naming change — Interactive MCP App → Pocketbook throughout (33 refs, 11 pages, titles/nav included). Slugs left unchanged to avoid breaking links; if you want the URLs renamed too that needs redirects and can be a follow-up. |
Per Drew: hero video on the Welcome page and the Cloud guide (both introduce Basic Memory broadly); the comments-and-suggestions clip on AI Collaboration (its second act is exactly that page's review workflow); the Pocketbook clip on the ChatGPT integration page (the take was recorded in ChatGPT). Audited and skipped: frontmatter on concepts/knowledge-format (wrong context), themes/import/palette elsewhere (passing mentions only). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…o frame All nine embeds now present identically: cream card, wordmark, serif title. Rendered from the bm-video StatementCard component. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
'One knowledge base for you, your AI, and your team.' — frame 2.2s of the hero itself. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Zero new recording — the marketing takes (semantic search, structured search, schemas, knowledge graph) recut with docs treatment: title cards, caption pills, click-to-play posters. Placements: schema-system, semantic-search (also on the v0.23 What's New page — the search release), metadata-search, and the web-app Explore-the-graph section. Raw full-display footage, so no cream padding — cards and pills carry the family style. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Spliced from Drew's two takes: browse existing schemas → ask the AI to infer a Research schema from the research notes → proposed YAML with reasoning → save and validate (5/5 pass) → the schema note in the Schemas folder and conforming notes in Source view. Replaces the marketing recut, which showed validation only. 90s. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…g cut Per Drew's frame-by-frame review: the Schemas-folder click is now on camera, redundant note-hopping and a wrong-note detour are excised, and the clip ends frozen on a research note's type: research frontmatter in Source view. 77s. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…ct; pill says 'This project' not 'Lighthouse' Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
From Drew's new take: sidebar Graph click on camera, Project dashboard, Structured tab, Explore with insights panel, zoom, node focus. Replaces the marketing recut that started inside the graph. 28s. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Folder filter → ⌘K with the scope-pill click on camera → in-project results → open the found note → ⋮ Pin Note → the Pinned tab payoff. Search shot ends before the full query surfaces cross-project results. 29s. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The Pinned tab briefly re-populated with all notes instead of staying filtered (~26.3s in the source take). Clip now ends at the last clean frame — Pinned tab correctly holding just the one note. 28s. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The Pinned TAB click (needed to demo the tab payoff) is what triggers a stale-list render bug downstream, so that beat is cut entirely rather than risked — the clip now ends on the sidebar's PINNED section holding the note. Closing pill retargeted to match. 22.6s. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Extended the pin beat to show the drag gesture on the sidebar's pinned entry (tooltip: 'Drag to move note') — it snaps back rather than completing an unpin, but it's the closest thing to 'unpin from the sidebar' in the take. Freeze point moved to a settled frame (20.6s, no tooltip) — the previous freeze at 19.3s landed mid-drag, which looked like a cursor glitch. 23.6s. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Single 14.7s take: sidebar Activity click → the feed dwelt at reading pace, showing recent notes and AI-made edits. 19s cut. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Recorded in the Claude desktop app (the web app shows Anthropic's prompt-injection warning banner whenever Pocketbook fills the composer). Cut for clicks rather than conversation: open → folder grid → folder → note → full-screen editor with an in-place edit → 'Use this folder' setting the agent's write location. 49s, native 1844x1080 canvas. Adds a short 'Open your knowledge base in the conversation' block to the page's Try It section — the page had no Pocketbook mention at all, so the clip needed somewhere to live. Copy matches the existing prompt pattern; flagging it for review since it's new prose, not just an embed. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Author
|
One flag for review in 3b1c114: the Claude Desktop integration page had no mention of Pocketbook at all, so the new clip needed somewhere to live. I added a short |
Per Drew: the cut skipped the click that hands a note to Claude. Now shows the full arc — note opens, 'use note in conversation' drops it into the composer, question typed and sent, Claude answers from the note's actual contents. 57s. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Opens File history, loads the version timeline, compares the diff, and pulls an earlier line forward with the revert controls until the note is restored. 32s. Scope note: the clip ends on the restored note without showing the Apply click, because Apply is currently broken (basic-memory-cloud #1829 — the revert controls themselves put unsaved changes in the editor, which disables Apply). Nothing shown is inaccurate, but re-take once that's fixed so the clip matches the page's prose, which names Apply as the commit step. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Author
|
