#187 - add gpt 5 family support by updating litellm and openai packag…

README.md

@@ -132,6 +132,14 @@ const result = await zerox({
 });
 ```
 
+// For GPT-5 models, you can control the reasoning effort:
+// Allowed values: "minimal", "low", "medium", "high"
+// Example:
+// ...
+//   model: ModelOptions.OPENAI_GPT_5,
+//   reasoning_effort: "medium",
+// ...
+
 The `maintainFormat` option tries to return the markdown in a consistent format by passing the output of a prior page in as additional context for the next page. This requires the requests to run synchronously, so it's a lot slower. But valuable if your documents have a lot of tabular data, or frequently have tables that cross pages.
 
 ```
@@ -213,13 +221,19 @@ Zerox supports a wide range of models across different providers:
   - GPT-4 Vision Mini (gpt-4o-mini)
   - GPT-4.1 (gpt-4.1)
   - GPT-4.1 Mini (gpt-4.1-mini)
+  - GPT-5 (gpt-5)
+  - GPT-5 Mini (gpt-5-mini)
+  - GPT-5 Nano (gpt-5-nano)
 
 - **OpenAI**
 
   - GPT-4 Vision (gpt-4o)
   - GPT-4 Vision Mini (gpt-4o-mini)
   - GPT-4.1 (gpt-4.1)
   - GPT-4.1 Mini (gpt-4.1-mini)
+  - GPT-5 (gpt-5)
+  - GPT-5 Mini (gpt-5-mini)
+  - GPT-5 Nano (gpt-5-nano)  
 
 - **AWS Bedrock**
 
@@ -380,7 +394,6 @@ print(result)
 
 ### Parameters
 
-```python
 async def zerox(
     cleanup: bool = True,
     concurrency: int = 10,
@@ -396,6 +409,12 @@ async def zerox(
   ...
 ```
 
+# For GPT-5 models, you can control the reasoning effort:
+# Allowed values: "minimal", "low", "medium", "high"
+# Example:
+# result = await zerox(file_path=..., model="gpt-5", reasoning_effort="medium")
+
+
 Parameters
 
 - **cleanup** (bool, optional):
@@ -417,6 +436,8 @@ Parameters
   The system prompt to use for the model, this overrides the default system prompt of Zerox.Generally it is not required unless you want some specific behavior. Defaults to None.
 - **select_pages** (Optional[Union[int, Iterable[int]]], optional):
   Pages to process, can be a single page number or an iterable of page numbers. Defaults to None
+- **reasoning_effort** (str, optional, GPT-5 only):
+  Controls the reasoning effort for GPT-5 models. Allowed values: "minimal", "low", "medium", "high". Defaults to None.
 - **kwargs** (dict, optional):
   Additional keyword arguments to pass to the litellm.completion method.
   Refer to the LiteLLM Documentation and Completion Input for details.

node-zerox/src/index.ts

@@ -64,6 +64,7 @@ export const zerox = async ({
   imageFormat = "png",
   imageHeight,
   llmParams = {},
+  reasoning_effort,
   maintainFormat = false,
   maxImageSize = 15,
   maxRetries = 1,
@@ -79,6 +80,10 @@ export const zerox = async ({
   trimEdges = true,
 }: ZeroxArgs): Promise<ZeroxOutput> => {
   let extracted: Record<string, unknown> | null = null;
+  // If reasoning is provided, add to llmParams
+  if (reasoning_effort) {
+    llmParams = { ...llmParams, reasoning_effort };
+  }
   let extractedLogprobs: LogprobPage[] = [];
   let inputTokenCount: number = 0;
   let outputTokenCount: number = 0;

node-zerox/src/models/openAI.ts

@@ -96,7 +96,7 @@ export default class OpenAIModel implements ModelInterface {
     priorPage,
     prompt,
   }: CompletionArgs): Promise<CompletionResponse> {
-    const systemPrompt = prompt || SYSTEM_PROMPT_BASE;
+  const systemPrompt = prompt || SYSTEM_PROMPT_BASE;
 
     // Default system message
     const messages: any = [{ role: "system", content: systemPrompt }];
@@ -120,13 +120,18 @@ export default class OpenAIModel implements ModelInterface {
     messages.push({ role: "user", content: imageContents });
 
     try {
+      // If model is GPT-5 and reasoning_effort is provided, add it to payload
+      let payload: any = {
+        messages,
+        model: this.model,
+        ...convertKeysToSnakeCase(this.llmParams ?? null),
+      };
+      if (this.model && this.model.startsWith("gpt-5") && this.llmParams?.reasoning_effort) {
+        payload.reasoning_effort = this.llmParams.reasoning_effort;
+      }
       const response = await axios.post(
         "https://api.openai.com/v1/chat/completions",
-        {
-          messages,
-          model: this.model,
-          ...convertKeysToSnakeCase(this.llmParams ?? null),
-        },
+        payload,
         {
           headers: {
             Authorization: `Bearer ${this.apiKey}`,

node-zerox/src/types.ts

@@ -28,6 +28,7 @@ export interface ZeroxArgs {
   imageHeight?: number;
   imageFormat?: "png" | "jpeg";
   llmParams?: Partial<LLMParams>;
+  reasoning_effort?: "minimal" | "low" | "medium" | "high";
   maintainFormat?: boolean;
   maxImageSize?: number;
   maxRetries?: number;
@@ -95,6 +96,11 @@ export enum ModelOptions {
   OPENAI_GPT_4O = "gpt-4o",
   OPENAI_GPT_4O_MINI = "gpt-4o-mini",
 
+  // OpenAI GPT-5 Models
+  OPENAI_GPT_5 = "gpt-5",
+  OPENAI_GPT_5_MINI = "gpt-5-mini",
+  OPENAI_GPT_5_NANO = "gpt-5-nano",
+
   // Google Gemini Models
   GOOGLE_GEMINI_1_5_FLASH = "gemini-1.5-flash",
   GOOGLE_GEMINI_1_5_FLASH_8B = "gemini-1.5-flash-8b",
@@ -222,6 +228,7 @@ export interface GoogleLLMParams extends BaseLLMParams {
 export interface OpenAILLMParams extends BaseLLMParams {
   logprobs: boolean;
   maxTokens: number;
+  reasoning_effort?: "minimal" | "low" | "medium" | "high";
 }
 
 // Union type of all provider params

package.json

@@ -23,7 +23,7 @@
     "heic-convert": "^2.1.0",
     "libreoffice-convert": "^1.6.0",
     "mime-types": "^2.1.35",
-    "openai": "^4.82.0",
+    "openai": "^5.15.0",
     "os": "^0.1.2",
     "p-limit": "^3.1.0",
     "path": "^0.12.7",