Merge pull request #1 from all-in-aigc/feature/support-openai

support openai api

.gitignore

@@ -33,3 +33,5 @@ yarn-error.log*
 # typescript
 *.tsbuildinfo
 next-env.d.ts
+
+pnpm-lock.yaml

README.md

@@ -81,14 +81,27 @@ HF_INFERENCE_ENDPOINT_URL="path to your inference endpoint url"
 
 To run this kind of LLM locally, you can use [TGI](https://github.com/huggingface/text-generation-inference) (Please read [this post](https://github.com/huggingface/text-generation-inference/issues/726) for more information about the licensing).
 
-### Option 3: Fork and modify the code to use a different LLM system
+### Option 3: Use an OpenAI API Key
 
-Another option could be to disable the LLM completely and replace it with another LLM protocol and/or provider (eg. OpenAI, Replicate), or a human-generated story instead (by returning mock or static data).
+This is a new option added recently, where you can use OpenAI API with an OpenAI API Key.
 
+To activate it, create a `.env.local` configuration file:
+
+```bash
+LLM_ENGINE="OPENAI"
+# default openai api base url is: https://api.openai.com/v1
+OPENAI_API_BASE_URL="Your OpenAI API Base URL"
+OPENAI_API_KEY="Your OpenAI API Key"
+OPENAI_API_MODEL="gpt-3.5-turbo"
+```
+
+### Option 4: Fork and modify the code to use a different LLM system
+
+Another option could be to disable the LLM completely and replace it with another LLM protocol and/or provider (eg. Claude, Replicate), or a human-generated story instead (by returning mock or static data).
 
 ### Notes
 
-It is possible that I modify the AI Comic Factory to make it easier in the future (eg. add support for OpenAI or Replicate)
+It is possible that I modify the AI Comic Factory to make it easier in the future (eg. add support for Claude or Replicate)
 
 ## The Rendering API
 

package.json

@@ -43,6 +43,7 @@
     "html2canvas": "^1.4.1",
     "lucide-react": "^0.260.0",
     "next": "13.4.10",
+    "openai": "^4.10.0",
     "pick": "^0.0.1",
     "postcss": "8.4.26",
     "react": "18.2.0",

src/app/queries/predict.ts

@@ -1,15 +1,19 @@
 "use server"
 
-import { LLMEngine } from "@/types"
 import { HfInference, HfInferenceEndpoint } from "@huggingface/inference"
 
+import type { ChatCompletionMessage } from "openai/resources/chat"
+import { LLMEngine } from "@/types"
+import OpenAI from "openai"
+
 const hf = new HfInference(process.env.HF_API_TOKEN)
 
 
 // note: we always try "inference endpoint" first
 const llmEngine = `${process.env.LLM_ENGINE || ""}` as LLMEngine
 const inferenceEndpoint = `${process.env.HF_INFERENCE_ENDPOINT_URL || ""}`
 const inferenceModel = `${process.env.HF_INFERENCE_API_MODEL || ""}`
+const openaiApiKey = `${process.env.OPENAI_API_KEY || ""}`
 
 let hfie: HfInferenceEndpoint
 
@@ -34,6 +38,16 @@ switch (llmEngine) {
       throw new Error(error)
     }
     break;
+
+  case "OPENAI":
+    if (openaiApiKey) {
+      console.log("Using an OpenAI API Key")
+    } else {
+      const error = "No OpenAI API key defined"
+      console.error(error)
+      throw new Error(error)
+    }
+    break;
 
   default:
     const error = "No Inference Endpoint URL or Inference API Model defined"
@@ -45,6 +59,10 @@ export async function predict(inputs: string) {
 
   console.log(`predict: `, inputs)
 
+  if (llmEngine==="OPENAI") {
+    return predictWithOpenAI(inputs)
+  } 
+
   const api = llmEngine ==="INFERENCE_ENDPOINT" ? hfie : hf
 
   let instructions = ""
@@ -92,4 +110,31 @@ export async function predict(inputs: string) {
     .replaceAll("<|assistant|>", "")
     .replaceAll('""', '"')
   )
+}
+
+async function predictWithOpenAI(inputs: string) {
+  const openaiApiBaseUrl = `${process.env.OPENAI_API_BASE_URL || "https://api.openai.com/v1"}`
+  const openaiApiModel = `${process.env.OPENAI_API_MODEL || "gpt-3.5-turbo"}`
+
+  const openai = new OpenAI({
+    apiKey: openaiApiKey,
+    baseURL: openaiApiBaseUrl,
+  })
+
+  const messages: ChatCompletionMessage[] = [
+    { role: "system", content: inputs },
+  ]
+
+  try {
+    const res = await openai.chat.completions.create({
+      messages: messages,
+      stream: false,
+      model: openaiApiModel,
+      temperature: 0.8
+    })
+
+    return res.choices[0].message.content
+  } catch (err) {
+    console.error(`error during generation: ${err}`)
+  }
 }