ocas-weave（奥卡斯编织）

Weave

Weave maintains a private, provenance-backed social graph of people, relationships, preferences, and shared experiences — queryable for meeting prep, gift ideas, hosting, introductions, city connections, and serendipity discovery. Every stored fact carries source type, reference, timestamp, and confidence score; the graph never silently merges two person records and never writes back to external systems without explicit per-sync approval. All queries use Cypher — no SQL. The database initializes automatically on first use.

When to use

• Store or update information about a person, relationship, or preference
• Look up who someone is, how they relate to others, or what they like
• Prepare for a meeting, dinner, or introduction
• Find who you know in a given city
• Generate gift ideas grounded in known preferences
• Discover serendipity connections between people
• Sync contacts from Google Contacts or Clay

When not to use

• Web research without a social graph need — use Sift
• OSINT investigations on people — use Scout
• CRM or sales pipeline automation
• Personality profiling without evidence

Responsibility boundary

Weave owns the private social graph: people, relationships, preferences, and shared experiences.

Weave does not own: general world knowledge (Elephas/Chronicle), OSINT research (Scout), web research (Sift), task management (Triage).

Weave is a standalone database. It does not write to Chronicle and has no runtime dependency on Chronicle. If a person in Weave also exists in Chronicle, Chronicle may store a weave:person_id reference on its Entity node. That is Chronicle's concern, not Weave's.

Storage layout

~/openclaw/db/ocas-weave/
  weave.lbug          — LadybugDB database (auto-created on first use)
  config.json         — connector and sync configuration
  staging/            — temporary import/export files

~/openclaw/journals/ocas-weave/
  YYYY-MM-DD/
    {run_id}.json     — one journal per run

Default config.json:

{
  "skill_id": "ocas-weave",
  "skill_version": "2.3.0",
  "config_version": "1",
  "created_at": "",
  "updated_at": "",
  "writeback": {
    "google_contacts": false,
    "clay": false
  },
  "last_sync": {
    "google_contacts": null,
    "clay": null
  },
  "retention": {
    "days": 0
  }
}

Database rules

LadybugDB is an embedded single-file database. One READ_WRITE process at a time. If another process holds the lock, operations fail immediately with a lock error — do not retry silently, surface the error.

Multiple READ_ONLY connections are safe simultaneously. COPY FROM is for bulk import (>100 rows). MERGE is for sporadic single-record upserts. Never loop MERGE over bulk data.

Auto-initialization

Every command that opens the database runs _ensure_init() first. No manual init command is needed on first use.

Read references/init_pattern.md for the _open_db implementation pattern. Full DDL is in references/schemas.md.

Commands

weave.upsert.person -- Add or update a person. Auto-inits DB on first call. MERGE on id. Read back after write; report failure if no row returned — never claim success unconfirmed.

weave.upsert.relationship -- Add or update a Knows edge. Confirm both Person nodes exist first. Halt and report which is missing.

weave.upsert.preference -- Store a provenance-backed preference. Each preference is a distinct CREATE (not merged). Link to Person via HasPreference edge.

weave.import.csv -- Bulk import contacts via COPY FROM. Read references/import_export.md. Pre-process CSV to staging dir first. Check CALL show_warnings() RETURN * after. Report: N imported, N skipped (with reasons), N failed.

weave.query -- Query the graph. Read references/query_patterns.md. Modes: lookup, connection, serendipity, city, summarize, gift. Return only stored facts with provenance. Never speculate.

weave.attach -- Query an external skill database read-only. Read references/cross_db.md.

weave.export -- Export data to staging dir via COPY TO. Read references/import_export.md.

weave.sync.google-contacts -- Bidirectional sync with Google Contacts. Read references/connectors.md. Inbound before outbound. Outbound requires writeback.google_contacts: true AND explicit per-sync approval.

weave.sync.clay -- Bidirectional sync with Clay. Read references/connectors.md. Clay is enrichment source — Weave provenance wins conflicts. Outbound requires writeback.clay: true AND explicit approval.

weave.project.vcard -- Generate vCard 4.0 draft. Read references/vcard_projection.md. Omit fields with confidence below 0.7. Label DRAFT. Requires explicit approval before writeback.

weave.writeback.contacts -- Push records to Google Contacts or Clay. Disabled by default. Requires config enablement AND per-action user approval.

weave.init -- Diagnostic and repair command. Checks schema, creates missing tables, verifies indexes. Use when troubleshooting, not as a prerequisite — the database initializes automatically on first use.

weave.status -- Report graph health and config state.

