Study pipeline management — content, review, sessions, and tracking.
# Study sessions (tmux + AI agent + Textual sidebar)
studyloop study "topic" --energy 7 # Full tmux environment in one command
studyloop study "topic" --mode co-study # Co-study mode (user drives)
studyloop study --resume # Reattach to existing session
studyloop study --end # End session cleanly
studyloop study "topic" --web # Also start web dashboard
studyloop park QUESTION [-t TOPIC] # Park tangential topic
# Low-level session commands (used internally by study)
studyloop session start -t TOPIC -e 7 # Start session (DB + IPC files)
studyloop session status # Timer, topics, parking lot
studyloop session end [-n NOTES] # End session, show summary
studyloop session effectiveness # Persona effectiveness metrics
# Content pipeline
studyloop content split SOURCE # Split PDF by chapters
studyloop content generate-cards DIR --course COURSE # Local quiz/flashcard JSON
studyloop content generate-practice DIR --course COURSE # Local hands-on practice JSON
studyloop content discover # Preview configured study sources
studyloop content ingest --dry-run # Plan source ingest
studyloop content import-review DIR --course COURSE # Import existing JSON artefacts
studyloop content process SOURCE # Legacy optional NotebookLM upload path
studyloop content from-obsidian DIR # Legacy optional NotebookLM path
# Sync & topics
studyloop sync [TOPIC] --all --dry-run # Legacy optional notebook sync
studyloop status [TOPIC] # Show sync status
studyloop topics # List configured topics
studyloop audio TOPIC # Legacy optional audio overview
studyloop dedup [TOPIC] --all --dry-run # Remove duplicate notebook sources
# Review
studyloop now # Recommend one next study action
studyloop now --energy low --time 15 # Smaller, lower-switching recommendation
studyloop now --modality hands-on --interleave adaptive
studyloop chat-note NOTE.md --mode recall # Build a Socratic context pack
studyloop chat-note NOTE.md --mode diagram --voice
studyloop practice verify TASKS.json --task 1 --notes "what passed"
studyloop practice verify TASKS.json --task 1 --run-command --workdir . --timeout 120
studyloop recap today # One win, repair target, due item, next action
studyloop recap today --speak # Speak through study-speak
studyloop recap today --audio-file recap.wav
studyloop mastery graph --topic python # Mermaid concept graph
studyloop mastery graph --topic python --format json
studyloop mastery weak-links --topic python
studyloop review # Check spaced repetition due dates
studyloop review --interleave adaptive --energy medium
studyloop progress # Summarize local course progress
studyloop progress --course python # Summarize one course
studyloop progress --json # Machine-readable progress summary
studyloop progress CONCEPT -t TOPIC -c confident # Record concept confidence
studyloop teachback CONCEPT -t TOPIC --score 3,3,4,3,2 --type structured
studyloop teachback-history CONCEPT [-t TOPIC]
studyloop struggles --days 30 # Find recurring struggle topics
studyloop wins --days 30 # Concepts mastered / confident (AuDHD wins)
studyloop resume # Where you left off (last session summary)
studyloop streaks # Study streak and consistency stats
studyloop backlog list # Pending study backlog (parked/struggled/manual)
studyloop bridge list # Network→DE (or cross-domain) knowledge bridges
studyloop clean --dry-run # Preview orphan tmux/session cleanup
studyloop extract-struggles --incremental --harness kiro --model MODEL_ID
# Study plans
studyloop plan interview # Interview questions + evidence-based seed suggestions
studyloop plan new --title TITLE [--why WHY] [--topic T] [--success S] [--milestone M]
studyloop plan new --title TITLE --activate # Activate on create (refused if incomplete)
