Repository navigation
Replies: 3 comments
|
Thanks for writing this up with the schema, the command shape, and the open questions named. That is the useful form of an Ideas thread. What already matches the wiki-style layout in
So the gap you named is real: there is no standard, resumable batch that discovers a human-maintained note tree, previews scope, runs ingest in chunks, and leaves the originals alone. Instruction-only "do not write here" is also not a hard boundary today. I am not deciding the product questions here (whether Looping Eugeniu on whether this is in scope for the wiki-style variant. |
|
As proposed in #329, maybe the better way to warmstart is to use the extension of the /obsidian-ingest command as the general case for warm-starting the repo. In #329 it is proposed to enhance the /obsidian-ingest command such that it cannot only ingest a single file but also a folder which is then searched and ingested. Using this on the First-Brain/ folder is equivalent to an explicit /obsidian-warm-start command. |
|
@2Obe this proposal is now built, in the shape you narrowed it to:
So no separate |
Uh oh!
There was an error while loading. Please reload this page.
Proposal: Warm-start an existing Obsidian vault into a wiki-style Second Brain
The existing wiki-style layout provides a useful separation between immutable sources in
raw/and AI-maintained knowledge inwiki/. I would like to discuss extending this architecture with a human-maintained source layer and a warm-start workflow for existing Obsidian users.The goal is to let users continue working with their personal notes (lets call this place "First Brain") while the agent derives and maintains a separate knowledge layer (the already existing second-brain logic) that can help answer questions about those notes.
Proposed vault schema
This would be a variant of the existing wiki-style schema, rather than a replacement.
First-Brain/remains an evolving human workspace, whileraw/preserves the source versions used to derive knowledge.Introducing this layout may involve moving existing notes initially. After that migration, agent processing should leave the human-maintained originals unchanged.
Existing foundation
I have explored this separation through custom
_CLAUDE.mdandAGENTS.mdinstructions, explicit Folder Map routing, and controlled ingestion.The current implementation already provides much of the foundation:
What appears to be missing is a standard workflow that brings these pieces together. Instruction-based protection also does not provide a hard write boundary.
Proposed command:
obsidian-warmstartRather than asking the agent to “ingest everything” in one untracked operation, a dedicated command could:
First-Brain/, respecting exclusions and sensitive-content rules.obsidian-ingestworkflow.raw/and derived knowledge towiki/, leaving originals unchanged.Conflicting claims involving First Brain notes should be recorded in the derived knowledge layer with references to the sources, without modifying the original notes.
Fit with the existing command model
The command would fit into the README’s Operations layer, with
category: meta, alongside initialization and reindexing:obsidian-initobsidian-ingestobsidian-warmstartobsidian-reindexobsidian-initcould detect a populatedFirst-Brain/and offer warm start as an explicit next step. Warm start should remain separately runnable because importing a large collection may require multiple sessions.Potential implementation locations:
references/vault-schema.md: document the human-source variant of wiki-style.commands/obsidian-warmstart.md: define the orchestration workflow.commands/obsidian-ingest.md: clarify source protection and local-source provenance.commands/obsidian-init.md: offer the warm-start workflow when appropriate.The agent would perform extraction and consolidation; the helper would manage bookkeeping.
Questions for discussion
wiki/, consulting First Brain for original evidence?First-Brain/the right convention, or should the source folder be configurable?Continuous background synchronization could follow later. The initial scope would be a controlled, resumable warm start—not an automatic watcher.
Would this be a valuable addition to the project, and does this approach fit its intended architecture?
If the idea is considered valuable, I will create the required issues from this discussion, separating the schema documentation, command orchestration, protection rules, and any follow-up synchronization work.
All reactions