update to latest llguidance

.gitmodules

@@ -1,6 +1,3 @@
-[submodule "toktrie"]
-	path = toktrie
-	url = https://github.com/microsoft/toktrie
 [submodule "llguidance"]
 	path = llguidance
 	url = https://github.com/microsoft/llguidance

Cargo.toml

@@ -3,23 +3,23 @@ members = [
     "trtllm_rs",
     "llgtrt",
     "llguidance/parser",
-    "llguidance/sample_parser",
-    "llguidance/rust",
-    "toktrie/core",
-    "toktrie/hf_tokenizers",
+    "llguidance/toktrie",
+    "llguidance/toktrie_hf_tokenizers",
     "derivre",
 ]
 exclude = [
-    "llguidance/rust",
+    "llguidance",
 ]
 resolver = "2"
 
 [profile.release]
 # debug = 1
 
-[patch.'https://github.com/microsoft/toktrie']
-toktrie = { path = "toktrie/core" }
-toktrie_hf_tokenizers = { path = "toktrie/hf_tokenizers" }
-
 [patch.'https://github.com/microsoft/derivre']
 derivre = { path = "derivre" }
+
+[workspace.dependencies]
+toktrie = { path = "llguidance/toktrie" }
+llguidance = { path = "llguidance/parser" }
+toktrie_hf_tokenizers = { path = "llguidance/toktrie_hf_tokenizers" }
+trtllm_rs = { path = "trtllm_rs" }

llgtrt/Cargo.toml

@@ -12,13 +12,13 @@ clap = { version = "4.4.7", features = ["derive"] }
 serde = { version = "1.0.190", features = ["derive"] }
 serde_json = "1.0.108"
 uuid = { version = "1.6.1", features = ["v4"] }
-trtllm_rs = { path = "../trtllm_rs" }
-toktrie = { path = "../toktrie/core" }
-toktrie_hf_tokenizers = { path = "../toktrie/hf_tokenizers" }
+trtllm_rs = { workspace = true }
 flexi_logger = "0.29.0"
 log = "0.4.22"
 rand = "0.8.5"
-llguidance_parser = { path = "../llguidance/parser" }
+llguidance = { workspace = true }
+toktrie = { workspace = true }
+toktrie_hf_tokenizers = { workspace = true }
 rayon = "1.10.0"
 futures-core = "0.3.30"
 minijinja = { version = "2.3.1", features = ["preserve_order", "loop_controls", "loader"] }

llgtrt/src/async_exec.rs

@@ -1,5 +1,5 @@
 use anyhow::{ensure, Result};
-use llguidance_parser::Constraint;
+use llguidance::Constraint;
 use rayon::prelude::*;
 use std::{
     any::Any,
@@ -127,7 +127,7 @@ impl PendingSeq {
             llg.compute_mask()?
         } else {
             // if we're still in prompt
-            if !llg.has_current_step_result() {
+            if llg.step_result().sample_mask.is_none() {
                 // first time, compute the mask
                 llg.compute_mask()?
             } else {

llgtrt/src/constraint_mgr.rs

@@ -1,6 +1,6 @@
 use crate::routes::api_ext::LlgLogLevel;
 use anyhow::Result;
-use llguidance_parser::{
+use llguidance::{
     api::{ParserLimits, TopLevelGrammar},
     Constraint, Logger, TokenParser,
 };

llgtrt/src/routes/api_ext.rs

@@ -1,4 +1,4 @@
-use llguidance_parser::api::TopLevelGrammar;
+use llguidance::api::TopLevelGrammar;
 use serde::{Deserialize, Serialize};
 use serde_json::{json, Value};
 

llgtrt/src/routes/completions.rs

@@ -7,8 +7,8 @@ use axum::response::sse::{Event, Sse};
 use axum::response::{IntoResponse, Response};
 use axum::Json;
 use futures_core::Stream;
-use llguidance_parser::api::{RegexSpec, TopLevelGrammar};
-use llguidance_parser::{lark_to_llguidance, Constraint, GrammarBuilder, JsonCompileOptions};
+use llguidance::api::{RegexSpec, TopLevelGrammar};
+use llguidance::{lark_to_llguidance, Constraint, GrammarBuilder, JsonCompileOptions};
 use serde_json::{json, Value};
 use std::fmt::Display;
 use std::sync::Arc;

llgtrt/src/routes/openai.rs

@@ -1,5 +1,5 @@
 use super::api_ext::LlgLogLevel;
-use llguidance_parser::api::TopLevelGrammar;
+use llguidance::api::TopLevelGrammar;
 use serde::{Deserialize, Deserializer, Serialize};
 use std::collections::HashMap;
 use toktrie::{TokTrie, TokenId};