diff --git a/docs/API/README.md b/docs/API/README.md
index 613b1e0..a837766 100644
--- a/docs/API/README.md
+++ b/docs/API/README.md
@@ -8,6 +8,11 @@ llama.rn
 
 - [LlamaContext](classes/LlamaContext.md)
 - [SchemaGrammarConverter](classes/SchemaGrammarConverter.md)
+- [SchemaGrammarConverterBuiltinRule](classes/SchemaGrammarConverterBuiltinRule.md)
+
+### Interfaces
+
+- [SchemaGrammarConverterPropOrder](interfaces/SchemaGrammarConverterPropOrder.md)
 
 ### Type Aliases
 
@@ -15,6 +20,19 @@ llama.rn
 - [CompletionParams](README.md#completionparams)
 - [ContextParams](README.md#contextparams)
 - [EmbeddingParams](README.md#embeddingparams)
+- [NativeCompletionParams](README.md#nativecompletionparams)
+- [NativeCompletionResult](README.md#nativecompletionresult)
+- [NativeCompletionResultTimings](README.md#nativecompletionresulttimings)
+- [NativeCompletionTokenProb](README.md#nativecompletiontokenprob)
+- [NativeCompletionTokenProbItem](README.md#nativecompletiontokenprobitem)
+- [NativeContextParams](README.md#nativecontextparams)
+- [NativeEmbeddingParams](README.md#nativeembeddingparams)
+- [NativeEmbeddingResult](README.md#nativeembeddingresult)
+- [NativeLlamaContext](README.md#nativellamacontext)
+- [NativeSessionLoadResult](README.md#nativesessionloadresult)
+- [NativeTokenizeResult](README.md#nativetokenizeresult)
+- [RNLlamaMessagePart](README.md#rnllamamessagepart)
+- [RNLlamaOAICompatibleMessage](README.md#rnllamaoaicompatiblemessage)
 - [TokenData](README.md#tokendata)
 
 ### Functions
@@ -45,37 +63,317 @@ llama.rn
 
 #### Defined in
 
-[index.ts:58](https://github.com/mybigday/llama.rn/blob/68acf1a/src/index.ts#L58)
+[index.ts:84](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/index.ts#L84)
 
 ___
 
 ### CompletionParams
 
-Ƭ **CompletionParams**: `Omit`<`NativeCompletionParams`, ``"emit_partial_completion"`` \| ``"prompt"``\> & { `chatTemplate?`: `string` ; `messages?`: `RNLlamaOAICompatibleMessage`[] ; `prompt?`: `string`  }
+Ƭ **CompletionParams**: `Omit`<[`NativeCompletionParams`](README.md#nativecompletionparams), ``"emit_partial_completion"`` \| ``"prompt"``\> & { `chatTemplate?`: `string` ; `messages?`: [`RNLlamaOAICompatibleMessage`](README.md#rnllamaoaicompatiblemessage)[] ; `prompt?`: `string`  }
 
 #### Defined in
 
-[index.ts:49](https://github.com/mybigday/llama.rn/blob/68acf1a/src/index.ts#L49)
+[index.ts:75](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/index.ts#L75)
 
 ___
 
 ### ContextParams
 
-Ƭ **ContextParams**: `Omit`<`NativeContextParams`, ``"pooling_type"``\> & { `pooling_type?`: ``"none"`` \| ``"mean"`` \| ``"cls"`` \| ``"last"`` \| ``"rank"``  }
+Ƭ **ContextParams**: `Omit`<[`NativeContextParams`](README.md#nativecontextparams), ``"cache_type_k"`` \| ``"cache_type_v"`` \| ``"pooling_type"``\> & { `cache_type_k?`: ``"f16"`` \| ``"f32"`` \| ``"q8_0"`` \| ``"q4_0"`` \| ``"q4_1"`` \| ``"iq4_nl"`` \| ``"q5_0"`` \| ``"q5_1"`` ; `cache_type_v?`: ``"f16"`` \| ``"f32"`` \| ``"q8_0"`` \| ``"q4_0"`` \| ``"q4_1"`` \| ``"iq4_nl"`` \| ``"q5_0"`` \| ``"q5_1"`` ; `pooling_type?`: ``"none"`` \| ``"mean"`` \| ``"cls"`` \| ``"last"`` \| ``"rank"``  }
 
 #### Defined in
 
-[index.ts:43](https://github.com/mybigday/llama.rn/blob/68acf1a/src/index.ts#L43)
+[index.ts:64](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/index.ts#L64)
 
