How lx-0/llm-wiki works
llm-wiki compiles personal data into a shared, queryable Obsidian wiki. The main workflow captures agent sessions and other substrates, stages or writes them into the vault, compiles them into linked Markdown, then serves human and agent queries. Documented side branches include curiosity scans, wiki health checks, approved suggestions, remote publishing, surfaced commitments, and analytical self-reports. The desktop app is an optional control surface. Several subsystem paths and links below are based on README documentation rather than inspected implementation; these are cited as documented relationships.
Main entry points: compile.py Knowledge compiler · paths.py Compiled wiki · session-end.py Session hooks · flush_pipeline.py Session staging
Components 25
Capture
- Session hooks
hooks/session-end.py - Transcript extraction
hooks/_transcript.py - Session staging
scripts/core/flush_pipeline.py - Substrate collectors
scripts/collectors - HTML ingestion
scripts/preprocessors/html_ingest.py - Vault materials
scripts/core/paths.py
Compilation engine
- Knowledge compiler
scripts/compile.py - Compile stages
scripts/compile_stages - Compiled wiki
scripts/core/paths.py
Wiki consumers
- Wiki query
scripts/query.py - Curiosity loop
scripts/curiosity/producer.py - Wiki health checks
scripts/lint.py - Commitment surfacing
scripts/compile_stages - Approved suggestions
scripts/suggestions/producer.py - Remote mirror
scripts/publish - Self-reports
scripts/reports/_engine
Desktop control
- Desktop control
desktop/src/main.ts - CLI execution seam
desktop/src/vault/wiki-exec.ts - Desktop query
desktop/src/vault/query.ts - Listener lifecycle
desktop/src/listeners/lifecycle.ts
Runtime foundations
- Runtime configuration
scripts/core/config.py - Vault paths
scripts/core/paths.py - Intent records
scripts/intents/_record.py
Other components
- Operator
- Agent sessions
Connections 29
- Agent sessions → Session hooks: ends session
README.md - Session hooks → Transcript extraction: extracts transcript
hooks/session-end.py - Session hooks → Session staging: stages context
hooks/session-end.py - Session staging → Vault materials: writes session log
scripts/core/flush_pipeline.py - Operator → Substrate collectors: provides substrates
README.md - Substrate collectors → Vault materials: writes raw captures
README.md - Operator → HTML ingestion: submits HTML
scripts/preprocessors/html_ingest.py - HTML ingestion → Vault materials: writes articles
scripts/preprocessors/html_ingest.py - Vault materials → Knowledge compiler: feeds materials
README.md - Knowledge compiler → Compile stages: runs stages
- Knowledge compiler → Compiled wiki: writes articles
README.md - Operator → Wiki query: asks question
scripts/query.py - Wiki query → Compiled wiki: reads index
scripts/query.py - Knowledge compiler → Curiosity loop: queues gap scans
README.md - Operator → Wiki health checks: checks wiki
README.md - Knowledge compiler → Commitment surfacing: surfaces commitments
README.md - Operator → Approved suggestions: approves actions
README.md - Operator → Remote mirror: initiates mirror
README.md - Remote mirror → Compiled wiki: mirrors Markdown
README.md - Operator → Self-reports: runs study
README.md - Self-reports → Compiled wiki: reads substrate
README.md - Operator → Desktop control: uses controls
desktop/src/main.ts - Desktop control → CLI execution seam: invokes CLI
desktop/src/main.ts - Desktop query → CLI execution seam: runs query
desktop/src/vault/query.ts - Desktop control → Listener lifecycle: controls listeners
desktop/src/main.ts - CLI execution seam → Compiled wiki: dispatches wiki command
desktop/src/vault/wiki-exec.ts - Runtime configuration → Vault paths: uses paths
scripts/core/config.py - Session staging → Vault paths: uses session location
scripts/core/flush_pipeline.py - Intent records → Vault paths: resolves record paths
scripts/intents/_record.py