Merge pull request #18 from neodon/openai-compatible-configs

add configuration for OpenAI-compatible embeddings providers

Author: nullure

.env.example

@@ -54,6 +54,15 @@ OM_ADV_EMBED_PARALLEL=false
 # Delay between embeddings in milliseconds
 OM_EMBED_DELAY_MS=200
 
+# OpenAI-compatible Embeddings Provider
+# OM_OPENAI_BASE_URL=https://api.openai.com/v1
+# Model override for all sector embeddings (leave empty to use defaults)
+# OM_OPENAI_MODEL=text-embedding-qwen3-embedding-4b
+
+# API Configuration
+# Max request body size in bytes (default: 1MB)
+OM_MAX_PAYLOAD_SIZE=1000000
+
 # --------------------------------------------
 # Embedding Provider API Keys
 # --------------------------------------------

backend/src/config/index.ts

@@ -12,12 +12,15 @@ export const env = {
     adv_embed_parallel: process.env.OM_ADV_EMBED_PARALLEL === 'true',
     embed_delay_ms: Number(process.env.OM_EMBED_DELAY_MS) || 200,
     openai_key: process.env.OPENAI_API_KEY || process.env.OM_OPENAI_API_KEY || '',
+    openai_base_url: process.env.OM_OPENAI_BASE_URL || 'https://api.openai.com/v1',
+    openai_model: process.env.OM_OPENAI_MODEL || undefined,
     gemini_key: process.env.GEMINI_API_KEY || process.env.OM_GEMINI_API_KEY || '',
     ollama_url: process.env.OLLAMA_URL || process.env.OM_OLLAMA_URL || 'http://localhost:11434',
     local_model_path: process.env.LOCAL_MODEL_PATH || process.env.OM_LOCAL_MODEL_PATH || '',
     vec_dim: Number(process.env.OM_VEC_DIM) || 768,
     min_score: Number(process.env.OM_MIN_SCORE) || 0.3,
     decay_lambda: Number(process.env.OM_DECAY_LAMBDA) || 0.02,
+    max_payload_size: Number(process.env.OM_MAX_PAYLOAD_SIZE) || 1_000_000,
     mode: (process.env.OM_MODE || 'standard').toLowerCase(),
     lg_namespace: process.env.OM_LG_NAMESPACE || 'default',
     lg_max_context: Number(process.env.OM_LG_MAX_CONTEXT) || 50,

backend/src/embedding/index.ts

@@ -39,17 +39,18 @@ export async function embedForSector(text: string, sector: string): Promise<numb
 async function embedWithOpenAI(text: string, sector: string): Promise<number[]> {
     if (!env.openai_key) throw new Error('OpenAI API key not configured')
 
-    const modelMap: Record<string, string> = {
+    const defaultModelMap: Record<string, string> = {
         episodic: 'text-embedding-3-small',
         semantic: 'text-embedding-3-small',
         procedural: 'text-embedding-3-small',
         emotional: 'text-embedding-3-small',
         reflective: 'text-embedding-3-large'
     }
 
-    const model = modelMap[sector] || 'text-embedding-3-small'
+    const model = env.openai_model || defaultModelMap[sector] || 'text-embedding-3-small'
 
-    const response = await fetch('https://api.openai.com/v1/embeddings', {
+    const baseUrl = env.openai_base_url.replace(/\/$/, '')
+    const response = await fetch(`${baseUrl}/embeddings`, {
         method: 'POST',
         headers: {
             'content-type': 'application/json',
@@ -58,7 +59,7 @@ async function embedWithOpenAI(text: string, sector: string): Promise<number[]>
         body: JSON.stringify({
             input: text,
             model,
-            dimensions: env.vec_dim <= 1536 ? env.vec_dim : undefined
+            dimensions: env.vec_dim
         })
     })
 
@@ -73,11 +74,12 @@ async function embedWithOpenAI(text: string, sector: string): Promise<number[]>
 async function embedBatchOpenAI(texts: Record<string, string>): Promise<Record<string, number[]>> {
     if (!env.openai_key) throw new Error('OpenAI API key not configured')
 
-    const model = 'text-embedding-3-small'
+    const model = env.openai_model || 'text-embedding-3-small'
     const inputTexts = Object.values(texts)
     const sectors = Object.keys(texts)
 
-    const response = await fetch('https://api.openai.com/v1/embeddings', {
+    const baseUrl = env.openai_base_url.replace(/\/$/, '')
+    const response = await fetch(`${baseUrl}/embeddings`, {
         method: 'POST',
         headers: {
             'content-type': 'application/json',
@@ -86,7 +88,7 @@ async function embedBatchOpenAI(texts: Record<string, string>): Promise<Record<s
         body: JSON.stringify({
             input: inputTexts,
             model,
-            dimensions: env.vec_dim <= 1536 ? env.vec_dim : undefined
+            dimensions: env.vec_dim
         })
     })
 
@@ -492,13 +494,15 @@ export function getEmbeddingInfo(): Record<string, any> {
     switch (env.emb_kind) {
         case 'openai':
             info.configured = !!env.openai_key
+            info.base_url = env.openai_base_url
+            info.model_override = env.openai_model || null
             info.batch_api = env.embed_mode === 'simple'
             info.models = {
-                episodic: 'text-embedding-3-small',
-                semantic: 'text-embedding-3-small',
-                procedural: 'text-embedding-3-small',
-                emotional: 'text-embedding-3-small',
-                reflective: 'text-embedding-3-large'
+                episodic: env.openai_model || 'text-embedding-3-small',
+                semantic: env.openai_model || 'text-embedding-3-small',
+                procedural: env.openai_model || 'text-embedding-3-small',
+                emotional: env.openai_model || 'text-embedding-3-small',
+                reflective: env.openai_model || 'text-embedding-3-large'
             }
             break
         case 'gemini':

backend/src/server/index.ts

@@ -23,7 +23,7 @@ import type {
     ingest_url_req
 } from '../types'
 
-const app = server()
+const app = server({ max_payload_size: env.max_payload_size })
 
 app.use((req: any, res: any, next: any) => {
     res.setHeader('Access-Control-Allow-Origin', '*')

backend/src/server/server.js

@@ -26,7 +26,7 @@ const path = require('path');
 const http = require('http');
 const WebSocket = require('ws');
 const { parse } = require('url');
-function server() {
+function server(config = {}) {
     const ROUTES = [];
     const WARES = [];
     const WS_ROUTES = [];
@@ -159,7 +159,7 @@ function server() {
     use((req, res, next) => {
         if (req.headers['content-type']?.includes('application/json')) {
             let d = '';
-            let max = 1e6;
+            let max = config.max_payload_size || 1_000_000;
             req.on('data', e => {
                 d += e;
                 if (d.length > max) {