 ___
 
 ### EmbeddingParams
 
-Ƭ **EmbeddingParams**: `NativeEmbeddingParams`
+Ƭ **EmbeddingParams**: [`NativeEmbeddingParams`](README.md#nativeembeddingparams)
+
+#### Defined in
+
+[index.ts:73](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/index.ts#L73)
+
+___
+
+### NativeCompletionParams
+
+Ƭ **NativeCompletionParams**: `Object`
+
+#### Type declaration
+
+| Name | Type | Description |
+| :------ | :------ | :------ |
+| `dry_allowed_length?` | `number` | Tokens that extend repetition beyond this receive exponentially increasing penalty: multiplier * base ^ (length of repeating sequence before token - allowed length). Default: `2` |
+| `dry_base?` | `number` | Set the DRY repetition penalty base value. Default: `1.75` |
+| `dry_multiplier?` | `number` | Set the DRY (Don't Repeat Yourself) repetition penalty multiplier. Default: `0.0`, which is disabled. |
+| `dry_penalty_last_n?` | `number` | How many tokens to scan for repetitions. Default: `-1`, where `0` is disabled and `-1` is context size. |
+| `dry_sequence_breakers?` | `string`[] | Specify an array of sequence breakers for DRY sampling. Only a JSON array of strings is accepted. Default: `['\n', ':', '"', '*']` |
+| `emit_partial_completion` | `boolean` | - |
+| `grammar?` | `string` | Set grammar for grammar-based sampling. Default: no grammar |
+| `ignore_eos?` | `boolean` | Ignore end of stream token and continue generating. Default: `false` |
+| `logit_bias?` | `number`[][] | Modify the likelihood of a token appearing in the generated text completion. For example, use `"logit_bias": [[15043,1.0]]` to increase the likelihood of the token 'Hello', or `"logit_bias": [[15043,-1.0]]` to decrease its likelihood. Setting the value to false, `"logit_bias": [[15043,false]]` ensures that the token `Hello` is never produced. The tokens can also be represented as strings, e.g.`[["Hello, World!",-0.5]]` will reduce the likelihood of all the individual tokens that represent the string `Hello, World!`, just like the `presence_penalty` does. Default: `[]` |
+| `min_p?` | `number` | The minimum probability for a token to be considered, relative to the probability of the most likely token. Default: `0.05` |
+| `mirostat?` | `number` | Enable Mirostat sampling, controlling perplexity during text generation. Default: `0`, where `0` is disabled, `1` is Mirostat, and `2` is Mirostat 2.0. |
+| `mirostat_eta?` | `number` | Set the Mirostat learning rate, parameter eta. Default: `0.1` |
+| `mirostat_tau?` | `number` | Set the Mirostat target entropy, parameter tau. Default: `5.0` |
+| `n_predict?` | `number` | Set the maximum number of tokens to predict when generating text. **Note:** May exceed the set limit slightly if the last token is a partial multibyte character. When 0,no tokens will be generated but the prompt is evaluated into the cache. Default: `-1`, where `-1` is infinity. |
+| `n_probs?` | `number` | If greater than 0, the response also contains the probabilities of top N tokens for each generated token given the sampling settings. Note that for temperature < 0 the tokens are sampled greedily but token probabilities are still being calculated via a simple softmax of the logits without considering any other sampler settings. Default: `0` |
+| `n_threads?` | `number` | - |
+| `penalize_nl?` | `boolean` | Penalize newline tokens when applying the repeat penalty. Default: `false` |
+| `penalty_freq?` | `number` | Repeat alpha frequency penalty. Default: `0.0`, which is disabled. |
+| `penalty_last_n?` | `number` | Last n tokens to consider for penalizing repetition. Default: `64`, where `0` is disabled and `-1` is ctx-size. |
+| `penalty_present?` | `number` | Repeat alpha presence penalty. Default: `0.0`, which is disabled. |
+| `penalty_repeat?` | `number` | Control the repetition of token sequences in the generated text. Default: `1.0` |
+| `prompt` | `string` | - |
+| `seed?` | `number` | Set the random number generator (RNG) seed. Default: `-1`, which is a random seed. |
+| `stop?` | `string`[] | Specify a JSON array of stopping strings. These words will not be included in the completion, so make sure to add them to the prompt for the next iteration. Default: `[]` |
+| `temperature?` | `number` | Adjust the randomness of the generated text. Default: `0.8` |
+| `top_k?` | `number` | Limit the next token selection to the K most probable tokens. Default: `40` |
+| `top_p?` | `number` | Limit the next token selection to a subset of tokens with a cumulative probability above a threshold P. Default: `0.95` |
+| `typical_p?` | `number` | Enable locally typical sampling with parameter p. Default: `1.0`, which is disabled. |
+| `xtc_probability?` | `number` | Set the chance for token removal via XTC sampler. Default: `0.0`, which is disabled. |
+| `xtc_threshold?` | `number` | Set a minimum probability threshold for tokens to be removed via XTC sampler. Default: `0.1` (> `0.5` disables XTC) |
+
+#### Defined in
+
+[NativeRNLlama.ts:50](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/NativeRNLlama.ts#L50)
+
+___
+
+### NativeCompletionResult
+
+Ƭ **NativeCompletionResult**: `Object`
+
+#### Type declaration
+
+| Name | Type |
+| :------ | :------ |
+| `completion_probabilities?` | [`NativeCompletionTokenProb`](README.md#nativecompletiontokenprob)[] |
+| `stopped_eos` | `boolean` |
+| `stopped_limit` | `number` |
+| `stopped_word` | `string` |
+| `stopping_word` | `string` |
+| `text` | `string` |
+| `timings` | [`NativeCompletionResultTimings`](README.md#nativecompletionresulttimings) |
+| `tokens_cached` | `number` |
+| `tokens_evaluated` | `number` |
+| `tokens_predicted` | `number` |
+| `truncated` | `boolean` |
+
+#### Defined in
+
+[NativeRNLlama.ts:195](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/NativeRNLlama.ts#L195)
+
+___
+
+### NativeCompletionResultTimings
+
+Ƭ **NativeCompletionResultTimings**: `Object`
+
+#### Type declaration
+
+| Name | Type |
+| :------ | :------ |
+| `predicted_ms` | `number` |
+| `predicted_n` | `number` |
+| `predicted_per_second` | `number` |
+| `predicted_per_token_ms` | `number` |
+| `prompt_ms` | `number` |
+| `prompt_n` | `number` |
+| `prompt_per_second` | `number` |
+| `prompt_per_token_ms` | `number` |
+
+#### Defined in
+
+[NativeRNLlama.ts:184](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/NativeRNLlama.ts#L184)
+
+___
+
+### NativeCompletionTokenProb
+
+Ƭ **NativeCompletionTokenProb**: `Object`
+
+#### Type declaration
+
+| Name | Type |
+| :------ | :------ |
+| `content` | `string` |
+| `probs` | [`NativeCompletionTokenProbItem`](README.md#nativecompletiontokenprobitem)[] |
+
+#### Defined in
+
+[NativeRNLlama.ts:179](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/NativeRNLlama.ts#L179)
+
+___
+
+### NativeCompletionTokenProbItem
+
+Ƭ **NativeCompletionTokenProbItem**: `Object`
+
+#### Type declaration
+
+| Name | Type |
+| :------ | :------ |
+| `prob` | `number` |
+| `tok_str` | `string` |
+
+#### Defined in
+
+[NativeRNLlama.ts:174](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/NativeRNLlama.ts#L174)
+
+___
+
+### NativeContextParams
+
+Ƭ **NativeContextParams**: `Object`
+
+#### Type declaration
+
+| Name | Type | Description |
+| :------ | :------ | :------ |
+| `cache_type_k?` | `string` | KV cache data type for the K (Experimental in llama.cpp) |
+| `cache_type_v?` | `string` | KV cache data type for the V (Experimental in llama.cpp) |
+| `embd_normalize?` | `number` | - |
+| `embedding?` | `boolean` | - |
+| `flash_attn?` | `boolean` | Enable flash attention, only recommended in GPU device (Experimental in llama.cpp) |
+| `is_model_asset?` | `boolean` | - |
+| `lora?` | `string` | - |
+| `lora_scaled?` | `number` | - |
+| `model` | `string` | - |
+| `n_batch?` | `number` | - |
+| `n_ctx?` | `number` | - |
+| `n_gpu_layers?` | `number` | - |
+| `n_threads?` | `number` | - |
+| `pooling_type?` | `number` | - |
+| `rope_freq_base?` | `number` | - |
+| `rope_freq_scale?` | `number` | - |
+| `use_mlock?` | `boolean` | - |
+| `use_mmap?` | `boolean` | - |
+| `use_progress_callback?` | `boolean` | - |
+| `vocab_only?` | `boolean` | - |
+
+#### Defined in
+
+[NativeRNLlama.ts:8](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/NativeRNLlama.ts#L8)
+
+___
+
+### NativeEmbeddingParams
+
+Ƭ **NativeEmbeddingParams**: `Object`
+
+#### Type declaration
+
+| Name | Type |
+| :------ | :------ |
+| `embd_normalize?` | `number` |
+
+#### Defined in
+
+[NativeRNLlama.ts:4](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/NativeRNLlama.ts#L4)
+
+___
+
+### NativeEmbeddingResult
+
+Ƭ **NativeEmbeddingResult**: `Object`
+
+#### Type declaration
+
+| Name | Type |
+| :------ | :------ |
+| `embedding` | `number`[] |
+
+#### Defined in
+
+[NativeRNLlama.ts:215](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/NativeRNLlama.ts#L215)
+
+___
+
+### NativeLlamaContext
+
+Ƭ **NativeLlamaContext**: `Object`
+
+#### Type declaration
+
+| Name | Type |
+| :------ | :------ |
+| `contextId` | `number` |
+| `gpu` | `boolean` |
+| `model` | `Object` |
+| `reasonNoGPU` | `string` |
+
+#### Defined in
+
+[NativeRNLlama.ts:219](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/NativeRNLlama.ts#L219)
+
+___
+
+### NativeSessionLoadResult
+
+Ƭ **NativeSessionLoadResult**: `Object`
+
+#### Type declaration
+
+| Name | Type |
+| :------ | :------ |
+| `prompt` | `string` |
+| `tokens_loaded` | `number` |
+
+#### Defined in
+
+[NativeRNLlama.ts:226](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/NativeRNLlama.ts#L226)
+
+___
+
+### NativeTokenizeResult
+
+Ƭ **NativeTokenizeResult**: `Object`
+
+#### Type declaration
+
+| Name | Type |
+| :------ | :------ |
+| `tokens` | `number`[] |
+
+#### Defined in
+
+[NativeRNLlama.ts:211](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/NativeRNLlama.ts#L211)
+
+___
+
+### RNLlamaMessagePart
+
+Ƭ **RNLlamaMessagePart**: `Object`
+
+#### Type declaration
+
+| Name | Type |
+| :------ | :------ |
+| `text?` | `string` |
+
+#### Defined in
+
+[chat.ts:3](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/chat.ts#L3)
+
+___
+
+### RNLlamaOAICompatibleMessage
+
+Ƭ **RNLlamaOAICompatibleMessage**: `Object`
+
+#### Type declaration
+
+| Name | Type |
+| :------ | :------ |
+| `content?` | `string` \| [`RNLlamaMessagePart`](README.md#rnllamamessagepart)[] \| `any` |
+| `role` | `string` |
 