Added the File history clip in 06f1953 (page: /cloud/file-history). One scope note worth knowing at review time: the clip ends on the restored note without showing the Apply click, because Apply is currently broken — see basic-memory-cloud#1829. The diff's Nothing in the clip is inaccurate — it shows the timeline, the diff, and pulling an old line forward, and the note genuinely ends up restored (the reverted text autosaves). But the page's prose names Apply as the commit step, so once #1829 is fixed this one is worth a short re-take so video and text line up. |
Full recovery arc in one take: create a snapshot, delete a note, restore just that file from the snapshot, and find it back in the folder. 52s. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Full link lifecycle in one take: share from the note menu, the warning copy, Create Link and Copy, the public page opened in a logged-out browser, then Settings → Shared Notes → Stop sharing until the list reads 'No shared notes yet'. 37s. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
That page is the decision tree between the two recovery tiers and had no media; both clips already exist, so each section now shows its own. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The connector flow is step 2 of the quickstart and the only multi-step flow that happens in a third-party UI — directory, install, OAuth consent — and it had one static screenshot. 49s, ending on Claude listing the real workspace. Password and code entry are rate-compressed; the consent screen plays at natural speed since that's the moment worth reading. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Per Drew: the first cut jumped from the Claude home straight to the Connectors page, skipping the account menu → Settings → Connectors navigation. That IS the instruction. Now all of it plays at 1x; only the take's incidental detour into Skills is rate-compressed. 57s. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Live footage now ends at source 78s and holds that frame — Drew starts typing a follow-up at ~80.5s, which shouldn't be in the clip. 55s. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…ction New: the ChatGPT plugin-directory flow — account menu → Settings → Plugins → Browse directory → Install → OAuth consent → connected → the @basic Memory Cloud mention in a chat. Every navigation click at 1x. Also moved the Claude clip from above 'Choose your assistant' down under '### For Claude', so each subsection carries its own video instead of one clip floating above both. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Drew re-recorded because the first take showed too much browser and bookmark chrome. The new take's ChatGPT windows are clean, but it ends mid-question with no answer — so this is a splice: take 2 for the whole connection flow, take 1 for the chat payoff. The 20s of emailed-code entry is compressed 5x; the Connect modal plays at 1x. 50s. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Per Drew: end once the @ mention is sent — the reply isn't needed, the connection is already proven by the install. And the blinking code field is down from 20s of source to about 2s on screen. 38s. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Dropped the spliced chat ending — Drew spotted the seam, and mixing two recordings read as exactly that. The clip is now one continuous take ending on 'Basic Memory Cloud plugin installed', which is the proof the page needs anyway. 35s. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Per Drew: simultaneous multi-user editing is no longer supported — it was causing sync errors — and the docs still promised it. Removed: - teams/about: the CRDT/Yjs paragraph promising live keystroke sync and per-collaborator presence avatars; heading is now 'Working together' and describes what actually happens (shared knowledge base + Activity feed). The AI-agents and Activity subsections are unchanged and still accurate. - teams/about intro: 'Edit notes together in real time' dropped. - welcome: Teams card no longer advertises 'real-time editing'. Left alone deliberately: every other 'real time' in the docs refers to file sync and indexing, which is a different feature and still true. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Author
|
Drew flagged that real-time collaborative editing is no longer supported (it was causing sync errors), but the docs still promised it. Removed in the latest commit:
Deliberately left alone: every other "real time" in the docs refers to file sync and indexing (local/5.user-guide, claude-desktop, docker, configuration, technical-information). That's a different feature and still accurate — worth confirming you agree with that line, since it's the judgment call in this change. Happy to split this into its own PR if you'd rather keep this one purely about videos. |
Both sets of install steps were verified by running them on a clean machine, and both were wrong. OpenClaw: there is no 'plugins enable --slot' flag (checked against 2026.7.1-2), so the documented command fails outright. Installing already enables the plugin, and the memory slot is a config key, so drop the enable step and show the plugins.slots.memory block instead. Hermes: 'hermes plugins install' cannot install this plugin on ANY released Hermes. The plugin ships manifest_version 2 and the installer caps at 1; the fix is still an open upstream PR, so the previous advice to 'update Hermes' was a dead end. Say so plainly and give the symlink path that works today, since the runtime loads v2 fine. Also removes the duplicate stale claim from Prerequisites, which contradicted the note. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Rehearsed from a genuine pre-install state: 'plugins install' enables the
plugin and claims the memory slot itself. My earlier correction told users
to set plugins.slots.memory by hand, which is unnecessary work and fails
outright if you write it before installing (the config then references a
plugin that does not exist and every command refuses to run).
