Catscope Optimizer

Run a web assembly bot that will run in the Catscope runtime over Solpipe. A local Solpipe bidder daemon sets up a gRPC tunnel to a validator. The Optimizer code base talks with the local Solpipe bidder daemon to access the target Validator (Pipeline=2Gi87bD2MNKFREReqkJgcLodkBuQEcYGKaaWz45i2Nbo). ./brain is the generic, mode-agnostic bot-orchestration layer (see its own package doc comment); each trading/monitoring mode is one of its sibling packages -- currently ./brain/marketwatchv1, ./brain/testperplatencyv1, and ./brain/xstockshealthv1. The solpipe bidder proxy described in the Catscope quick start tutorial must be running before the optimizer will work.

Build the CLI:

make

• the resuling binary will be in ./bin/optimizer

Every command below takes <parent-key>, a file path to a Solana-CLI-format keypair JSON file (the fee payer/treasury parent -- not the bidder proxy's own fee payer). They all also expect a local Solpipe bidder daemon already running at its default sockets (~/.solpipe.bidder.manage.sock, ~/.solpipe.bidder.proxy.sock) -- override via the MANAGER_SOCKET/PROXY_SOCKET/STATE_URL/TXPROC_URL environment variables if yours runs elsewhere. Set PROGRAM_SOLPIPE to the real on-chain Solpipe program id if the compiled-in default doesn't match the deployment you're pointed at. Any of these can go in a .env file in the working directory instead of being exported by hand.

Every brain mode reads its market/lending data from a local SQLite database at ~/.optimizer/prefetch.db, not live from chain -- populate (or refresh) it first:

./bin/optimizer download-arb <parent-key>

This walks every tracked DEX (Orca, Raydium AMM/CLMM/CPMM, Pumpswap, Pumpfun, Phoenix) and lending protocol (Solend, Kamino, Marginfi, Drift, Jet) via the bidder daemon's own state connection and writes the result into prefetch.db. Use --exact=orca,raydium_amm,... to scope it to specific sources, or --force to re-fetch everything even if already populated.

As its last step this command also compiles a release wasm bot image from the data it just wrote, as a smoke test -- see "Reaching each brain" below for what that needs (REPO, a wasm32-wasip2 Rust toolchain). If that compile fails, the command reports an error, but prefetch.db itself has already been populated by that point regardless.

Each brain mode needs two things beyond a populated prefetch.db:

• REPO (environment variable), pointing at a local checkout of catscope-rust-bot with a wasm32-wasip2-capable Rust toolchain (rustup target add wasm32-wasip2). Every mode compiles that repo into a release .wasm binary -- statically embedding the prefetch data above -- and uploads it to the validator pipeline before doing anything else. Clone the catscope-rust-bot repository from here.
• Optionally --working-dir (defaults to ~/.optimizer) and --state-url (defaults to the bidder daemon's own state connection).

# brain/marketwatchv1 -- read-only, never trades; live terminal dashboard
# of SOL/JUP/JTO/BONK/WIF/RAY volatility/correlation.
REPO=/path/to/catscope-rust-bot ./optimizer market-watch <parent-key>

# brain/testperplatencyv1 -- real-transaction, 100x-cycled deposit/withdraw
# latency test. Scope to one protocol with
# --protocol=solend|kamino|marginfi|native|native_lite, or leave unset
# for the original full unscoped sequence.
REPO=/path/to/catscope-rust-bot ./optimizer testperp-latency <parent-key>

# brain/xstockshealthv1 -- passive watcher: logs a health-factor board for
# Kamino xStocks obligations/reserves, sends no transactions. Serves a
# local page at --dashboard-addr (default 127.0.0.1:8091).
REPO=/path/to/catscope-rust-bot ./optimizer xstocks-health <parent-key>

Run ./bin/optimizer <command> --help for every flag a given mode accepts, and see ./brain's own package doc comment for how a new mode plugs into this same upload/handshake flow.