 #### Defined in
 
-[index.ts:47](https://github.com/mybigday/llama.rn/blob/68acf1a/src/index.ts#L47)
+[chat.ts:7](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/chat.ts#L7)
 
 ___
 
@@ -87,12 +385,12 @@ ___
 
 | Name | Type |
 | :------ | :------ |
-| `completion_probabilities?` | `NativeCompletionTokenProb`[] |
+| `completion_probabilities?` | [`NativeCompletionTokenProb`](README.md#nativecompletiontokenprob)[] |
 | `token` | `string` |
 
 #### Defined in
 
-[index.ts:33](https://github.com/mybigday/llama.rn/blob/68acf1a/src/index.ts#L33)
+[index.ts:54](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/index.ts#L54)
 
 ## Functions
 
@@ -107,7 +405,7 @@ ___
 | `«destructured»` | `Object` |
 | › `allowFetch?` | `boolean` |
 | › `dotall?` | `boolean` |
-| › `propOrder?` | `PropOrder` |
+| › `propOrder?` | [`SchemaGrammarConverterPropOrder`](interfaces/SchemaGrammarConverterPropOrder.md) |
 | › `schema` | `any` |
 
 #### Returns
@@ -116,7 +414,7 @@ ___
 
 #### Defined in
 
-[grammar.ts:824](https://github.com/mybigday/llama.rn/blob/68acf1a/src/grammar.ts#L824)
+[grammar.ts:826](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/grammar.ts#L826)
 
 ___
 
@@ -137,7 +435,7 @@ ___
 
 #### Defined in
 
-[index.ts:225](https://github.com/mybigday/llama.rn/blob/68acf1a/src/index.ts#L225)
+[index.ts:251](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/index.ts#L251)
 
 ___
 
@@ -157,7 +455,7 @@ ___
 
 #### Defined in
 
-[index.ts:210](https://github.com/mybigday/llama.rn/blob/68acf1a/src/index.ts#L210)
+[index.ts:236](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/index.ts#L236)
 
 ___
 
@@ -171,7 +469,7 @@ ___
 
 #### Defined in
 
-[index.ts:275](https://github.com/mybigday/llama.rn/blob/68acf1a/src/index.ts#L275)
+[index.ts:301](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/index.ts#L301)
 
 ___
 
@@ -191,4 +489,4 @@ ___
 
 #### Defined in
 
-[index.ts:196](https://github.com/mybigday/llama.rn/blob/68acf1a/src/index.ts#L196)
+[index.ts:222](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/index.ts#L222)
diff --git a/docs/API/classes/LlamaContext.md b/docs/API/classes/LlamaContext.md
index f1e2263..5265475 100644
--- a/docs/API/classes/LlamaContext.md
+++ b/docs/API/classes/LlamaContext.md
@@ -38,11 +38,11 @@
 
 | Name | Type |
 | :------ | :------ |
-| `«destructured»` | `NativeLlamaContext` |
+| `«destructured»` | [`NativeLlamaContext`](../README.md#nativellamacontext) |
 
 #### Defined in
 
-[index.ts:79](https://github.com/mybigday/llama.rn/blob/68acf1a/src/index.ts#L79)
+[index.ts:105](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/index.ts#L105)
 
 ## Properties
 
@@ -52,7 +52,7 @@
 
 #### Defined in
 
-[index.ts:71](https://github.com/mybigday/llama.rn/blob/68acf1a/src/index.ts#L71)
+[index.ts:97](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/index.ts#L97)
 
