Skip to content

Release 0.3.2 — seamless local Neo4j spin-up - #3

Merged
cklam12345 merged 6 commits into
mainfrom
release/0.3.2
Jul 11, 2026
Merged

Release 0.3.2 — seamless local Neo4j spin-up#3
cklam12345 merged 6 commits into
mainfrom
release/0.3.2

Conversation

@cklam12345

Copy link
Copy Markdown
Owner

Merges the release line into main (main was 5 commits behind: 0.3.0 → 0.3.1 → 0.3.2).

0.3.2 — seamless local Neo4j spin-up

Addresses QA feedback that bringing up the local Neo4j instance was not seamless. Four fixes:

  • Real readiness gatewisdom local up now waits for an authenticated Cypher RETURN 1 (_wait_until_ready), not just an open Bolt TCP port. Neo4j opens 7687 seconds before it accepts queries, so quickstart's follow-up doctor check used to fail right after printing "ready".
  • MCP auto-start — on a connection failure the MCP server attempts local.up() once when Docker is available, with stdout redirected to stderr so it never corrupts the JSON-RPC stdio stream. MCP-first users no longer hit a dead end.
  • Streaming pull — first-run docker pull streams layer progress instead of hiding it behind a silent captured call.
  • Password-drift protection — an existing container's baked-in password is reused instead of overwriting the saved one with a --password that can't take effect.

Testing

  • 134 unit tests pass (127 + 7 new).
  • Verified live end-to-end against real Docker/Neo4j:
    • Cold spin-up (~62s incl. pull); doctor/status connect immediately after "ready" — old race gone.
    • MCP _get_driver() auto-starts a stopped backend and returns a working driver.
    • Streaming pull progress; idempotent re-run; password-drift warns and keeps existing password.

🤖 Generated with Claude Code

cklam12345 and others added 6 commits May 24, 2026 22:14
Addresses QA feedback that spinning up the local Neo4j instance was not
seamless. Four fixes:

- Readiness: `wisdom local up` now waits for an authenticated Cypher
  `RETURN 1` (_wait_until_ready), not just an open Bolt TCP port. Neo4j
  opens 7687 seconds before it accepts queries, so quickstart's follow-up
  doctor check used to fail right after printing "ready".
- MCP auto-start: on a connection failure the MCP server now attempts
  `local.up()` once when Docker is available, with stdout redirected to
  stderr so it never corrupts the JSON-RPC stdio stream.
- Cold-start pull now streams docker progress instead of hiding it behind
  a silent captured call.
- Password drift: reuse an existing container's baked-in password instead
  of overwriting the saved one with a new --password that can't take effect.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
@cklam12345
cklam12345 merged commit dc084b3 into main Jul 11, 2026
10 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant