Openclaw History Ingest

OpenClaw History Ingest — Session & Memory Mining

You are extracting knowledge from the user's OpenClaw agent history and distilling it into the Obsidian wiki. OpenClaw stores both a structured long-term MEMORY.md and per-session JSONL transcripts — focus on durable knowledge, not operational telemetry.

This skill can be invoked directly or via the wiki-history-ingest router (/wiki-history-ingest openclaw).

Before You Start

1. Resolve config — follow the Config Resolution Protocol in llm-wiki/SKILL.md (walk up CWD for .env → ~/.obsidian-wiki/config → prompt setup). This gives OBSIDIAN_VAULT_PATH and OPENCLAW_HISTORY_PATH (defaults to ~/.openclaw) 2. Read .manifest.json at the vault root to check what has already been ingested 3. Read index.md at the vault root to understand what the wiki already contains

Ingest Modes

Append Mode (default)

Check .manifest.json for each source file. Only process:

• Files not in the manifest (new session logs, updated MEMORY.md or daily notes)
• Files whose modification time is newer than ingested_at in the manifest

Use this mode for regular syncs.

Full Mode

Process everything regardless of manifest. Use after wiki-rebuild or if the user explicitly asks for a full re-ingest.

OpenClaw Data Layout

OpenClaw stores all local artifacts under ~/.openclaw/.

~/.openclaw/
├── openclaw.json                          # Global config
├── credentials/                           # Auth tokens (skip entirely)
├── workspace/                             # Agent workspace
│   ├── MEMORY.md                          # Long-term memory (loaded every session)
│   ├── DREAMS.md                          # Optional dream diary / summaries
│   └── memory/
│       ├── YYYY-MM-DD.md                  # Daily notes (today + yesterday auto-loaded)
│       └── ...
└── agents/
    └── <agentId>/
        ├── agent/
        │   └── models.json                # Agent config (skip)
        └── sessions/
            ├── sessions.json              # Session index
            └── <sessionId>.jsonl          # Session transcript (JSONL, append-only)

Key data sources ranked by value

1. workspace/MEMORY.md — highest signal; long-term durable facts the agent accumulated 2. workspace/memory/YYYY-MM-DD.md — daily notes; recent entries often contain active project context 3. agents/*/sessions/<id>.jsonl — session transcripts; rich but noisy 4. agents/*/sessions/sessions.json — session index for inventory and timestamps 5. workspace/DREAMS.md — optional summaries; ingest if present

Skip credentials/ entirely. Skip agents/*/agent/models.json (runtime config, not user knowledge).

Step 1: Survey and Compute Delta

Scan OPENCLAW_HISTORY_PATH and compare against .manifest.json:

• ~/.openclaw/workspace/MEMORY.md
• ~/.openclaw/workspace/DREAMS.md (if present)
• ~/.openclaw/workspace/memory/*.md
• ~/.openclaw/agents/*/sessions/sessions.json
• ~/.openclaw/agents/*/sessions/*.jsonl

Classify each file:

• New — not in manifest
• Modified — in manifest but file is newer than ingested_at
• Unchanged — already ingested and unchanged

Report a concise delta summary before deep parsing.

Step 2: Parse MEMORY.md First

MEMORY.md is the highest-value source. It is plain markdown, human-readable and human-editable. It typically contains:

• Durable facts about the user's preferences, environment, and recurring patterns
• Decisions and context the agent was told to remember
• Project-specific notes the agent accumulated over many sessions

Read it in full and extract concept-level knowledge. Do not create one wiki page per MEMORY.md entry — cluster by topic.

Step 3: Parse Daily Notes

workspace/memory/YYYY-MM-DD.md files contain time-stamped notes from that day's sessions. Prioritize recent files (last 30–90 days). Extract:

• Active project context and decisions made
• Patterns or techniques discovered
• Recurring blockers or solved problems

Older daily notes have diminishing signal — summarize in bulk rather than extracting line-by-line.