 ___
 
@@ -62,7 +62,7 @@ ___
 
 #### Defined in
 
-[index.ts:69](https://github.com/mybigday/llama.rn/blob/68acf1a/src/index.ts#L69)
+[index.ts:95](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/index.ts#L95)
 
 ___
 
@@ -78,7 +78,7 @@ ___
 
 #### Defined in
 
-[index.ts:75](https://github.com/mybigday/llama.rn/blob/68acf1a/src/index.ts#L75)
+[index.ts:101](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/index.ts#L101)
 
 ___
 
@@ -88,7 +88,7 @@ ___
 
 #### Defined in
 
-[index.ts:73](https://github.com/mybigday/llama.rn/blob/68acf1a/src/index.ts#L73)
+[index.ts:99](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/index.ts#L99)
 
 ## Methods
 
@@ -111,13 +111,13 @@ ___
 
 #### Defined in
 
-[index.ts:171](https://github.com/mybigday/llama.rn/blob/68acf1a/src/index.ts#L171)
+[index.ts:197](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/index.ts#L197)
 
 ___
 
 ### completion
 
-▸ **completion**(`params`, `callback?`): `Promise`<`NativeCompletionResult`\>
+▸ **completion**(`params`, `callback?`): `Promise`<[`NativeCompletionResult`](../README.md#nativecompletionresult)\>
 
 #### Parameters
 
@@ -128,11 +128,11 @@ ___
 
 #### Returns
 
-`Promise`<`NativeCompletionResult`\>
+`Promise`<[`NativeCompletionResult`](../README.md#nativecompletionresult)\>
 
 #### Defined in
 
-[index.ts:115](https://github.com/mybigday/llama.rn/blob/68acf1a/src/index.ts#L115)
+[index.ts:141](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/index.ts#L141)
 
 ___
 
@@ -152,28 +152,28 @@ ___
 
 #### Defined in
 
-[index.ts:160](https://github.com/mybigday/llama.rn/blob/68acf1a/src/index.ts#L160)
+[index.ts:186](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/index.ts#L186)
 
 ___
 
 ### embedding
 
-▸ **embedding**(`text`, `params?`): `Promise`<`NativeEmbeddingResult`\>
+▸ **embedding**(`text`, `params?`): `Promise`<[`NativeEmbeddingResult`](../README.md#nativeembeddingresult)\>
 
 #### Parameters
 
 | Name | Type |
 | :------ | :------ |
 | `text` | `string` |
-| `params?` | `NativeEmbeddingParams` |
+| `params?` | [`NativeEmbeddingParams`](../README.md#nativeembeddingparams) |
 
 #### Returns
 
-`Promise`<`NativeEmbeddingResult`\>
+`Promise`<[`NativeEmbeddingResult`](../README.md#nativeembeddingresult)\>
 
 #### Defined in
 
-[index.ts:164](https://github.com/mybigday/llama.rn/blob/68acf1a/src/index.ts#L164)
+[index.ts:190](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/index.ts#L190)
 
 ___
 
@@ -185,7 +185,7 @@ ___
 
 | Name | Type |
 | :------ | :------ |
-| `messages` | `RNLlamaOAICompatibleMessage`[] |
+| `messages` | [`RNLlamaOAICompatibleMessage`](../README.md#rnllamaoaicompatiblemessage)[] |
 | `template?` | `string` |
 
 #### Returns
@@ -194,13 +194,13 @@ ___
 
 #### Defined in
 
-[index.ts:105](https://github.com/mybigday/llama.rn/blob/68acf1a/src/index.ts#L105)
+[index.ts:131](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/index.ts#L131)
 
 ___
 
 ### loadSession
 
-▸ **loadSession**(`filepath`): `Promise`<`NativeSessionLoadResult`\>
+▸ **loadSession**(`filepath`): `Promise`<[`NativeSessionLoadResult`](../README.md#nativesessionloadresult)\>
 
 Load cached prompt & completion state from a file.
 
@@ -212,11 +212,11 @@ Load cached prompt & completion state from a file.
 
 #### Returns
 
-`Promise`<`NativeSessionLoadResult`\>
+`Promise`<[`NativeSessionLoadResult`](../README.md#nativesessionloadresult)\>
 
 #### Defined in
 
-[index.ts:89](https://github.com/mybigday/llama.rn/blob/68acf1a/src/index.ts#L89)
+[index.ts:115](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/index.ts#L115)
 
 ___
 
@@ -230,7 +230,7 @@ ___
 
 #### Defined in
 
-[index.ts:191](https://github.com/mybigday/llama.rn/blob/68acf1a/src/index.ts#L191)
+[index.ts:217](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/index.ts#L217)
 
 ___
 
@@ -254,7 +254,7 @@ Save current cached prompt & completion state to a file.
 
 #### Defined in
 
-[index.ts:98](https://github.com/mybigday/llama.rn/blob/68acf1a/src/index.ts#L98)
+[index.ts:124](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/index.ts#L124)
 
 ___
 
@@ -268,13 +268,13 @@ ___
 
 #### Defined in
 
-[index.ts:152](https://github.com/mybigday/llama.rn/blob/68acf1a/src/index.ts#L152)
+[index.ts:178](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/index.ts#L178)
 
 ___
 
 ### tokenize
 
-▸ **tokenize**(`text`): `Promise`<`NativeTokenizeResult`\>
+▸ **tokenize**(`text`): `Promise`<[`NativeTokenizeResult`](../README.md#nativetokenizeresult)\>
 
 #### Parameters
 
@@ -284,8 +284,8 @@ ___
 
 #### Returns
 
-`Promise`<`NativeTokenizeResult`\>
+`Promise`<[`NativeTokenizeResult`](../README.md#nativetokenizeresult)\>
 
 #### Defined in
 
-[index.ts:156](https://github.com/mybigday/llama.rn/blob/68acf1a/src/index.ts#L156)
+[index.ts:182](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/index.ts#L182)
diff --git a/docs/API/classes/SchemaGrammarConverter.md b/docs/API/classes/SchemaGrammarConverter.md
index 8753f12..c3aa05f 100644
--- a/docs/API/classes/SchemaGrammarConverter.md
+++ b/docs/API/classes/SchemaGrammarConverter.md
@@ -42,11 +42,11 @@
 | `options` | `Object` |
 | `options.allow_fetch?` | `boolean` |
 | `options.dotall?` | `boolean` |
-| `options.prop_order?` | `PropOrder` |
+| `options.prop_order?` | [`SchemaGrammarConverterPropOrder`](../interfaces/SchemaGrammarConverterPropOrder.md) |
 