studyloop plan list [--status draft|active|paused|complete|abandoned] [--json]
studyloop plan show PLAN_ID [--markdown] [--json]
studyloop plan status PLAN_ID active # Change lifecycle state
studyloop plan milestone PLAN_ID INDEX [--done|--undone] # Toggle or set a milestone
studyloop plan evaluate PLAN_ID [--phase start|mid|end] [--record] [--study-id ID]
studyloop plan reindex # Rebuild the plan index in the sessions DB
studyloop plan path # Print the plan document directory
# Topic exercises (blank slate / completion / multiple choice)
studyloop exercise new --topic TOPIC [--plan ID] [--requirement R] [--reference FILE]
studyloop exercise from-milestone PLAN_ID [--index N] # Draft from a plan milestone
studyloop exercise import PATH # Import a hand-authored exercise document
studyloop exercise list [--plan ID] [--topic T] [--json]
studyloop exercise show SET_ID [--markdown] # Learner-safe (answers withheld)
studyloop exercise show SET_ID --with-answers # Author use only
studyloop exercise review SET_ID [--kind blank_slate|completion|multiple_choice]
studyloop exercise review SET_ID --file ANSWER.py --record # Score + write confidence
studyloop exercise path # Print the exercise document directory
# Focus & retention
studyloop focus # Show current focus topics (max 3)
studyloop focus suggest [--days 30] # Suggest focus from sessions, struggles, config
studyloop focus set "python" "sql window functions" # Replace focus (1-3 topics)
studyloop focus set TOPIC --no-refocus # Save focus only, skip the data movement
studyloop focus apply [--days 30] [--dry-run] # Run/retry the deferred data movement
studyloop focus clear # Clear all focus topics
studyloop prune [--days 30] # Preview trimming old local sessions (dry run)
studyloop prune --days 30 --apply # Actually delete (only verified-in-full sessions)
# Configuration & health
studyloop setup # Interactive setup wizard
studyloop install tools # Install global CLI entrypoints from repo
studyloop install agents # Install agent definitions for detected tools
studyloop config init # Advanced/legacy config initializer
studyloop config show # Display current configuration
studyloop self-test # Lightweight post-install smoke check
studyloop doctor # Full health check
studyloop update # Check for available updates
studyloop upgrade # Apply all available updates
# Backup & restore
studyloop backup [--tag NAME] # Snapshot DB + config to backups/
studyloop restore # List available backups
studyloop restore BACKUP --confirm # Restore from backup (safety backup first)
# Web
studyloop web [--port PORT] [--lan] [--password SECRET] # Launch study web app (PWA)
studyloop web --ttyd-port 7681 # ttyd server transport port (0 = read from config)
studyloop web --dev # Dev mode: swap xterm.js for an alternative rendererThe primary entry point is studyloop study, which creates a complete tmux-based study environment:
studyloop study "Python Decorators" --energy 7 # Socratic mentor session
studyloop study "Spark Internals" --mode co-study # User-driven co-study
studyloop study "topic" --timer pomodoro # Override default timer
studyloop study "topic" --agent claude --web # Explicit agent + web dashboard
studyloop study "topic" --agent codex # Explicit Codex CLI session
studyloop study "topic" --lan # LAN access with password auth (implies --web)
studyloop study "topic" --lan --password SECRET # Explicit password for LAN auth
studyloop study "topic" --agent opencode # Preview OpenCode integration
studyloop study "topic" --agent pi # Preview pi integration
studyloop study --resume # Resume conversation (-r)
studyloop study --end # End session cleanly
studyloop park "How does asyncio compare?" # Park mid-sessionRun studyloop study without a topic to open the textual picker for body double, topic directory, course vendor, or course.
