feat: add FalkorDB bulk-loader export format

[shahar-biron]

Summary

• add a new falkordb output format to CLI/config/exporters
• emit FalkorDB bulk-loader schema headers and row layouts for accounts, transactions, and fraud edges
• include is_fraud markers on transaction exports (false for normal, true for injected fraud)
• expand tests and documentation for FalkorDB bulk-loader ingestion

Validation

• python3 -m ruff check src/gen_fraud_graph/cli.py src/gen_fraud_graph/config.py src/gen_fraud_graph/exporters.py src/gen_fraud_graph/generator.py src/gen_fraud_graph/typologies.py tests/test_cli.py tests/test_exporters.py tests/test_generator.py tests/test_typologies.py
• PYTHONPATH=src python3 -m pytest --no-cov tests/test_exporters.py tests/test_generator.py tests/test_typologies.py tests/test_cli.py
• PYTHONPATH=src python3 -m pytest --no-cov tests

Artifacts

• Conversation: https://app.warp.dev/conversation/369146ed-df6b-47b5-8d6a-12f6807fd109
• Plan: https://app.warp.dev/drive/notebook/HgQHdcCbXvH2uJdworqGiz

Co-Authored-By: Oz oz-agent@warp.dev

[github-actions]

All contributors have signed the CLA ✍️ ✅
_{Posted by the CLA Assistant Lite bot.}

[shahar-biron]

I have read the CLA Document and I hereby sign the CLA

[shahar-biron]

recheck

[opensource-SantanderAI]

Thanks @shahar-biron — the FalkorDB bulk-loader export is a useful addition and the tests/docs are appreciated.

Two things are needed before we can merge:

1. Rebase and resolve a merge conflict in typologies.py

We just merged #16 (the evaluation harness + --hardness knob), which changed the same fraud-transaction row-building code you touched, so your branch now conflicts. cli.py, config.py and generator.py merge cleanly; the only real conflict is one block in src/gen_fraud_graph/typologies.py.

It's a semantic conflict, not a textual one, so we'd rather you resolve it (you know your FalkorDB layout best). For guidance: #16 replaced the static self.amount with self._jittered_amount() (so the hardness knob can spread fraud amounts). Please keep your per-format row layout and use the jittered amount, e.g.:

if fmt == "neptune":
    row = [tx_id, src, dst, "TRANSFER", self._jittered_amount(), timestamp, desc]
elif fmt == "falkordb":
    row = [src, dst, tx_id, self._jittered_amount(), timestamp, desc]
else:
    row = [tx_id, src, dst, self._jittered_amount(), timestamp, desc]

Please also confirm your falkordb export behaves correctly now that fraud amounts are no longer a fixed sentinel (i.e. at --hardness medium/high). After rebasing, re-run the suite locally (ruff check, black --check, pytest) and push so CI is green again.

2. Remove the tool/vendor attribution

Please drop the app.warp.dev links (conversation + notebook) from the PR description and the Co-Authored-By: Oz <oz-agent@warp.dev> trailer from the commit message. We don't keep third-party tool links or agent co-author trailers in the history — authorship should be yours.

Once it's rebased, conflict-free, CI-green and the attribution is cleaned up, ping us and we'll re-review and merge. Thanks again!