llguidance 0.7.14__tar.gz → 0.7.16__tar.gz

{llguidance-0.7.14 → llguidance-0.7.16}/CHANGELOG.md

@@ -4,6 +4,16 @@ All notable changes to this project will be documented in this file. Dates are d
 
 If a release doesn't introduce any interesting changes (build fixes etc.), it's skipped.
 
+#### [0.7.16](https://github.com/guidance-ai/llguidance/compare/v0.7.15...0.7.16) 2025-04-17
+
+- fix special token tokenization [`ae7870f`](https://github.com/guidance-ai/llguidance/commit/ae7870f05ca0de68599088607ba742b7071f92ad)
+
+#### [0.7.15](https://github.com/guidance-ai/llguidance/compare/v0.7.14...0.7.15) 2025-04-16
+
+- support for patternProperties in JSON schema [`#167`](https://github.com/guidance-ai/llguidance/pull/167)
+- add lenient option to JSON schemas [`#163`](https://github.com/guidance-ai/llguidance/pull/163) [`#136`](https://github.com/guidance-ai/llguidance/issues/136)
+- Add llg_validate_grammar() in C FFI [`e5c21cf`](https://github.com/guidance-ai/llguidance/commit/e5c21cf480a17e6b310e46b24b272576cfd9c4c6)
+
 #### [0.7.14](https://github.com/guidance-ai/llguidance/compare/v0.7.13...0.7.14) 2025-04-11
 
 - support %lark { ... } syntax for nested grammars [`#157`](https://github.com/guidance-ai/llguidance/pull/157)

{llguidance-0.7.14 → llguidance-0.7.16}/Cargo.lock

@@ -1178,7 +1178,7 @@ checksum = "23fb14cb19457329c82206317a5663005a4d404783dc74f4252769b0d5f42856"
 
 [[package]]
 name = "llguidance"
-version = "0.7.14"
+version = "0.7.16"
 dependencies = [
  "anyhow",
  "derivre",
@@ -1197,7 +1197,7 @@ dependencies = [
 
 [[package]]
 name = "llguidance_py"
-version = "0.7.14"
+version = "0.7.16"
 dependencies = [
  "anyhow",
  "bytemuck",
@@ -2362,7 +2362,7 @@ dependencies = [
 
 [[package]]
 name = "toktrie"
-version = "0.7.14"
+version = "0.7.16"
 dependencies = [
  "anyhow",
  "bytemuck",
@@ -2373,7 +2373,7 @@ dependencies = [
 
 [[package]]
 name = "toktrie_hf_downloader"
-version = "0.7.14"
+version = "0.7.16"
 dependencies = [
  "anyhow",
  "hf-hub",
@@ -2384,7 +2384,7 @@ dependencies = [
 
 [[package]]
 name = "toktrie_hf_tokenizers"
-version = "0.7.14"
+version = "0.7.16"
 dependencies = [
  "anyhow",
  "log",

{llguidance-0.7.14 → llguidance-0.7.16}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: llguidance
-Version: 0.7.14
+Version: 0.7.16
 License-File: LICENSE
 Summary: Bindings for the Low-level Guidance (llguidance) Rust library for use within Guidance
 Author: Michal Moskal
@@ -20,6 +20,7 @@ Project-URL: issue_tracker, https://github.com/microsoft/llguidance/issues
 
