Skip to content

README.md

Latest commit

 

History

History
117 lines (82 loc) · 3.18 KB

README.md

File metadata and controls

117 lines (82 loc) · 3.18 KB

OpenClaw → OpenViking Memory Migration

Imports your existing OpenClaw memory files directly into OpenViking's memory system.

Zero LLM calls. Content is preserved verbatim; only embeddings are generated.

What gets migrated

OpenClaw file OpenViking category Why
MEMORY.md, memory.md entities Curated durable knowledge — projects, people, concepts
YYYY-MM-DD.md (daily logs) events Time-stamped records, decisions, milestones
YYYY-MM-DD-slug.md (session summaries) cases Problem + solution context from specific sessions
Everything else entities Safe fallback for arbitrary markdown files

Requirements

pip install openviking

A valid ~/.openviking/ov.conf (or equivalent) with an embedding model configured is required for the real run.

Usage

# Preview — no data written
python migrate.py --dry-run

# Migrate with defaults
#   OpenClaw dir : ~/.openclaw/workspace
#   OV data dir  : ./data
#   identity     : account=default, user=default, agent=default
python migrate.py

# Custom paths and identity
python migrate.py \
  --openclaw-dir ~/myworkspace \
  --ov-data-dir  ./ov-data \
  --account-id   myaccount \
  --user-id      myuser \
  --agent-id     myagent

# Force all files into a single category
python migrate.py --category events

Options

Flag Default Description
--openclaw-dir PATH ~/.openclaw/workspace Path to OpenClaw workspace
--ov-data-dir PATH ./data OpenViking data directory
--account-id TEXT default Account ID (alphanumeric/underscore/hyphen)
--user-id TEXT default User ID
--agent-id TEXT default Agent ID
--category TEXT (auto) Override category for all files
--dry-run off Preview without writing

Dry-run example

OpenClaw → OpenViking Migration (DRY RUN)
Found 3 file(s) in /Users/alice/.openclaw/workspace

  MEMORY.md                    →  entities      (3,421 chars)
  memory/2026-03-15.md         →  events        (1,832 chars)
  memory/2026-03-15-bug-fix.md →  cases         (942 chars)

Would import: 3 file(s) | 0 LLM calls | ~3 embedding job(s) queued | 6,195 chars total
Run without --dry-run to proceed.

How it works

For each file the script:

  1. Reads the Markdown content.
  2. Classifies it into an OV memory category (or uses --category).
  3. Builds a one-line abstract (first non-empty line, ≤ 100 chars).
  4. Calls MemoryExtractor.create_memory() — writes the file to VikingFS.
  5. Calls SessionCompressor._index_memory() — enqueues an embedding job.

Memories appear in OV's memory-specific retrieval (viking://user/<user>/memories/) as soon as the embedding worker processes the queue.

Running tests

PYTHONPATH=. .venv/bin/python -m pytest tests/unit/test_openclaw_migration.py -v --no-cov

Verifying the import

After a real run, confirm memories were written:

import openviking as ov

client = ov.OpenViking(path="./data")
client.initialize()
results = client.find("viking://user/default/memories/entities/")
print(results)