What studyloop study creates:
- tmux session with agent pane (left) + Textual sidebar (right)
- AI agent launched with mode-specific persona (clean pane, no visible command)
- Persistent session directory at
~/.config/studyloop/sessions/{name}/— preserves AI conversation history (.claude/,.kiro/, etc.) - Sidebar shows timer, activity feed, counters (keyboard:
ppause,rreset,Qend session) - IPC files for dashboard viewports
- Optional web dashboard at
/sessionvia--web --webauto-opens a browser to the dashboard on startup--lanexposes the dashboard on your LAN with HTTP Basic Auth, prints usable local/LAN URLs, username, and password (implies--web). Password is auto-generated if not set via--passwordorlan_passwordin config--password SECRETsets the LAN authentication password (used with--lan)
Session lifecycle:
- Start:
studyloop study "topic"— creates tmux session, agent, sidebar - Exit: quit Claude normally (
/exit, Ctrl+C) — auto-cleans up tmux, IPC files, switches back to previous session. Session directory preserved. - Resume:
studyloop study --resume— if tmux alive, reattaches. If ended, rebuilds tmux and passes-rto the agent to continue the conversation from history. - End explicitly:
studyloop study --endor sidebarQ— same cleanup as quitting Claude
Modes:
| Mode | Flag | Timer default | Agent role |
|---|---|---|---|
| Study | (default) | Elapsed | Socratic mentor drives |
| Co-study | --mode co-study |
Pomodoro | User drives, agent available |
Low-level session commands (used internally by studyloop study):
studyloop session start -t "Decorators" -e 7 # Start session record
studyloop session status # Show current state
studyloop session end -n "Got through closures" # End with notes
studyloop session effectiveness # Win rate per persona version
studyloop session effectiveness -p abc123... # Filter by persona hashstudyloop self-test # Lightweight post-install smoke check
studyloop self-test --json # JSON output for scripts/agents
studyloop self-test --quiet # One-line summary
studyloop doctor # Full health check (Rich table)
studyloop doctor --json # JSON output (for AI agents and CI)
studyloop doctor --quiet # One-line summary
studyloop doctor --category core # Check specific category only
studyloop doctor --category voice # Check local Kokoro files, afplay, and the Kokoro server
studyloop doctor --fix # Apply safe automatic fixes
studyloop update --json # Machine-readable update info
studyloop upgrade --dry-run # Preview what would change
studyloop upgrade --component packages # Upgrade only packages
studyloop upgrade --component database # Run DB migrations only
studyloop upgrade --component agents # Update agent definitions onlyUse studyloop self-test immediately after install when you only need to
confirm the CLI imports, config can be read, the sessions database path is
usable, and the web module imports. It is deliberately lightweight: it does
not run doctor --fix, start web servers, contact external services, or write
agent/harness files.
Exit codes for studyloop self-test:
| Code | Meaning |
|---|---|
0 |
All lightweight checks pass |
1 |
One or more warnings, with no failures |
2 |
One or more checks failed |
Exit codes for studyloop doctor:
| Code | Meaning |
|---|---|
0 |
All checks pass — installation is healthy |
1 |
Warnings or failures that can be fixed — run studyloop doctor --fix |
2 |
Core failure — a fundamental component is broken (e.g. wrong Python version) |
Check categories: core (Python, packages, config), database (review DB, sessions DB), config (Obsidian vault + .obsidian/ marker, Obsidian export config, review dirs, pandoc), deps (optional packages), agents (AI tool definitions), voice (local Kokoro model files, afplay, and Kokoro-server reachability when configured), harness (session-export wiring), updates (source-install/version metadata).
Review schedule: 1 → 3 → 7 → 14 → 30 days
studyloop review shows what's due based on active learning evidence in study_progress, such as recorded concept progress and teach-back scores.
