Work Generation¶

Every Kakitu block requires a valid Proof-of-Work (PoW) value before it can be broadcast. Work is a spam-prevention mechanism — it doesn't affect consensus or add fees, but it requires computational effort to produce.

Overview¶

Work generation uses the Blake2b hash function. A valid work value satisfies:

blake2b(nonce || prev_block_hash) >= difficulty_threshold

For the first block in an account (open block):

blake2b(nonce || public_key) >= difficulty_threshold

Difficulty Thresholds (Epoch v2)¶

Block Type	Threshold
Send / Change	`fffffff800000000`
Receive / Open / Epoch	`fffffe0000000000`

Send/change blocks require higher difficulty because they can be used for spam; receive blocks are less costly to generate work for.

Hardware Recommendations¶

Hardware	Work/second (approx.)
Nvidia RTX 3080	~60 work/s
Nvidia RTX 2080 Ti	~47 work/s
AMD RX 6700 XT	~40 work/s
AMD Ryzen 5 3600 (CPU)	~0.6 work/s

GPUs are strongly preferred. CPU work generation is acceptable for low-volume operations but will cause RPC delays under load.

Recommended Configurations¶

High-Volume Node (Exchange / Service)¶

Dedicated machine with GPU running the Kakitu Work Server
Configure as a work_peer in the node's config
Disable local CPU work: node.work_threads = 0

Low-Volume Node (Personal / Small Integration)¶

Run the Kakitu Work Server on the same machine as the node
Use GPU acceleration (--gpu 0:0)
Set node.work_threads = 0

Kakitu Work Server¶

The Kakitu Work Server is the preferred work generation tool. It runs independently of the node and can service multiple nodes simultaneously.

Install¶

# Download from releases
wget https://github.com/kakitucurrency/kakitu-work-server/releases/latest/download/kakitu_work_server_linux_x86_64
chmod +x kakitu_work_server_linux_x86_64
sudo mv kakitu_work_server_linux_x86_64 /usr/local/bin/kakitu_work_server

Run with GPU¶

kakitu_work_server --listen 127.0.0.1:7078 --gpu 0:0

Run with CPU only¶

kakitu_work_server --listen 127.0.0.1:7078 --cpu-threads 4

Configure as a work peer¶

In config-node.toml:

[node]
work_threads = 0
work_peers = ["::ffff:127.0.0.1:7078"]

Enabling OpenCL (GPU) in the Node Directly¶

If you prefer to use OpenCL directly in the node rather than the work server:

[opencl]
enable = true
platform = 0
device = 0
threads = 1048576

[node]
work_threads = 0

List available OpenCL devices:

kakitu_node --diagnostics

Work Generation via RPC¶

Generate work manually:

curl -s -d '{
  "action": "work_generate",
  "hash": "PREVIOUS_BLOCK_HASH_OR_PUBLIC_KEY",
  "difficulty": "fffffff800000000"
}' http://localhost:7076

{
  "work": "2bf29ef00786a6bc",
  "difficulty": "ffffffff8ba4a60c",
  "multiplier": "8.00",
  "hash": "PREVIOUS_BLOCK_HASH_OR_PUBLIC_KEY"
}

Note

work_generate requires enable_control = true in config-rpc.toml.

Work Validation via RPC¶

curl -s -d '{
  "action": "work_validate",
  "work": "2bf29ef00786a6bc",
  "hash": "PREVIOUS_BLOCK_HASH_OR_PUBLIC_KEY",
  "difficulty": "fffffff800000000"
}' http://localhost:7076

{
  "valid_all": "1",
  "valid_receive": "1",
  "difficulty": "ffffffff8ba4a60c",
  "multiplier": "8.00"
}

Pre-caching Work¶

For optimal latency, generate work for the next block immediately after each confirmed transaction — before the next send is needed.

After a block is confirmed:

Store the block hash
Immediately submit a work_generate request using that hash
Cache the result
When the next block is needed, use the cached work

For wallets where the next block type is unknown, generate at the higher send/change difficulty threshold to ensure the work is valid for any block type.

Work Cancellation¶

Cancel an in-progress work generation:

curl -s -d '{
  "action": "work_cancel",
  "hash": "PREVIOUS_BLOCK_HASH"
}' http://localhost:7076