1# LLGuidance Support in llama.cpp
 2
 3[LLGuidance](https://github.com/guidance-ai/llguidance) is a library for constrained decoding (also called constrained sampling or structured outputs) for Large Language Models (LLMs). Initially developed as the backend for the [Guidance](https://github.com/guidance-ai/guidance) library, it can also be used independently.
 4
 5LLGuidance supports JSON Schemas and arbitrary context-free grammars (CFGs) written in a [variant](https://github.com/guidance-ai/llguidance/blob/main/docs/syntax.md) of Lark syntax. It is [very fast](https://github.com/guidance-ai/jsonschemabench/tree/main/maskbench) and has [excellent](https://github.com/guidance-ai/llguidance/blob/main/docs/json_schema.md) JSON Schema coverage but requires the Rust compiler, which complicates the llama.cpp build process.
 6
 7## Building
 8
 9To enable LLGuidance support, build llama.cpp with the `LLAMA_LLGUIDANCE` option:
10
11```sh
12cmake -B build -DLLAMA_LLGUIDANCE=ON
13make -C build -j
14```
15
16For Windows use `cmake --build build --config Release` instead of `make`.
17
18This requires the Rust compiler and the `cargo` tool to be [installed](https://www.rust-lang.org/tools/install).
19
20## Interface
21
22There are no new command-line arguments or modifications to `common_params`. When enabled, grammars starting with `%llguidance` are passed to LLGuidance instead of the [current](../grammars/README.md) llama.cpp grammars. Additionally, JSON Schema requests (e.g., using the `-j` argument in `llama-cli`) are also passed to LLGuidance.
23
24For your existing GBNF grammars, you can use [gbnf_to_lark.py script](https://github.com/guidance-ai/llguidance/blob/main/python/llguidance/gbnf_to_lark.py) to convert them to LLGuidance Lark-like format.
25
26## Performance
27
28Computing a "token mask" (i.e., the set of allowed tokens) for a llama3 tokenizer with 128k tokens takes, on average, 50μs of single-core CPU time for the [JSON Schema Bench](https://github.com/guidance-ai/jsonschemabench). The p99 time is 0.5ms, and the p100 time is 20ms. These results are due to the lexer/parser split and several [optimizations](https://github.com/guidance-ai/llguidance/blob/main/docs/optimizations.md).
29
30## JSON Schema
31
32LLGuidance adheres closely to the JSON Schema specification. For example:
33
34- `additionalProperties` defaults to `true`, unlike current grammars, though you can set `"additionalProperties": false` if needed.
35- any whitespace is allowed.
36- The definition order in the `"properties": {}` object is maintained, regardless of whether properties are required (current grammars always puts required properties first).
37
38Unsupported schemas result in an error message—no keywords are silently ignored.
39
40## Why Not Reuse GBNF Format?
41
42GBNF lacks the concept of a lexer.
43
44Most programming languages, including JSON, use a two-step process: a lexer (built with regular expressions) converts a byte stream into lexemes, which are then processed by a CFG parser. This approach is faster because lexers are cheaper to evaluate, and there is ~10x fewer lexemes than bytes.
45LLM tokens often align with lexemes, so the parser is engaged in under 0.5% of tokens, with the lexer handling the rest.
46
47However, the user has to provide the distinction between lexemes and CFG symbols. In [Lark](https://github.com/lark-parser/lark), lexeme names are uppercase, while CFG symbols are lowercase.
48The [gbnf_to_lark.py script](https://github.com/guidance-ai/llguidance/blob/main/scripts/gbnf_to_lark.py) can often take care of this automatically.
49See [LLGuidance syntax docs](https://github.com/guidance-ai/llguidance/blob/main/docs/syntax.md#terminals-vs-rules) for more details.
50
51## Error Handling
52
53Errors are currently printed to `stderr`, and generation continues. Improved error handling may be added in the future.