Use studyloop review --interleave adaptive --energy low|medium|high when you want review to show the active interleaving mix. Low energy keeps the mix close to current or due repair; medium and high energy allow more transfer and weak-link work.
studyloop now is the shared recommendation engine for the CLI and web API (GET /api/now). It returns one primary recommendation and up to two alternates. Each recommendation includes the concept, topic or course, reason, action type, estimated minutes, source, score, and the command to record evidence when done.
studyloop now
studyloop now --energy low --time 15
studyloop now --modality hands-on --interleave adaptive
studyloop now --json
studyloop now --speakDefault ranking is due review first, then struggling or low teach-back score, then active-course continuity, then modality match. Low energy suppresses hard context switching.
studyloop chat-note turns one markdown/text note into a compact Socratic context pack. V1 prints or speaks the mentor prompt; it does not run a separate chat backend.
studyloop chat-note ~/Obsidian/Personal/Study/Python/decorators.md
studyloop chat-note NOTE.md --mode diagram
studyloop chat-note NOTE.md --mode trace --jsonModes are recall, diagram, trace, teachback, and repair. The command validates that the note is inside configured vault/content roots, chunks by headings and code blocks, and ends with a suggested studyloop progress or studyloop teachback command.
studyloop practice verify records an attempt against a generated practice deck. Command verification only runs when --run-command is explicit; non-command tasks use notes plus expected-artifact checks as a rubric. Newer generated decks can also carry rubric items, evidence prompts, setup commands, and per-task command timeouts.
studyloop practice verify course-practice.json --task 1 --notes "diagram matched"
studyloop practice verify course-practice.json --task 2 --run-command --workdir . --timeout 120studyloop recap today compresses the day into one win, one repair target, one due item, and one next action. --speak calls study-speak, so the local-Kokoro / Kokoro-server / macOS backend configuration is inherited. --audio-file saves the same recap as a local audio file, preferring the Kokoro server when configured and falling back to macOS say.
studyloop recap today
studyloop recap today --json
studyloop recap today --speak
studyloop recap today --audio-file ~/Desktop/studyloop-recap.wavstudyloop mastery exposes concept dependencies and blockers. The Web UI's
Mastery tab uses the same data through GET /api/mastery/graph and
GET /api/mastery/weak-links; those web endpoints accept limit so broad
topics stay fast to render in Mermaid.
studyloop mastery graph --topic python
studyloop mastery graph --topic python --format json
studyloop mastery weak-links --topic pythonThe graph seeds lightweight concept edges from headings, tags, backlinks, existing concept relations, and knowledge bridges, then renders Mermaid by default.
studyloop progress has two modes:
- With no concept argument, it summarizes local course progress from
content.base_pathand the review database. - With a concept argument, it records confidence for a study concept.
studyloop progress # Course summary table
studyloop progress --course python # One course only
studyloop progress --json # JSON for scripts/agents
studyloop progress "list comprehensions" \
--topic python \
--confidence confident # Record concept confidenceThe summary includes local source count, unique review cards, due cards, mastered cards, review sessions, and review accuracy. Course filtering uses the course slug, such as python or data-engineering.
These commands surface progress without opening the Web UI. Shipped mentor definitions run the relevant checks at session start.
studyloop wins # Progress overview + recent mastered/confident concepts
studyloop wins --days 14 # Wins in the last N days (default 30)
studyloop resume # Last session source, topics, in-progress concepts, streak
studyloop streaks # Current/longest streak, weekly sessions, energy patternsstudyloop resume is not the same as studyloop study --resume: the latter reattaches a tmux/agent session; studyloop resume prints a text summary from session history.
studyloop struggles --days 30 # Topics mentioned in 3+ sessions (table)
studyloop extract-struggles --incremental --harness kiro --model MODEL_ID
studyloop extract-struggles --incremental --session-id ID --model MODEL_ID
studyloop extract-struggles --full --harness codex --model MODEL_ID
studyloop extract-struggles --dry-run --harness claude --model MODEL_IDExtraction always calls the explicitly selected live Bedrock model. You can set
STUDYLOOP_EXTRACTOR_MODEL instead of repeating --model. There is no fixture
or fallback backend: missing model access fails without writing progress rows.
Latest-session and full-history runs require --harness; an explicit session ID
does not scan unrelated coding history.