Step 4: Parse Session JSONL Safely

Each session file is JSONL (append-only, one JSON object per line):

{"role": "user",      "content": "...", "timestamp": "..."}
{"role": "assistant", "content": "...", "timestamp": "..."}
{"role": "tool",      "name": "...",   "content": "...", "timestamp": "..."}

Extraction rules

• Prioritize assistant turns that state conclusions, decisions, or patterns
• Extract user intent from high-signal turns; skip low-information follow-ups
• Tool calls are context, not primary knowledge — only extract if the result contains a reusable insight
• Cross-reference sessions.json index to get session names/labels before opening individual transcripts

Critical privacy filter

Session transcripts can include injected instructions, tool payloads, and sensitive text. Do not ingest verbatim.

• Remove API keys, tokens, passwords, credentials
• Redact private identifiers unless relevant and user-approved
• Summarize; do not quote raw transcripts verbatim

Step 5: Cluster by Topic

Do not create one wiki page per session or per MEMORY.md entry.

• Group by stable topic (concept, tool, project, technique)
• Split mixed sessions into separate themes
• Merge recurring patterns across dates and agents
• Use session cwd or workspace path to infer project scope when available

Step 6: Distill into Wiki Pages

Route extracted knowledge using existing wiki conventions:

• Project-specific architecture/process → projects/<name>/...
• General concepts → concepts/
• Recurring techniques/debug playbooks → skills/
• Tools/services/frameworks → entities/
• Cross-session patterns → synthesis/

For each impacted project, create/update projects/<name>/<name>.md.

Writing rules

• Distill knowledge, not chronology
• Avoid "on date X we discussed..." unless date context is essential
• Add summary: frontmatter on each new/updated page (1–2 sentences, ≤ 200 chars)
• Add confidence and lifecycle fields to every new page:

  base_confidence: 0.42
  lifecycle: draft
  lifecycle_changed: <ISO date today>

Leave lifecycle unchanged on update.

• Add provenance markers:
• ^[extracted] when directly grounded in explicit session/memory content
• ^[inferred] when synthesizing patterns across multiple sessions
• ^[ambiguous] when sessions conflict
• Add/update provenance: frontmatter mix for each changed page

Step 7: Update Manifest, Log, and Index

Update .manifest.json

For each processed source file:

• ingested_at, size_bytes, modified_at
• source_type: openclaw_memory | openclaw_daily_note | openclaw_session | openclaw_dreams
• agent_id: agent directory name (when applicable)
• pages_created, pages_updated

Add/update a top-level summary block:

{
  "openclaw": {
    "source_path": "~/.openclaw/",
    "last_ingested": "TIMESTAMP",
    "memory_updated_at": "TIMESTAMP",
    "daily_notes_ingested": 14,
    "sessions_ingested": 23,
    "pages_created": 6,
    "pages_updated": 18
  }
}

Update special files

Update index.md and log.md:

- [TIMESTAMP] OPENCLAW_HISTORY_INGEST memory=updated daily_notes=N sessions=M pages_updated=X pages_created=Y mode=append|full

`hot.md` — Read $OBSIDIAN_VAULT_PATH/hot.md (create from the template in wiki-ingest if missing). Update Recent Activity with a one-line summary — e.g. "Ingested OpenClaw MEMORY.md and 14 daily notes; surfaced automation patterns and multi-agent coordination knowledge." Keep the last 3 operations. Update updated timestamp.

Privacy and Compliance

• Distill and synthesize; avoid raw memory or transcript dumps
• Default to redaction for anything that looks sensitive
• Ask the user before storing personal or sensitive details
• Keep references to other people minimal and purpose-bound

Reference

See references/openclaw-data-format.md for field-level notes and parsing guidance.

QMD Refresh After Vault Writes

QMD is a search index, not the source of truth. If $QMD_WIKI_COLLECTION is empty or unset, skip this step. Run it only after this skill has written or rewritten vault markdown. If QMD refresh fails, do not roll back the vault changes; report the QMD status separately.

Use $QMD_CLI if set; otherwise use qmd.