Keeps a config block only for the real optional case, pointing the plugin
at an existing project instead of the generated openclaw-{hostname} one,
and notes that OpenClaw needs a working model provider first.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Both plugins default to a local project, which neither page said. A reader following either page end to end got a local-only agent and no indication that Cloud was even involved, despite Cloud being the product we point these integrations at. Adds a Cloud section to each with commands verified against the live CLI: sign in, create or reuse a cloud-routed project, name it in the plugin config, confirm via the Route column in 'bm project list'. The two differ and the pages now reflect that. OpenClaw follows whatever routing bm has for the named project. Hermes needs an explicit 'mode: cloud' in ~/.hermes/basic-memory.json, and in cloud mode will not create the project for you the way local mode does. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Setting the project is a one-line CLI call, so hand-editing JSON is not required. Keeps the JSON as a reference for what the command writes. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Both pages had zero media. Adds the setup clip to the Installation sections and the recall/write clip where each page describes what the agent gets, with title-card posters matching the rest of the docs videos. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
nellins
changed the title
The page said 'click Invite Member' with no mention that it greys out at zero available seats. Hit while filming: seat usage read 1 of 1, the button did nothing, and nothing on either the members or billing page connected the two. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
One clip on both sides of the same story: under "Inviting members" for the person sending the invite, and at the top of the join page for the person receiving it. Both pages had no media. Covers invite by email, the pending row, the invitation arriving, accepting and creating an account, landing in a workspace that already has the team's projects, opening a shared note, and the member reading Active back on the inviter's side. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The page described three visibility tiers (workspace / shared / private) set once at creation from the CLI. The app has two (Team / Invite-only), changeable any time from Settings -> Projects via Edit Access, with a Manage Members button for choosing who can see an invite-only project. None of that was documented, and the "Private — just you" tier does not exist in the UI at all. Rewrites the section against what the app does, and answers the question that prompted this: keeping work away from the team is usually a matter of using your personal workspace, which teammates never see. An invite-only project is for work that should live in the team workspace but stay narrow. Also warns that bm project add --visibility is silently ignored — it validates the value and then creates a Team-visible project regardless (basic-memory#1343) — and that the access level cannot be read back from the CLI at all. Sweeps the same stale vocabulary out of the Viewer role description, Shared Notes, and two section cards. Embeds the Teams invite video on the Teams announcement and Teams overview pages, which had no media. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
…ckstart The team-projects page explains Team vs Invite-only entirely in prose; the clip shows a project vanishing from a teammate's sidebar and coming back, which is the part that is hard to picture from a table. Also drops the invite video into the cloud quickstart's "Invite your team" step, where the flow is described but never shown. Teams coverage is now: invite on the announcement, overview, quickstart, inviter and invitee pages; access on team-projects. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Embeds instructional video clips across the docs, and corrects the integration install instructions that turned out to be wrong when actually run.
Videos
23 clips across the Cloud pages, plus two new ones for the OpenClaw integration:
docs-openclaw-cloud— the full path from signed-out to an agent writing into a Basic Memory Cloud project. Five commands, no config editing.docs-openclaw-remembers— a session picking up open tasks from Cloud, then writing a fact back, ending on the changed note in the web app.All clips use the house title-card posters, play with controls, and never autoplay.
Install corrections
These came out of running the documented steps on a clean machine. Each one failed:
plugins enable --slot memorydoes not exist. No such flag in 2026.7.1-2. Installing already enables the plugin and claims the memory slot, so both steps were unnecessary. Worse, setting the slot by hand before installing leaves the config referencing a missing plugin, and everyopenclawcommand then refuses to run.manifest_version: 2; the installer caps at 1. The page told readers to "update Hermes," but the fix is an unmerged upstream PR, so no such release exists. Now states this plainly and gives the symlink path that works. Tracked in Hermes plugin can't be installed via 'hermes plugins install' on any released Hermes (manifest_version 2 vs installer cap of 1) basic-memory#1339.Cloud routing
Both integration pages described cloud routing only in the abstract, while the plugins default to a local project. A reader following either page end to end got a local-only agent with no indication Cloud was involved. Both now have a "Using Basic Memory Cloud" section with commands verified against the live CLI.
The two integrations differ, and the pages now say so: OpenClaw follows whatever routing
bmhas for the named project, while Hermes needs an explicitmode: cloudand will not create the project for you.