Cross-session backlog over parked topics, detected struggles, and manual entries (parked_topics store). Distinct from studyloop topics, which lists configured course topics in config.yaml.
studyloop backlog list
studyloop backlog list --tech Python --source struggled
studyloop backlog list --all # Include resolved/dismissed
studyloop backlog add "Python decorators" --tech Python --note "After Ch. 4"
studyloop backlog resolve 42
studyloop backlog suggest
studyloop backlog suggest --limit 5 --topic "Python Patterns"Mid-session parking uses studyloop park (writes to the same backlog with source parked).
Structured, evaluable plans: a mission, topics, success criteria, and milestones. Plans live as Markdown documents (studyloop plan path) with a derived index in the sessions DB, so a plan is readable and editable outside the tool.
studyloop plan interview # What to ask, plus evidence-based seed suggestions
studyloop plan new --title "SQL window functions" \
--why "Stop guessing at analytic queries" \
--topic "sql" --success "Explain PARTITION BY unprompted" \
--milestone "Frames (concepts: window frame, rows vs range)" \
--target-date 2026-10-01 --energy-floor 4
studyloop plan list --status active
studyloop plan show PLAN_ID --markdown
studyloop plan milestone PLAN_ID 0 --done
studyloop plan evaluate PLAN_ID --phase mid --record
studyloop plan status PLAN_ID complete
studyloop plan reindex # Rebuild the DB index from the documentsOmitted answers are left explicitly blank in the document rather than invented, and readiness reports what is still missing. Activation (--activate, or plan status … active) is refused while a plan lacks a mission, success criteria, or milestones — an unevaluable plan must not look active.
studyloop plan interview exists so an agent can learn what to ask before proposing a plan.
Each exercise set carries all three formats — blank slate, completion, and multiple choice. The completion exercise is sliced from the reference solution, so one authored task yields both code formats.
studyloop exercise new --topic "generators" \
--requirement "Yields lazily" --requirement "Handles empty input" \
--reference solution.py --reveal 0.4 --language python
studyloop exercise from-milestone PLAN_ID --index 0 # Seed from a plan milestone
studyloop exercise import hand-written.md # `- [x]` marks the correct option
studyloop exercise list --plan PLAN_ID
studyloop exercise show SET_ID --markdown # Learner-safe: no answers
studyloop exercise show SET_ID --with-answers # Author use only
studyloop exercise review SET_ID --kind completion --file attempt.py --record
studyloop exercise review SET_ID --kind multiple_choice --answer 0:a --answer 1:b,cshow withholds reference solutions and marked-correct choices unless --with-answers is passed, so pasting the output into a study session cannot hand over the solution. review output carries questions, never the solution; a completion attempt is scored on what the learner added, with criteria the starter code already satisfied marked given and excluded. --record writes the derived confidence to study_progress.
Multiple-choice sets are authored in plain Markdown, so a set can be written in any text editor with no tooling. These commands are CLI-only — there is no browser panel for exercises yet.
Focus shapes what studyloop now and review sessions recommend. It never deletes data on its own; studyloop prune is the age-based command that does.
studyloop focus # Show current focus (max 3 topics)
studyloop focus suggest --days 30 # Propose focus from sessions, struggles, config
studyloop focus set "python" "sql window functions"
studyloop focus set "python" --no-refocus # Save focus only, skip the data movement
studyloop focus apply --dry-run # Preview the deferred data movement
studyloop focus clear
studyloop prune --days 30 # Dry run by default
studyloop prune --days 30 --apply # Actually deleteSaving a focus is followed by the refocus data movement: focus-topic conversations newer than --days are pulled from the full DB into the local DB, then non-focus sessions older than --days are pruned. If the full DB is unreachable (external volume unmounted) the focus still saves and the movement is deferred — run studyloop focus apply later.
Both prune and the refocus prune obey one safety invariant: a session is deleted only when the configured full DB (database.full_db_path) holds the same session with a matching content hash and at least as many messages. Unverified sessions are skipped and reported. Learning data — progress, concepts, reviews — is never touched.