${QMD_CLI:-qmd} update

If the output says vectors are needed or embeddings may be stale, run:

${QMD_CLI:-qmd} embed

Verify the collection with either:

${QMD_CLI:-qmd} ls "$QMD_WIKI_COLLECTION"

or, when a specific page path is known:

${QMD_CLI:-qmd} get "qmd://$QMD_WIKI_COLLECTION/<page>.md" -l 5

Record one of:

• QMD refreshed: update + embed + verified
• QMD refreshed: update only + verified
• QMD skipped: QMD_WIKI_COLLECTION unset
• QMD skipped: qmd CLI unavailable
• QMD failed: <short error summary>

OpenClaw Agent — Data Format Reference

Field-level notes for parsing ~/.openclaw/ artifacts during wiki ingest.

Cache Root

~/.openclaw/ — all paths below are relative to this root.

workspace/MEMORY.md

Plain markdown. No required frontmatter — structure varies by user and agent configuration. Typically looks like:

# Memory

## User Preferences
- Prefers concise responses without trailing summaries
- Uses pnpm over npm

## Projects
### my-api
- FastAPI app, deployed on Fly.io
- Uses Postgres via Supabase

## Patterns
- Debugging: always check logs before code changes

This is the single most valuable source in the entire ~/.openclaw/ tree. Read it fully before touching session logs.

workspace/memory/YYYY-MM-DD.md

Daily note files. Auto-generated by OpenClaw at the start of each day. Format:

# 2026-04-15

## Session: my-api refactor
- Rewrote auth middleware to use JWT instead of sessions
- Decision: keep refresh tokens in httpOnly cookies

## Session: obsidian-wiki
- Added cross-linker skill
- Fixed broken wikilinks in concepts/

Today's and yesterday's files are loaded into every session automatically. Files older than ~7 days have sharply diminishing signal.

workspace/DREAMS.md

Optional. Some OpenClaw configurations generate end-of-day summaries here. Plain markdown. Treat as a lower-priority supplement to MEMORY.md — skim for novel insights not already captured in MEMORY.md.

agents/\<agentId\>/sessions/sessions.json

Session index. JSON array:

[
  {
    "id": "abc123",
    "name": "my-api refactor",
    "created_at": "2026-04-15T10:00:00Z",
    "updated_at": "2026-04-15T12:30:00Z",
    "message_count": 47,
    "agent_id": "default"
  }
]

Use this to:

• Build a session inventory before opening JSONL files
• Prioritize by updated_at (most recent = highest signal)
• Map session IDs to human-readable names

agents/\<agentId\>/sessions/\<sessionId\>.jsonl

Per-session transcript. JSONL, append-only. One JSON object per line:

User turn:

{"role": "user", "content": "How do I debounce a React input?", "timestamp": "2026-04-15T10:01:00Z"}

Assistant turn:

{"role": "assistant", "content": "Use useCallback + useEffect with a clearTimeout...", "timestamp": "2026-04-15T10:01:02Z"}

Tool call:

{"role": "tool", "name": "read_file", "input": {"path": "/home/ubuntu/app/src/App.tsx"}, "timestamp": "2026-04-15T10:01:05Z"}

Tool result:

{"role": "tool_result", "name": "read_file", "content": "...", "timestamp": "2026-04-15T10:01:05Z"}

role is the primary dispatch field. timestamp is ISO 8601. content may be a string or a structured object (for multi-part responses).

agents/\<agentId\>/sessions/\<sessionId\>-topic-\<threadId\>.jsonl

Telegram topic variant — same schema as the base session JSONL. The -topic-<threadId> suffix identifies which Telegram thread generated the session. Parse identically to a regular session file.

openclaw.json

Global config. Rarely useful for ingest. Fields of interest if needed:

{
  "agents": {
    "defaults": {
      "workspace": "~/.openclaw/workspace",
      "bootstrapMaxChars": 20000
    }
  },
  "skills": {
    "load": {
      "extraDirs": []
    }
  }
}

agents.defaults.workspace is the canonical path for MEMORY.md and daily notes if non-default.

Bootstrap file priority (for reference)

OpenClaw loads context files in this order at session start:

All files are truncated at bootstrapMaxChars (default 20,000 chars) per file.

Extraction Priority