 #### Defined in
 
-[grammar.ts:211](https://github.com/mybigday/llama.rn/blob/68acf1a/src/grammar.ts#L211)
+[grammar.ts:213](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/grammar.ts#L213)
 
 ## Properties
 
@@ -56,7 +56,7 @@
 
 #### Defined in
 
-[grammar.ts:201](https://github.com/mybigday/llama.rn/blob/68acf1a/src/grammar.ts#L201)
+[grammar.ts:203](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/grammar.ts#L203)
 
 ___
 
@@ -66,17 +66,17 @@ ___
 
 #### Defined in
 
-[grammar.ts:203](https://github.com/mybigday/llama.rn/blob/68acf1a/src/grammar.ts#L203)
+[grammar.ts:205](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/grammar.ts#L205)
 
 ___
 
 ### \_propOrder
 
-• `Private` **\_propOrder**: `PropOrder`
+• `Private` **\_propOrder**: [`SchemaGrammarConverterPropOrder`](../interfaces/SchemaGrammarConverterPropOrder.md)
 
 #### Defined in
 
-[grammar.ts:199](https://github.com/mybigday/llama.rn/blob/68acf1a/src/grammar.ts#L199)
+[grammar.ts:201](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/grammar.ts#L201)
 
 ___
 
@@ -90,7 +90,7 @@ ___
 
 #### Defined in
 
-[grammar.ts:207](https://github.com/mybigday/llama.rn/blob/68acf1a/src/grammar.ts#L207)
+[grammar.ts:209](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/grammar.ts#L209)
 
 ___
 
@@ -100,7 +100,7 @@ ___
 
 #### Defined in
 
-[grammar.ts:209](https://github.com/mybigday/llama.rn/blob/68acf1a/src/grammar.ts#L209)
+[grammar.ts:211](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/grammar.ts#L211)
 
 ___
 
@@ -114,7 +114,7 @@ ___
 
 #### Defined in
 
-[grammar.ts:205](https://github.com/mybigday/llama.rn/blob/68acf1a/src/grammar.ts#L205)
+[grammar.ts:207](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/grammar.ts#L207)
 
 ## Methods
 
@@ -127,7 +127,7 @@ ___
 | Name | Type |
 | :------ | :------ |
 | `name` | `string` |
-| `rule` | `undefined` \| `BuiltinRule` |
+| `rule` | `undefined` \| [`SchemaGrammarConverterBuiltinRule`](SchemaGrammarConverterBuiltinRule.md) |
 
 #### Returns
 
@@ -135,7 +135,7 @@ ___
 
 #### Defined in
 
-[grammar.ts:693](https://github.com/mybigday/llama.rn/blob/68acf1a/src/grammar.ts#L693)
+[grammar.ts:695](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/grammar.ts#L695)
 
 ___
 
@@ -156,7 +156,7 @@ ___
 
 #### Defined in
 
-[grammar.ts:224](https://github.com/mybigday/llama.rn/blob/68acf1a/src/grammar.ts#L224)
+[grammar.ts:226](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/grammar.ts#L226)
 
 ___
 
@@ -179,7 +179,7 @@ ___
 
 #### Defined in
 
-[grammar.ts:710](https://github.com/mybigday/llama.rn/blob/68acf1a/src/grammar.ts#L710)
+[grammar.ts:712](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/grammar.ts#L712)
 
 ___
 
@@ -200,7 +200,7 @@ ___
 
 #### Defined in
 
-[grammar.ts:312](https://github.com/mybigday/llama.rn/blob/68acf1a/src/grammar.ts#L312)
+[grammar.ts:314](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/grammar.ts#L314)
 
 ___
 
@@ -220,7 +220,7 @@ ___
 
 #### Defined in
 
-[grammar.ts:518](https://github.com/mybigday/llama.rn/blob/68acf1a/src/grammar.ts#L518)
+[grammar.ts:520](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/grammar.ts#L520)
 
 ___
 
@@ -241,7 +241,7 @@ ___
 
 #### Defined in
 
-[grammar.ts:323](https://github.com/mybigday/llama.rn/blob/68acf1a/src/grammar.ts#L323)
+[grammar.ts:325](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/grammar.ts#L325)
 
 ___
 
@@ -255,7 +255,7 @@ ___
 
 #### Defined in
 
-[grammar.ts:813](https://github.com/mybigday/llama.rn/blob/68acf1a/src/grammar.ts#L813)
+[grammar.ts:815](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/grammar.ts#L815)
 
 ___
 
@@ -276,7 +276,7 @@ ___
 
 #### Defined in
 
-[grammar.ts:247](https://github.com/mybigday/llama.rn/blob/68acf1a/src/grammar.ts#L247)
+[grammar.ts:249](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/grammar.ts#L249)
 
 ___
 
@@ -297,4 +297,4 @@ ___
 
 #### Defined in
 
-[grammar.ts:529](https://github.com/mybigday/llama.rn/blob/68acf1a/src/grammar.ts#L529)
+[grammar.ts:531](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/grammar.ts#L531)
diff --git a/docs/API/classes/SchemaGrammarConverterBuiltinRule.md b/docs/API/classes/SchemaGrammarConverterBuiltinRule.md
new file mode 100644
index 0000000..b6b8afc
--- /dev/null
+++ b/docs/API/classes/SchemaGrammarConverterBuiltinRule.md
@@ -0,0 +1,51 @@
+[llama.rn](../README.md) / SchemaGrammarConverterBuiltinRule
+
+# Class: SchemaGrammarConverterBuiltinRule
+
+## Table of contents
+
+### Constructors
+
+- [constructor](SchemaGrammarConverterBuiltinRule.md#constructor)
+
+### Properties
+
+- [content](SchemaGrammarConverterBuiltinRule.md#content)
+- [deps](SchemaGrammarConverterBuiltinRule.md#deps)
+
+## Constructors
+
+### constructor
+
+• **new SchemaGrammarConverterBuiltinRule**(`content`, `deps`)
+
+#### Parameters
+
+| Name | Type |
+| :------ | :------ |
+| `content` | `string` |
+| `deps` | `string`[] |
+
+#### Defined in
+
+[grammar.ts:82](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/grammar.ts#L82)
+
+## Properties
+
+### content
+
+• **content**: `string`
+
+#### Defined in
+
+[grammar.ts:78](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/grammar.ts#L78)
+
+___
+
+### deps
+
+• **deps**: `string`[]
+
+#### Defined in
+
+[grammar.ts:80](https://github.com/mybigday/llama.rn/blob/0c04b5e/src/grammar.ts#L80)
diff --git a/docs/API/interfaces/SchemaGrammarConverterPropOrder.md b/docs/API/interfaces/SchemaGrammarConverterPropOrder.md
new file mode 100644
index 0000000..767b48f
--- /dev/null
+++ b/docs/API/interfaces/SchemaGrammarConverterPropOrder.md
@@ -0,0 +1,7 @@
+[llama.rn](../README.md) / SchemaGrammarConverterPropOrder
+
+# Interface: SchemaGrammarConverterPropOrder
+
+## Indexable
+
+▪ [key: `string`]: `number`
diff --git a/example/src/App.tsx b/example/src/App.tsx
index 4d44a75..2cb4c59 100644
--- a/example/src/App.tsx
+++ b/example/src/App.tsx
@@ -93,7 +93,10 @@ export default function App() {
     console.log(`Model info (took ${Date.now() - t0}ms): `, info)
   }
 
