[Solana] Serum DEX Connector Progress & Follow-Up Thread

This post aims to inform about the current state of development on the Serum DEX Connector and what our current plans are.

Past Achievements

Pull request #5441 has been successfully merged, even though it did not deliver on what we hoped to achieve. Its major contribution lies in providing new interfaces for central limit order book (CLOB) DEXes through Gateway v2 and other, non-EVM-like chains.

All CLOB endpoints are implemented for Serum DEX and preliminary support for the amm_arb strategy has been established. Support for Solana wallets on the client-side has been fixed and is now fully functional.

Difficulties & Lections Learned

A major roadblock we had to resolve were randomly failing tests, for which we finally found a solution by assigning more memory to the test processes. In retrospect this seems trivial, but has been made difficult to detect by the size of the PR, the rapidly evolving development branch and frequent merges, which were not always conflict-free due to the necessary interface additions and refactorings.

The Hummingbot Foundation recommended to split up the full Solana/Serum DEX integration into multiple phases, as we are expecting more difficulties on the client’s base level. We agreed that smaller PRs will be easier to review and test, therefore a multi-phase approach in order to deliver value to the Hummingbot codebase through every merged PR.

Future Course

We, the developers of the Serum DEX Connector, are interested in launching a working MM strategy until early October. Our proposal, in accordance with the multi-stage approach, is to deliver a simple, Serum-specific strategy similar to PMM. This strategy will be launched live in October and be the ultimate test of reliability for the Serum Connector. We want to contribute in a “Phase 2” of the project this Serum MM strategy + extensive testing and an additional report like this, at the end of Phase 2.

With a potential customer in mind, we do not aim to extract more HBOTs for this work, but instead ask for continued technical support and help in achieving this daunting goal. There is roughly one month left to complete Phase 2 and we took action to ensure that enough manpower is available for this project. We are immensely grateful for in-depth reviews on our PRs and hope that this feedback cycle remains swift throughout Phase 2.

For Phase 3 we aim for full-support of all current, ExchangeBase strategies like PMM, Avellaneda-Stoikov, etc. For this, we will create an according HIP for more HBOT, as it may require some in-depth refactorings from which future Connector developers will profit, if they develop on DEX CLOBs or similar decentralized exchanges.

Special thanks go in no particular order to Fede, Faouzi, Rapcmia, Vic-en, James & Abel from Coinalpha as well as Mike Feng for the contributed efforts.

1 Like