The prompt contains the surrounding code context and marker placeholders. The completion starts at the first marker and returns the replacement span with markers included.

Typical Input

text

<[fim-suffix]>
  return total;
}

<[fim-prefix]>
function sum(items) {
  let total = 0;
  for (const item of items) {
    <|marker_1|><|marker_2|>
  }

<[fim-middle]>
<|marker_1|>

Typical Output

text

<|marker_1|>
total += item.value;
<|marker_2|>

llama.cpp

bash

llama-server \
  -m lacuna-v1-Q4_K_M.gguf \
  --ctx-size 4096 \
  --host 127.0.0.1 \
  --port 8080

Completion Settings

Setting	Value
`temperature`	`0`
`top_p`	`1`
`max_tokens`	`256`

Use a larger max_tokens value for longer or multi-region edits.

Limitations

Lacuna V1 is experimental. Quantized output can differ from the full model. Validate output before applying edits automatically.

License

Apache-2.0.

Region	Open marker	Close marker
1	`<\|marker_1\|>`	`<\|marker_2\|>`
2	`<\|marker_3\|>`	`<\|marker_4\|>`
N	`<\|marker_2N-1\|>`	`<\|marker_2N\|>`

jolovicdev/Lacuna-V1-GGUF-Q4_K_M

Lacuna V1 GGUF Q4_K_M

At A Glance

Marker Contract