How ar9av/obsidian-wiki works
obsidian-wiki is an agent-guided knowledge system: an agent uses Markdown skills to capture and maintain knowledge in a local Obsidian vault, then retrieves connected knowledge and session context. The Python package provides CLI, API, and optional remote-server access; the browser extension adds web capture and form filling. The graph preserves these documented product branches, but sampled excerpts do not establish the detailed wiring for most skill-driven workflows, so those links are kept at the documented agent-to-skill boundary.
Main entry points: brain_fill_host.py Form-fill host · session_graph.py Session graph · graphrag.py Graph-aware retrieval · server.py HTTP and MCP
Components 25
Agent workflows
- Wiki skills
.skills - Ingest skills
.skills/wiki-ingest - Update skill
.skills/wiki-update - Query and narrate
.skills/wiki-query - Vault maintenance
.skills/wiki-lint - Conversation capture
.skills/wiki-capture
Knowledge vault
- Markdown vault
- Graph-aware retrieval
obsidian_wiki/graphrag.py - Graph analysis
obsidian_wiki/graph_analysis.py - Context packing
obsidian_wiki/context_pack.py
Memory and sessions
- Memory operations
obsidian_wiki/memory.py - Session sources
obsidian_wiki/session_sources.py - Session index
obsidian_wiki/session_index.py - Session graph
obsidian_wiki/session_graph.py - Session search
obsidian_wiki/session_query.py
Interfaces
- CLI
obsidian_wiki/cli.py - Python memory API
obsidian_wiki/client.py - HTTP and MCP
obsidian_wiki/server.py
Browser tools
- Capture extension
extensions/brain/background.js - Capture and fill UI
extensions/brain/popup.js - Form-fill host
extensions/brain/host/brain_fill_host.py
Other components
- Vault user
- AI agent
- Web page
- Agent CLI providers
Connections 33
- Vault user → AI agent: requests work
- AI agent → Wiki skills: reads and runs
- AI agent → Ingest skills: dispatches ingestion
- AI agent → Update skill: dispatches updates
- AI agent → Query and narrate: asks knowledge
- AI agent → Vault maintenance: maintains vault
- AI agent → Conversation capture: captures conversation
- Ingest skills → Markdown vault: compiles knowledge
- Update skill → Markdown vault: updates pages
- Conversation capture → Markdown vault: saves knowledge
- Query and narrate → Graph-aware retrieval: queries graph
- Graph-aware retrieval → Markdown vault: reads page metadata
- Graph-aware retrieval → Graph analysis: analyzes structure
- Vault maintenance → Markdown vault: checks and cleans
- AI agent → CLI: runs commands
- CLI → Markdown vault: operates on
- Python memory API → Memory operations: delegates memory
- Memory operations → Markdown vault: reads and writes
- HTTP and MCP → Markdown vault: serves vault access
- HTTP and MCP → Graph-aware retrieval: delegates search
- HTTP and MCP → Context packing: delegates packing
- Session sources → Session index: provides session text
- Session graph → Session sources: reads histories
- Session graph → Session index: builds vectors
- Session graph → Graph analysis: detects communities
- Session search → Session graph: loads graph
- Vault user → Capture extension: captures web content
- Web page → Capture extension: provides page content
- Capture extension → Markdown vault: writes raw capture
- Capture and fill UI → Form-fill host: sends native messages
- Form-fill host → Context packing: runs context-pack
- Form-fill host → Agent CLI providers: requests form answers
- Form-fill host → Markdown vault: resolves vault