-  const handleInitContext = async (file: DocumentPickerResponse) => {
+  const handleInitContext = async (
+    file: DocumentPickerResponse,
+    loraFile?: DocumentPickerResponse,
+  ) => {
     await handleReleaseContext()
     await getModelInfo(file.uri)
     const msgId = addSystemMessage('Initializing context...')
@@ -103,7 +106,9 @@ export default function App() {
         model: file.uri,
         use_mlock: true,
         n_gpu_layers: Platform.OS === 'ios' ? 99 : 0, // > 0: enable GPU
+
         // embedding: true,
+        lora: loraFile?.uri,
       },
       (progress) => {
         setMessages((msgs) => {
@@ -147,39 +152,48 @@ export default function App() {
       })
   }
 
+  const copyFileIfNeeded = async (
+    type = 'model',
+    file: DocumentPickerResponse,
+  ) => {
+    if (Platform.OS === 'android' && file.uri.startsWith('content://')) {
+      const dir = `${ReactNativeBlobUtil.fs.dirs.CacheDir}/${type}s`
+      const filepath = `${dir}/${file.uri.split('/').pop() || type}.gguf`
+
+      if (!(await ReactNativeBlobUtil.fs.isDir(dir)))
+        await ReactNativeBlobUtil.fs.mkdir(dir)
+
+      if (await ReactNativeBlobUtil.fs.exists(filepath))
+        return { uri: filepath } as DocumentPickerResponse
+
+      await ReactNativeBlobUtil.fs.unlink(dir) // Clean up old files in models
+
+      addSystemMessage(`Copying ${type} to internal storage...`)
+      await ReactNativeBlobUtil.MediaCollection.copyToInternal(
+        file.uri,
+        filepath,
+      )
+      addSystemMessage(`${type} copied!`)
+      return { uri: filepath } as DocumentPickerResponse
+    }
+    return file
+  }
+
   const handlePickModel = async () => {
-    DocumentPicker.pick({
+    const modelRes = await DocumentPicker.pick({
       type: Platform.OS === 'ios' ? 'public.data' : 'application/octet-stream',
-    })
-      .then(async (res) => {
-        let [file] = res
-        if (file) {
-          if (Platform.OS === 'android' && file.uri.startsWith('content://')) {
-            const dir = `${ReactNativeBlobUtil.fs.dirs.CacheDir}/models`
-            if (!(await ReactNativeBlobUtil.fs.isDir(dir)))
-              await ReactNativeBlobUtil.fs.mkdir(dir)
-
-            const filepath = `${dir}/${
-              file.uri.split('/').pop() || 'model'
-            }.gguf`
-            if (await ReactNativeBlobUtil.fs.exists(filepath)) {
-              handleInitContext({ uri: filepath } as DocumentPickerResponse)
-              return
-            } else {
-              await ReactNativeBlobUtil.fs.unlink(dir) // Clean up old files in models
-            }
-            addSystemMessage('Copying model to internal storage...')
-            await ReactNativeBlobUtil.MediaCollection.copyToInternal(
-              file.uri,
-              filepath,
-            )
-            addSystemMessage('Model copied!')
-            file = { uri: filepath } as DocumentPickerResponse
-          }
-          handleInitContext(file)
-        }
-      })
-      .catch((e) => console.log('No file picked, error: ', e.message))
+    }).catch((e) => console.log('No model file picked, error: ', e.message))
+    if (!modelRes?.[0]) return
+    const modelFile = await copyFileIfNeeded('model', modelRes?.[0])
+
+    let loraFile
+    // Example: Apply lora adapter (Currently only select one lora file) (Uncomment to use)
+    // const loraRes = await DocumentPicker.pick({
+    //   type: Platform.OS === 'ios' ? 'public.data' : 'application/octet-stream',
+    // }).catch(e => console.log('No lora file picked, error: ', e.message))
+    // if (loraRes?.[0]) loraFile = await copyFileIfNeeded('lora', loraRes[0])
+
+    handleInitContext(modelFile, loraFile)
   }
 
   const handleSendPress = async (message: MessageType.PartialText) => {
@@ -383,22 +397,32 @@ export default function App() {
         {
           messages: msgs,
           n_predict: 100,
+          grammar,
+          seed: -1,
+          n_probs: 0,
+
+          // Sampling params
+          top_k: 40,
+          top_p: 0.5,
+          min_p: 0.05,
           xtc_probability: 0.5,
           xtc_threshold: 0.1,
+          typical_p: 1.0,
           temperature: 0.7,
-          top_k: 40, // <= 0 to use vocab size
-          top_p: 0.5, // 1.0 = disabled
-          typical_p: 1.0, // 1.0 = disabled
-          penalty_last_n: 256, // 0 = disable penalty, -1 = context size
-          penalty_repeat: 1.18, // 1.0 = disabled
-          penalty_freq: 0.0, // 0.0 = disabled
-          penalty_present: 0.0, // 0.0 = disabled
-          mirostat: 0, // 0/1/2
-          mirostat_tau: 5, // target entropy
-          mirostat_eta: 0.1, // learning rate
-          penalize_nl: false, // penalize newlines
-          seed: -1, // random seed
-          n_probs: 0, // Show probabilities
+          penalty_last_n: 64,
+          penalty_repeat: 1.0,
+          penalty_freq: 0.0,
+          penalty_present: 0.0,
+          dry_multiplier: 0,
+          dry_base: 1.75,
+          dry_allowed_length: 2,
+          dry_penalty_last_n: -1,
+          dry_sequence_breakers: ["\n", ":", "\"", "*"],
+          mirostat: 0,
+          mirostat_tau: 5,
+          mirostat_eta: 0.1,
+          penalize_nl: false,
+          ignore_eos: false,
           stop: [
             '</s>',
             '<|end|>',
@@ -410,12 +434,6 @@ export default function App() {
             '<|end_of_turn|>',
             '<|endoftext|>',
           ],
-          dry_multiplier: 1,
-          dry_base: 1.75,
-          dry_allowed_length: 200,    
-          dry_penalty_last_n: -1,
-          dry_sequence_breakers: ["\n", ":", "\"", "*"],
-          grammar,
           // n_threads: 4,
           // logit_bias: [[15043,1.0]],
         },
diff --git a/ios/RNLlamaContext.mm b/ios/RNLlamaContext.mm
index ac4e601..c2cb593 100644
--- a/ios/RNLlamaContext.mm
+++ b/ios/RNLlamaContext.mm
@@ -292,6 +292,19 @@ - (NSDictionary *)completion:(NSDictionary *)params
     if (params[@"xtc_probability"]) sparams.xtc_probability = [params[@"xtc_probability"] doubleValue];
     if (params[@"typical_p"]) sparams.typ_p = [params[@"typical_p"] doubleValue];
 
