mirror of
https://github.com/luxfi/precompile.git
synced 2026-09-24 00:26:21 +00:00
86 lines
2.6 KiB
Go
86 lines
2.6 KiB
Go
// Copyright (C) 2026, Lux Industries Inc. All rights reserved.
|
|
// SPDX-License-Identifier: Apache-2.0
|
|
//
|
|
// blue — talk to Beluga's on-chain brain.
|
|
//
|
|
// This invokes the SAME deterministic AI-inference precompile (0x0300…0003) that a
|
|
// transaction to the Beluga chain executes — in-process, so you get the exact bytes
|
|
// every validator would compute. The current brain is the MVP proof-of-correctness
|
|
// model (untrained, 48-token vocab): it "speaks" in deterministic token ids, not English.
|
|
// A trained model (e.g. zen-nano, int-quantized) plugs in via the model-registry
|
|
// precompile (0x0300…0002) to make blue actually converse.
|
|
//
|
|
// Usage:
|
|
//
|
|
// go run ./cmd/blue/ # blue continues the default prompt
|
|
// go run ./cmd/blue/ 1 7 13 2 # send a token-id prompt (ids 0..47)
|
|
// go run ./cmd/blue/ -n 16 5 9 2 # ask for 16 new tokens
|
|
package main
|
|
|
|
import (
|
|
"encoding/binary"
|
|
"fmt"
|
|
"os"
|
|
"strconv"
|
|
|
|
"github.com/luxfi/geth/common"
|
|
"github.com/luxfi/precompile/inference"
|
|
)
|
|
|
|
func main() {
|
|
nNew := uint32(10)
|
|
var prompt []uint32
|
|
for i := 1; i < len(os.Args); i++ {
|
|
a := os.Args[i]
|
|
if (a == "-n" || a == "--new") && i+1 < len(os.Args) {
|
|
v, err := strconv.Atoi(os.Args[i+1])
|
|
if err != nil || v < 1 {
|
|
fail("bad -n value %q", os.Args[i+1])
|
|
}
|
|
nNew = uint32(v)
|
|
i++
|
|
continue
|
|
}
|
|
v, err := strconv.Atoi(a)
|
|
if err != nil || v < 0 || v > 47 {
|
|
fail("bad token %q — use integer token ids 0..47 (blue's MVP vocab is 48)", a)
|
|
}
|
|
prompt = append(prompt, uint32(v))
|
|
}
|
|
if len(prompt) == 0 {
|
|
prompt = []uint32{1, 7, 13, 2}
|
|
}
|
|
|
|
// Encode the precompile calldata exactly as an on-chain caller would:
|
|
// selector(4) | nNew(uint32 BE, 4) | prompt tokens (uint32 BE each)
|
|
input := make([]byte, 8+len(prompt)*4)
|
|
binary.BigEndian.PutUint32(input[0:4], inference.SelectorGenerate)
|
|
binary.BigEndian.PutUint32(input[4:8], nNew)
|
|
for i, t := range prompt {
|
|
binary.BigEndian.PutUint32(input[8+i*4:], t)
|
|
}
|
|
|
|
gas := inference.InferencePrecompile.RequiredGas(input)
|
|
ret, _, err := inference.InferencePrecompile.Run(nil, common.Address{}, inference.ContractAddress, input, gas, false)
|
|
if err != nil {
|
|
fail("blue: %v", err)
|
|
}
|
|
|
|
out := make([]uint32, len(ret)/4)
|
|
for i := range out {
|
|
out[i] = binary.BigEndian.Uint32(ret[i*4:])
|
|
}
|
|
said := out[len(prompt):]
|
|
|
|
fmt.Printf(" you → %v\n", prompt)
|
|
fmt.Printf(" blue → %v\n", said)
|
|
fmt.Printf("\n (full: %v)\n", out)
|
|
fmt.Printf(" precompile %s · gas %d · deterministic: every validator computes this exact output\n",
|
|
inference.ContractAddress.Hex(), gas)
|
|
}
|
|
|
|
func fail(format string, a ...any) {
|
|
fmt.Fprintf(os.Stderr, format+"\n", a...)
|
|
os.Exit(1)
|
|
}
|