Record analogies between domains (e.g. networking → data engineering). Stored in the session/review database.
studyloop bridge add "ECMP" -s networking "Spark partitions" -t python -m "Both distribute load"
studyloop bridge list
studyloop bridge list --source-domain networking --target-domain pythonstudyloop clean --dry-run # Preview zombie tmux sessions, orphan dirs, stale state
studyloop clean # Apply cleanup (respects lock; safe with --resume)Also runs automatically before studyloop study when zombies are detected.
studyloop web launches a progressive web app for flashcard and quiz review. By default it binds to 127.0.0.1; use --lan to expose it on your network with HTTP Basic Auth.
studyloop web # Serve on 127.0.0.1:8567
studyloop web --port 9000 # Custom port
studyloop web --lan # LAN access with auth
studyloop web --lan --password SECRET| Key | Action | When |
|---|---|---|
Space/Enter |
Flip card | Flashcard, before reveal |
Y |
I knew it | Flashcard, after reveal |
N |
Didn't know | Flashcard, after reveal |
1-4 |
Select quiz option | Quiz mode |
S |
Skip card | During review |
T |
Read aloud (once) | During review |
R |
Retry wrong answers | After session |
Esc |
Back to home | Anywhere |
Features: Source/chapter filter, card count limiter (10/20/50/100/All), due cards badge, session history, 90-day study heatmap, Pomodoro timer (25min/5min), OpenDyslexic font toggle, dark/light theme, PWA installable.
Live session dashboard (/session): Real-time SSE activity feed, energy-adaptive timer, topic counters, and a terminal panel showing the live session. The panel renders one of two surfaces: xterm.js driven by a PTY streamed over a WebSocket (transport: "pty"), or ACP chat for structured-event agents (transport: "acp"). It is draggable (stacked or side-by-side), has a layout toggle and panel-swap buttons, and can be popped out to a separate window (pop-out auto-closes when returning inline). The panel reattaches by itself after a page refresh — it reads GET /api/session/state on init and adopts a live session it owns.
There is no browser terminal fallback: the ttyd iframe surface was retired.
Installing ttyd no longer enables anything user-visible in the dashboard. The
ttyd server transport still exists for maintainers, but a session started
that way has no browser renderer and reports an explicit unavailable state.
Voice: Server-side Kokoro. The browser posts to StudyLoop's own authenticated /api/tts/speak, which proxies to the Kokoro server named by tts.openvox_base_url; with none reachable it falls back to the OS's Web Speech voices, then to silence. Off until enabled in the header.
- Read once — speaker icon on card or
Tkey - Announcements — header speaker toggle (Pomodoro transitions and confirmations). It does not read cards automatically;
Vis unbound - Stop — header stop button interrupts playback mid-utterance
Web + terminal config (~/.config/studyloop/config.yaml):
web_port: 8567 # web dashboard port (default 8567)
ttyd_port: 7681 # ttyd server transport port (default 7681) — maintainer-only; no browser surface
browser: "" # auto-open browser: chrome, safari, firefox, brave, or empty for system default
lan_password: "" # persistent password for --lan mode (auto-generated if empty)Agent priority config (~/.config/studyloop/config.yaml):
agents:
priority: [kiro, codex, claude, opencode, pi]AI session export, search, and cross-machine sync.