+    if (params[@"dry_multiplier"]) sparams.dry_multiplier = [params[@"dry_multiplier"] doubleValue];
+    if (params[@"dry_base"]) sparams.dry_base = [params[@"dry_base"] doubleValue];
+    if (params[@"dry_allowed_length"]) sparams.dry_allowed_length = [params[@"dry_allowed_length"] intValue];
+    if (params[@"dry_penalty_last_n"]) sparams.dry_penalty_last_n = [params[@"dry_penalty_last_n"] intValue];
+
+    // dry break seq
+    if (params[@"dry_sequence_breakers"] && [params[@"dry_sequence_breakers"] isKindOfClass:[NSArray class]]) {
+        NSArray *dry_sequence_breakers = params[@"dry_sequence_breakers"];
+        for (NSString *s in dry_sequence_breakers) {
+            sparams.dry_sequence_breakers.push_back([s UTF8String]);
+        }
+    }
+
     if (params[@"grammar"]) {
         sparams.grammar = [params[@"grammar"] UTF8String];
     }
diff --git a/package.json b/package.json
index e97b8c6..6ac6ec5 100644
--- a/package.json
+++ b/package.json
@@ -1,6 +1,6 @@
 {
   "name": "llama.rn",
-  "version": "0.4.0",
+  "version": "0.4.1",
   "description": "React Native binding of llama.cpp",
   "main": "lib/commonjs/index",
   "module": "lib/module/index",
diff --git a/src/NativeRNLlama.ts b/src/NativeRNLlama.ts
index 04e06e8..5278f39 100644
--- a/src/NativeRNLlama.ts
+++ b/src/NativeRNLlama.ts
@@ -49,37 +49,124 @@ export type NativeContextParams = {
 
 export type NativeCompletionParams = {
   prompt: string
-  grammar?: string
-  stop?: Array<string> // -> antiprompt
-
   n_threads?: number
+  /**
+   * Set grammar for grammar-based sampling.  Default: no grammar
+   */
+  grammar?: string
+  /**
+   * Specify a JSON array of stopping strings.
+   * These words will not be included in the completion, so make sure to add them to the prompt for the next iteration. Default: `[]`
+   */
+  stop?: Array<string>
+  /**
+   * Set the maximum number of tokens to predict when generating text.
+   * **Note:** May exceed the set limit slightly if the last token is a partial multibyte character.
+   * When 0,no tokens will be generated but the prompt is evaluated into the cache. Default: `-1`, where `-1` is infinity.
+   */
   n_predict?: number
+  /**
+   * If greater than 0, the response also contains the probabilities of top N tokens for each generated token given the sampling settings.
+   * Note that for temperature < 0 the tokens are sampled greedily but token probabilities are still being calculated via a simple softmax of the logits without considering any other sampler settings.
+   * Default: `0`
+   */
   n_probs?: number
+  /**
+   * Limit the next token selection to the K most probable tokens.  Default: `40`
+   */
   top_k?: number
+  /**
+   * Limit the next token selection to a subset of tokens with a cumulative probability above a threshold P. Default: `0.95`
+   */
   top_p?: number
+  /**
+   * The minimum probability for a token to be considered, relative to the probability of the most likely token. Default: `0.05`
+   */
   min_p?: number
-  xtc_threshold?: number
+  /**
+   * Set the chance for token removal via XTC sampler. Default: `0.0`, which is disabled.
+   */
   xtc_probability?: number
+  /**
+   * Set a minimum probability threshold for tokens to be removed via XTC sampler. Default: `0.1` (> `0.5` disables XTC)
+   */
+  xtc_threshold?: number
+  /**
+   * Enable locally typical sampling with parameter p. Default: `1.0`, which is disabled.
+   */
   typical_p?: number
-  temperature?: number // -> temp
+  /**
+   * Adjust the randomness of the generated text. Default: `0.8`
+   */
+  temperature?: number
+  /**
+   * Last n tokens to consider for penalizing repetition. Default: `64`, where `0` is disabled and `-1` is ctx-size.
+   */
   penalty_last_n?: number
+  /**
+   * Control the repetition of token sequences in the generated text. Default: `1.0`
+   */
   penalty_repeat?: number
+  /**
+   * Repeat alpha frequency penalty. Default: `0.0`, which is disabled.
+   */
   penalty_freq?: number
+  /**
+   * Repeat alpha presence penalty. Default: `0.0`, which is disabled.
+   */
   penalty_present?: number
+  /**
+   * Penalize newline tokens when applying the repeat penalty. Default: `false`
+   */
+  penalize_nl?: boolean
+  /**
+   * Enable Mirostat sampling, controlling perplexity during text generation. Default: `0`, where `0` is disabled, `1` is Mirostat, and `2` is Mirostat 2.0.
+   */
   mirostat?: number
+  /**
+   * Set the Mirostat target entropy, parameter tau. Default: `5.0`
+   */
   mirostat_tau?: number
+  /**
+   * Set the Mirostat learning rate, parameter eta. Default: `0.1`
+   */
   mirostat_eta?: number
-  penalize_nl?: boolean
-  seed?: number
-
+  /**
+   * Set the DRY (Don't Repeat Yourself) repetition penalty multiplier. Default: `0.0`, which is disabled.
+   */
   dry_multiplier?: number
+  /**
+   * Set the DRY repetition penalty base value. Default: `1.75`
+   */
   dry_base?: number
-  dry_allowed_length?: number    
+  /**
+   * Tokens that extend repetition beyond this receive exponentially increasing penalty: multiplier * base ^ (length of repeating sequence before token - allowed length). Default: `2`
+   */
+  dry_allowed_length?: number
+  /**
+   * How many tokens to scan for repetitions. Default: `-1`, where `0` is disabled and `-1` is context size.
+   */
   dry_penalty_last_n?: number
+  /**
+   * Specify an array of sequence breakers for DRY sampling. Only a JSON array of strings is accepted. Default: `['\n', ':', '"', '*']`
+   */
   dry_sequence_breakers?: Array<string>
-
+  /**
+   * Ignore end of stream token and continue generating. Default: `false`
+   */
   ignore_eos?: boolean
+  /**
+   * Modify the likelihood of a token appearing in the generated text completion.
+   * For example, use `"logit_bias": [[15043,1.0]]` to increase the likelihood of the token 'Hello', or `"logit_bias": [[15043,-1.0]]` to decrease its likelihood.
+   * Setting the value to false, `"logit_bias": [[15043,false]]` ensures that the token `Hello` is never produced. The tokens can also be represented as strings,
+   * e.g.`[["Hello, World!",-0.5]]` will reduce the likelihood of all the individual tokens that represent the string `Hello, World!`, just like the `presence_penalty` does.
+   * Default: `[]`
+   */
   logit_bias?: Array<Array<number>>
+  /**
+   * Set the random number generator (RNG) seed. Default: `-1`, which is a random seed.
+   */
+  seed?: number
 