CALL show_tables() RETURN *;
MATCH (p:Person) RETURN count(p) AS people;
MATCH ()-[r:Knows]->() RETURN count(r) AS relationships;
MATCH (pref:Preference) RETURN count(pref) AS preferences;
CALL show_warnings() RETURN *;

weave.journal -- Write journal for the current run. Read references/journal.md. Called at end of every run. Journals are immutable after write.

weave.update -- Pull latest skill package from GitHub source. Preserves journals and data.

Run completion

After every Weave command that reads or writes data:

1. Persist any new or updated records to the database
2. Log material decisions to decisions.jsonl
3. Write journal via weave.journal — Observation Journal for queries/upserts/imports, Action Journal for syncs/writebacks

Provenance

Every written fact requires: source_type (direct / inferred / imported / user-stated), source_ref, record_time (ISO 8601), confidence (0.0–1.0). Use event_time when the real-world occurrence has a distinct time. Never write facts without provenance.

Constraints

• Never use SQL.
• Never report a write as successful before read-back confirms it.
• Never parse or modify .lbug, .wal, .shadow, or .tmp files directly.
• Never write to Chronicle or any other skill's database.
• Never silently collapse two Person records into one.
• Use ontology standard relationship types in Knows.rel_type.
• Store useful, durable, socially actionable facts only.
• No outbound sync without explicit per-sync user approval.
• Surface lock errors immediately.
• Write a journal at the end of every run. Runs missing journals are invalid.

OKRs

Universal OKRs from spec-ocas-journal.md apply. Weave-specific:

skill_okrs:
  - name: person_record_completeness
    metric: fraction of Person nodes with name + (email or phone) + record_time
    direction: maximize
    target: 0.80
    evaluation_window: 30_runs
  - name: sync_success_rate
    metric: fraction of sync runs with zero failed records
    direction: maximize
    target: 0.90
    evaluation_window: 30_runs
  - name: import_skip_rate
    metric: fraction of imported rows skipped due to missing required fields
    direction: minimize
    target: 0.05
    evaluation_window: 30_runs
  - name: query_provenance_coverage
    metric: fraction of returned facts carrying source_ref and record_time
    direction: maximize
    target: 1.0
    evaluation_window: 30_runs

Optional skill cooperation

• Elephas — read Chronicle read-only for entity enrichment (optional, degrades gracefully if absent)
• Scout — receive OSINT findings about people as upsert candidates
• Dispatch — provide social graph context for communication drafting

Journal outputs

• Observation Journal — query runs, upsert runs, import runs
• Action Journal — sync runs, writeback runs

Initialization

On first invocation of any Weave command, _open_db() handles auto-initialization:

1. Create ~/openclaw/db/ocas-weave/ and subdirectories (staging/)
2. Write default config.json with ConfigBase fields if absent
3. Create ~/openclaw/journals/ocas-weave/
4. Open database (auto-creates weave.lbug and runs DDL if tables absent)
5. Register cron job weave:update if not already present (check openclaw cron list first)
6. Log initialization as a DecisionRecord

Background tasks

openclaw cron add --name weave:update --schedule "0 0 * * *" --command "weave.update" --sessionTarget isolated --lightContext true --timezone America/Los_Angeles

Self-update

weave.update pulls the latest package from the source: URL in this file's frontmatter. Runs silently — no output unless the version changed or an error occurred.

1. Read source: from frontmatter → extract {owner}/{repo} from URL
2. Read local version from skill.json
3. Fetch remote version: gh api "repos/{owner}/{repo}/contents/skill.json" --jq '.content' | base64 -d | python3 -c "import sys,json;print(json.load(sys.stdin)['version'])"
4. If remote version equals local version → stop silently
5. Download and install:

   TMPDIR=$(mktemp -d)
   gh api "repos/{owner}/{repo}/tarball/main" > "$TMPDIR/archive.tar.gz"
   mkdir "$TMPDIR/extracted"
   tar xzf "$TMPDIR/archive.tar.gz" -C "$TMPDIR/extracted" --strip-components=1
   cp -R "$TMPDIR/extracted/"* ./
   rm -rf "$TMPDIR"

1. On failure → retry once. If second attempt fails, report the error and stop.
2. Output exactly: I updated Weave from version {old} to {new}

Visibility

public

Support file map