session-export [--sources SOURCE ...] # Export AI sessions to SQLite
session-export [--obsidian] [--obsidian-vault PATH] [--obsidian-backfill] [--obsidian-dry-run]
session-query search-cmd QUERY # Full-text search across sessions
session-query list --since 7d # List recent sessions
session-query show SESSION_ID # Show session details
session-query context SESSION_ID # Generate context for resuming
session-query continue SESSION_ID # Continuation context for resuming work
session-query stats-cmd # Database statistics
session-query tag|note SESSION_ID # Manage session tags / notes
session-query check-size # Check DB size against thresholds
session-query profiles # Manage export profiles/templates
session-sync push|pull|sync REMOTE # Sync database across machines
session-maint vacuum|reindex|schema|archive # Database maintenance
session-maint delete --confirm # Permanently delete old sessions
session-maint find-duplicates|fts-check|compact # Integrity and rescue
session-maint sync-full|snapshot|prune # Full-DB sync, snapshot, verified prune
study-speak "text" [-b openvox|kokoro|qwen3|macos] [-v VOICE] [-s SPEED]
# Speak text aloud using local TTSsearch-cmd and stats-cmd carry the -cmd suffix because the module already imports search and stats from query_logic, and Typer derives each command name from its function name. The CLI names are literal, not typos.
The --sources values below are the complete set the CLI accepts — anything else is rejected with Invalid sources: {...}.
| Source | Tool |
|---|---|
claude |
Claude Code |
codex |
OpenAI Codex CLI |
kiro |
Kiro CLI |
opencode |
OpenCode |
pi |
pi coding agent |
By default session-export runs incrementally: it only imports sessions
that are new or whose source file changed since the last export. The summary
reports four outcomes:
| Outcome | Meaning |
|---|---|
added |
New session imported for the first time |
updated |
Existing session re-imported because its source changed |
skipped |
Already up-to-date since last export (unchanged) — not re-read |
empty |
No extractable messages (header-only, content-less, or only tool-results) |
skipped is the steady-state for a repeat run — a large skipped count is
normal and means deduplication is working, not that anything failed. empty
is tracked separately so a session with no usable text is never mistaken for
one that was simply unchanged.
Use --full to ignore change-detection and re-import every session (every
unchanged session then re-imports as updated instead of skipped).
studyloop install tools
studyloop install agents
studyloop install agents --tool codex
studyloop install agents --tool claude --tool kiro
studyloop install agents --tool opencode
studyloop install agents --tool pi
session-export
session-export --sources claude --sources codex
session-export --sources opencode --sources pi
session-export --claude-only
session-export --codex-only
session-export --kiro-only
session-export --opencode-only
session-export --pi-only
session-export --fullsession-export can also write one Markdown "session-memory" note per session
into an Obsidian vault, alongside the SQLite export. Notes carry Dataview-ready
YAML frontmatter (type: agent-memory), [[wikilink]] backlinks to matching
vault topic notes, and per-project MOC index notes. Output lands in
<vault>/AgentMemory/ (and <vault>/AgentMemory/MOC/).
Enable it per-run with --obsidian, or set obsidian.export_enabled: true in
~/.config/studyloop/config.yaml (see Configuration).
session-export --obsidian # write notes for sessions touched this run
session-export --obsidian --obsidian-vault ~/Obsidian/Personal # override vault path
session-export --obsidian --obsidian-backfill # write notes for ALL historical sessions (idempotent)
session-export --obsidian --obsidian-dry-run # preview what would be written; write nothing
session-export --no-obsidian # force-disable even if config enables it| Flag | Effect |
|---|---|
--obsidian / --no-obsidian |
Enable/disable export for this run; overrides the obsidian.export_enabled config gate. |
--obsidian-vault PATH |
Override the vault root (default: obsidian.vault_path, falling back to obsidian_base). |
--obsidian-backfill |
Export every session in the DB, not just those touched this run. Idempotent — unchanged notes are skipped. |
--obsidian-dry-run |
Print the written/skipped/MOC counts without writing any files. |
Notes are idempotent: a content hash in each note's frontmatter means re-runs
skip unchanged sessions. A normal --obsidian run only writes sessions added or
updated in that run; use --obsidian-backfill for the one-time full history import.
uv pip install agent-session-tools[semantic] # Vector embeddings search
uv pip install agent-session-tools[tokens] # Token counting