   emit_partial_completion: boolean
 }
diff --git a/src/grammar.ts b/src/grammar.ts
index 6a845cf..d65d30b 100644
--- a/src/grammar.ts
+++ b/src/grammar.ts
@@ -74,7 +74,7 @@ function buildRepetition(
   return result
 }
 
-class BuiltinRule {
+export class SchemaGrammarConverterBuiltinRule {
   content: string
 
   deps: string[]
@@ -85,9 +85,11 @@ class BuiltinRule {
   }
 }
 
+const BuiltinRule = SchemaGrammarConverterBuiltinRule
+
 const UP_TO_15_DIGITS = buildRepetition('[0-9]', 0, 15)
 
-const PRIMITIVE_RULES: { [key: string]: BuiltinRule } = {
+const PRIMITIVE_RULES: { [key: string]: SchemaGrammarConverterBuiltinRule } = {
   boolean: new BuiltinRule('("true" | "false") space', []),
   'decimal-part': new BuiltinRule(`[0-9] ${UP_TO_15_DIGITS}`, []),
   'integral-part': new BuiltinRule(`[0-9] | [1-9] ${UP_TO_15_DIGITS}`, []),
@@ -126,7 +128,7 @@ const PRIMITIVE_RULES: { [key: string]: BuiltinRule } = {
 }
 
 // TODO: support "uri", "email" string formats
-const STRING_FORMAT_RULES: { [key: string]: BuiltinRule } = {
+const STRING_FORMAT_RULES: { [key: string]: SchemaGrammarConverterBuiltinRule } = {
   date: new BuiltinRule(
     '[0-9] [0-9] [0-9] [0-9] "-" ( "0" [1-9] | "1" [0-2] ) "-" ( "0" [1-9] | [1-2] [0-9] | "3" [0-1] )',
     [],
@@ -173,7 +175,7 @@ const formatLiteral = (literal: string): string => {
 const generateConstantRule = (value: any): string =>
   formatLiteral(JSON.stringify(value))
 
-interface PropOrder {
+export interface SchemaGrammarConverterPropOrder {
   [key: string]: number
 }
 
@@ -196,7 +198,7 @@ function* groupBy(iterable: Iterable<any>, keyFn: (x: any) => any) {
 }
 
 export class SchemaGrammarConverter {
-  private _propOrder: PropOrder
+  private _propOrder: SchemaGrammarConverterPropOrder
 
   private _allowFetch: boolean
 
@@ -209,7 +211,7 @@ export class SchemaGrammarConverter {
   private _refsBeingResolved: Set<string>
 
   constructor(options: {
-    prop_order?: PropOrder
+    prop_order?: SchemaGrammarConverterPropOrder
     allow_fetch?: boolean
     dotall?: boolean
   }) {
@@ -690,7 +692,7 @@ export class SchemaGrammarConverter {
     }
   }
 
-  _addPrimitive(name: string, rule: BuiltinRule | undefined) {
+  _addPrimitive(name: string, rule: SchemaGrammarConverterBuiltinRule | undefined) {
     if (!rule) {
       throw new Error(`Rule ${name} not known`)
     }
@@ -828,7 +830,7 @@ export const convertJsonSchemaToGrammar = ({
   allowFetch,
 }: {
   schema: any
-  propOrder?: PropOrder
+  propOrder?: SchemaGrammarConverterPropOrder
   dotall?: boolean
   allowFetch?: boolean
 }): string | Promise<string> => {
diff --git a/src/index.ts b/src/index.ts
index c77655f..d85d929 100644
--- a/src/index.ts
+++ b/src/index.ts
@@ -11,11 +11,32 @@ import type {
   NativeEmbeddingResult,
   NativeSessionLoadResult,
   NativeEmbeddingParams,
+  NativeCompletionTokenProbItem,
+  NativeCompletionResultTimings,
 } from './NativeRNLlama'
+import type { SchemaGrammarConverterPropOrder, SchemaGrammarConverterBuiltinRule } from './grammar'
 import { SchemaGrammarConverter, convertJsonSchemaToGrammar } from './grammar'
-import type { RNLlamaOAICompatibleMessage } from './chat'
+import type { RNLlamaMessagePart, RNLlamaOAICompatibleMessage } from './chat'
 import { formatChat } from './chat'
 
+export type {
+  NativeContextParams,
+  NativeLlamaContext,
+  NativeCompletionParams,
+  NativeCompletionTokenProb,
+  NativeCompletionResult,
+  NativeTokenizeResult,
+  NativeEmbeddingResult,
+  NativeSessionLoadResult,
+  NativeEmbeddingParams,
+  NativeCompletionTokenProbItem,
+  NativeCompletionResultTimings,
+  RNLlamaMessagePart,
+  RNLlamaOAICompatibleMessage,
+  SchemaGrammarConverterPropOrder,
+  SchemaGrammarConverterBuiltinRule,
+}
+
 export { SchemaGrammarConverter, convertJsonSchemaToGrammar }
 
 const EVENT_ON_INIT_CONTEXT_PROGRESS = '@RNLlama_onInitContextProgress'