pin cocoindex>=1.0.7 (fixes #361) + bump to 0.6.7

[HumanBean17]

Summary

Fixes #361 — TypeError: mount_table_target() got an unexpected keyword argument 'num_transactions_before_optimize' during init / increment.

Root cause

PR #309 added num_transactions_before_optimize=_NUM_TXN_BEFORE_OPTIMIZE to the three lancedb.mount_table_target(...) calls in java_index_flow_lancedb.py to disable cocoindex's in-flow background table.optimize() (which races concurrent table.delete() → LanceDB commit conflict, upstream lancedb#1504). That keyword was introduced by cocoindex in 1.0.7, but pyproject.toml still pinned cocoindex[lancedb]>=1.0.0a43,<2. Users resolving an older cocoindex (1.0.0–1.0.6, or a stale copy in a shared env — e.g. the reporter's py-envs/common py3.13) hit the TypeError at flow-build time.

Fix

Bump the cocoindex floor to 1.0.7 — the version that introduced the keyword (verified via inspect.signature on the locally installed 1.0.7). lance_optimize.py (connect_async / table.optimize()) is 1.0.7-era too, so 1.0.7 is the correct floor.

Workaround for affected users (no upgrade needed on their side)

pip install --upgrade "cocoindex[lancedb]>=1.0.7"

Changes

• pyproject.toml: cocoindex[lancedb]>=1.0.0a43,<2 → >=1.0.7,<2
• pyproject.toml: version 0.6.6 → 0.6.7

User-visible behaviour changes

• Fresh pip install java-codebase-rag now pulls cocoindex ≥ 1.0.7, so the init/increment TypeError no longer reproduces.
• New PyPI release 0.6.7 (published from this branch).

Validation

• .venv/bin/ruff check . — passes.
• Confirmed cocoindex 1.0.7's mount_table_target signature includes num_transactions_before_optimize: int = 50.

🤖 Generated with Claude Code