 --- 
 
+* 2025-04-11 integration [merged](https://github.com/chromium/chromium/commit/07ca6337c2f714ba0477202414bd2b1692e70594) into Chromium
 * 2025-03-25 integration [merged](https://github.com/vllm-project/vllm/pull/14779) into vLLM (v0.8.2)
 * 2025-02-26 integration [merged](https://github.com/sgl-project/sglang/pull/3298) into SGLang (v0.4.4)
 * 2025-02-01 integration [merged](https://github.com/ggml-org/llama.cpp/pull/10224) into llama.cpp (b4613)
@@ -61,16 +62,17 @@ The library is currently integrated in:
 - [llama.cpp](https://github.com/ggerganov/llama.cpp/pull/10224) - 
   available via `-DLLAMA_LLGUIDANCE=ON` option for `cmake`;
   llama.cpp can be also used Guidance Python package
+- **Chromium** - [merged](https://github.com/chromium/chromium/commit/07ca6337c2f714ba0477202414bd2b1692e70594),
+  to be used for [JSON Schema enforcement](https://github.com/webmachinelearning/prompt-api?tab=readme-ov-file#structured-output-or-json-output) for `window.ai` in Chromium-based browsers
 - [SGLang](https://github.com/sgl-project/sglang/pull/3298) -
   use `--grammar-backend llguidance`; when passing Lark grammar make
   sure to prefix them with `%llguidance {}`, just as in llama.cpp
-- vLLM - [V0 PR](https://github.com/vllm-project/vllm/pull/14589) and [V1 PR](https://github.com/vllm-project/vllm/pull/14779)
+- **vLLM** - [V0 PR](https://github.com/vllm-project/vllm/pull/14589) and [V1 PR](https://github.com/vllm-project/vllm/pull/14779)
 - [LLGTRT](https://github.com/guidance-ai/llgtrt) - OpenAI-compatible REST server using NVIDIA's [TensorRT-LLM](https://github.com/NVIDIA/TensorRT-LLM)
 - [mistral.rs](https://github.com/EricLBuehler/mistral.rs/pull/899)
 
 The integration is ongoing in:
-- onnxruntime-genai - [draft PR](https://github.com/microsoft/onnxruntime-genai/pull/1038)
-- Chromium - [ongoing PR](https://chromium-review.googlesource.com/c/chromium/src/+/6232561)
+- **onnxruntime-genai** - [draft PR](https://github.com/microsoft/onnxruntime-genai/pull/1038)
 
 ## Technical details
 

{llguidance-0.7.14 → llguidance-0.7.16}/README.md

@@ -8,6 +8,7 @@
 
 --- 
 
+* 2025-04-11 integration [merged](https://github.com/chromium/chromium/commit/07ca6337c2f714ba0477202414bd2b1692e70594) into Chromium
 * 2025-03-25 integration [merged](https://github.com/vllm-project/vllm/pull/14779) into vLLM (v0.8.2)
 * 2025-02-26 integration [merged](https://github.com/sgl-project/sglang/pull/3298) into SGLang (v0.4.4)
 * 2025-02-01 integration [merged](https://github.com/ggml-org/llama.cpp/pull/10224) into llama.cpp (b4613)
@@ -49,16 +50,17 @@ The library is currently integrated in:
 - [llama.cpp](https://github.com/ggerganov/llama.cpp/pull/10224) - 
   available via `-DLLAMA_LLGUIDANCE=ON` option for `cmake`;
   llama.cpp can be also used Guidance Python package
+- **Chromium** - [merged](https://github.com/chromium/chromium/commit/07ca6337c2f714ba0477202414bd2b1692e70594),
+  to be used for [JSON Schema enforcement](https://github.com/webmachinelearning/prompt-api?tab=readme-ov-file#structured-output-or-json-output) for `window.ai` in Chromium-based browsers
 - [SGLang](https://github.com/sgl-project/sglang/pull/3298) -
   use `--grammar-backend llguidance`; when passing Lark grammar make
   sure to prefix them with `%llguidance {}`, just as in llama.cpp
-- vLLM - [V0 PR](https://github.com/vllm-project/vllm/pull/14589) and [V1 PR](https://github.com/vllm-project/vllm/pull/14779)
+- **vLLM** - [V0 PR](https://github.com/vllm-project/vllm/pull/14589) and [V1 PR](https://github.com/vllm-project/vllm/pull/14779)
 - [LLGTRT](https://github.com/guidance-ai/llgtrt) - OpenAI-compatible REST server using NVIDIA's [TensorRT-LLM](https://github.com/NVIDIA/TensorRT-LLM)
 - [mistral.rs](https://github.com/EricLBuehler/mistral.rs/pull/899)
 
 The integration is ongoing in:
-- onnxruntime-genai - [draft PR](https://github.com/microsoft/onnxruntime-genai/pull/1038)
-- Chromium - [ongoing PR](https://chromium-review.googlesource.com/c/chromium/src/+/6232561)
+- **onnxruntime-genai** - [draft PR](https://github.com/microsoft/onnxruntime-genai/pull/1038)
 
