ObsDog Docs · Overview

From storing documents
to improving knowledge.

ObsDog is a local-first knowledge repository for people and AI. It measures what was retrieved, noisy, used, and useful—not only what exists.

Local first Block native Human + AI
Landscape

ObsDog connects the gaps between existing tools.

It does not merely reproduce a document editor or an LLM tracing tool. ObsDog connects knowledge itself with evidence of how that knowledge performs in real work.

ServiceWhat it does wellThe connection ObsDog adds
Local knowledge graphObsidian

Markdown, links, plugins, and personal knowledge workflows

Search exposure, use, and usefulness measurement are not its primary scope

Authoring, sharing, permissions, page analytics, and team workflows

They do not center one evidence chain across block revisions and AI retrieval

Model calls, retrieval traces, datasets, and evaluations

Long-lived human knowledge history and maintenance are not the primary object

Knowledge observabilityObsDog

Connect docs, blocks, retrieval, use, evaluation, and change through stable identity

Currently a private prototype; public distribution and cloud sync are planned

Why it is different

The differentiator is the connection, not one isolated feature.

Stable knowledge identity

Nested authored blocks preserve identity and lineage as they move and change.

Evidence, not inference

A retrieved block is never silently promoted to used or useful.

Feedback becomes maintenance

Useful, not useful, noise, verification, and reasons feed actionable review queues.

Humans and AI, explicitly

Human, agent, system, and evaluator activity remain attributable and distinct.

Measurement contract

Five distinct states explain whether knowledge worked.

  1. 01
    retrieved

    Returned as a search candidate

  2. 02
    exposed

    Shown to a person or agent

  3. 03
    selected

    Deliberately chosen

  4. 04
    used

    Cited, copied, linked, or applied

  5. 05
    useful

    Evaluated as helpful for the task

Each state is an independent observation. Exposed does not mean used, and used does not guarantee useful.

Product status

Start by making solo local use excellent.

Foundation today
  • Account-free local CLI and macOS boundary
  • Stable block, revision, and annotation contracts
  • Search-trace and feedback dashboard prototype
  • stdio and protected HTTP MCP foundations
Next milestones
  • Real local storage and search vertical slice
  • Optional personal device sync
  • Team sharing, permissions, and concurrent editing
  • Signed CLI, macOS, and iOS distribution

The current UI uses representative prototype data. Login, cloud sync, billing, and public packages are not presented as available.

Observable Docs

Measure whether knowledge works—not merely how much exists.

Explore how search quality, noise, usefulness, and knowledge health connect through representative data.

Explore the dashboard