Feature request: other API

[Yp-pro]

Hello,

Thanks for your job! Amazing idea! But I don't have OpenAI API, I use Gemini. It'd be great to add an option to use any other APIs instead of OpenAI's one.

Thank you.

[m-damien]

Thanks! This would be a good modification and it would also future-proof the system. All prompts are going through the function "executePrompt" in https://github.com/m-damien/Textoshop/blob/main/src/model/Model.tsx so just changing this function should be enough to support other APIs.

PRs are welcome if someone wants to tackle this!

[0x4E69676874466F78]

For openai-compatible api (e.g. llama.cpp) you can:

diff --git a/src/model/Model.tsx b/src/model/Model.tsx
index 9f661a0..bb365ed 100644
--- a/src/model/Model.tsx
+++ b/src/model/Model.tsx
@@ -30,6 +30,10 @@ const params = new URLSearchParams(search);
 const key = params.get('k');
 
 let openaiKey = ""
+let openaiBaseURL = ""
+if ("VITE_OPENAI_BASE_URL" in import.meta.env) {
+    openaiBaseURL = import.meta.env.VITE_OPENAI_BASE_URL;
+}
 if (!key) {
     if ("VITE_OPENAI_API_KEY" in import.meta.env) {
         openaiKey = import.meta.env.VITE_OPENAI_API_KEY;
@@ -41,6 +45,7 @@ if (!key) {
 }
 
 export const openai = new OpenAI({
+    baseURL: openaiBaseURL,
     apiKey: openaiKey,
     dangerouslyAllowBrowser: true
 });

and set environment variable "VITE_OPENAI_BASE_URL" like
Environment="VITE_OPENAI_BASE_URL=http://localhost:8080/v1" (if you used systemd service for example)
or
export VITE_OPENAI_BASE_URL=http://localhost:8080/v1 before npm run dev

[0x4E69676874466F78]

Also, I forgot to say, model name is hardcoded:

Textoshop/src/model/tools/promptTools/TextResizer.tsx

Line 85 in b1b4b6f

model: 'gpt-4o-mini-2024-07-18',

You need to either edit it or set an alias in llama-swap.