 ## Technical details
 

{llguidance-0.7.14 → llguidance-0.7.16}/docs/json_schema.md

@@ -15,7 +15,7 @@ Following JSON schema features are supported.
 Core features:
 
 - `anyOf`
-- `oneOf` - not supported right now, use `anyOf` instead, [issue](https://github.com/microsoft/llguidance/issues/77)
+- `oneOf` - converted to `anyOf` only when provably equivalent
 - `allOf` - intersection of certain schemas is not supported right now
 - `$ref` - external/remote refs unsupported
 - `const`
@@ -34,13 +34,14 @@ Object features:
 
 - `properties` - order of properties is fixed to the order in schema
 - `additionalProperties`
+- `patternProperties` - they have to be disjoint
 - `required`
 
 String features:
 
 - `minLength`
 - `maxLength`
-- `pattern` (though we always anchor them, [issue](https://github.com/microsoft/llguidance/issues/66))
+- `pattern`
 - `format`, with the following formats: `date-time`, `time`, `date`, `duration`, `email`, `hostname`, `ipv4`, `ipv6`, `uuid`,
 
 Number features (for both integer and number):
@@ -55,6 +56,8 @@ Number features (for both integer and number):
 
 - order of object properties is fixed, see below
 - string `format` is enforced by default, with unrecognized or unimplemented formats returning errors
+- for properties specified with `additionalProperties` or `patternProperties`, the grammar does not enforce unique keys;
+  the ones listed in `properties` are enforced uniquely, in given order
 
 ## Whitespace handling
 
@@ -71,6 +74,7 @@ Following keys are available inside of it:
 - `whitespace_pattern`, optional string, overrides `whitespace_flexible`;
   `whitespace_flexible: true` is equivalent to `whitespace_pattern: r"[\x20\x0A\x0D\x09]+"`
 - `coerce_one_of`, defaults to `false`; when set to `true`, the `"oneOf"` will be treated as `"anyOf"`
+- `lenient`, defaults to `false`; when set to `true`, the unsupported keywords and formats will be ignored; implies `coerce_one_of: true`
 
 For example:
 
@@ -97,6 +101,7 @@ The `"x-guidance"` key is only recognized at the top level of the schema.
 
 Properties follow order in `properties` map.
 When schemas are merged with `allOf` etc., the `properties` maps are merged in order.
+Any `additionalProperties` or `patternProperties` come in any order, but after `properties` and `required`.
 
 Easiest way to override this, is to include `"my_property": true` in appropriate position in `"properties"`,
 before `anyOf/allOf/oneOf/$ref`.
@@ -108,7 +113,7 @@ Basically, the schema is processed line-by-line, left-to-right, and property ord
 
 The enforced property order during generation is as follows:
 1. Each property in the `"properties"` object, in order of appearance
-2. Each property in `"required"`, in order of appearance (if not already in `"properties"` they are constrained with `"additionalPropertie"`)
+2. Each property in `"required"`, in order of appearance (if not already in `"properties"` they are constrained with `"additionalProperties"`)
 
 When two schemas are joined (more than two is defined inductively), the resulting `"properties"` object will have order given by:
 1. Each property in the left schema, in order of appearance

{llguidance-0.7.14 → llguidance-0.7.16}/docs/syntax.md

@@ -56,8 +56,8 @@ This will accepting strings like `There is no function I can call`
 and `{"name":"get_weather", "parameters": {"city":"Seattle"}}`.
 If the string starts with `{`, it will be forced to follow the given JSON schema.
 
-If you have more functions, you should use `%json { "anyOf": [ ... ] }`.
-Do not use `fun_call1 | fun_call2`, as it [currently doesn't work](https://github.com/guidance-ai/llguidance/issues/113).
+If you have more functions, you can use either `%json { "anyOf": [ ... ] }`,
+or `fun_call1 | fun_call2` where each `fun_callX` is defined as `%json { ... }`.
 
 ### Special tokens