Custom API (GWDG), Embeddings & Upload Issues

[Teclablue]

What happened?

Hi LibreChat team,

we're running LibreChat in a Docker setup and use a custom AI backend via the GWDG Academic Cloud (Germany). We're not using OpenAI but a compatible API. Here's our stack:
🔧 Setup

LibreChat Version: dev-current 

API Endpoint: https://chat-ai.academiccloud.de/v1 and /v2 (custom OpenAI-compatible API by GWDG)

Embeddings: Mistral (e5-mistral-7b-instruct) via custom endpoint

RAG: enabled in .env using RAG_OPENAI_API_KEY, RAG_OPENAI_BASEURL

Web search APIs: Serper, Firecrawl, Jina — all keys provided

What works

Login with OpenID (Microsoft Entra ID)

Image upload & vision-based interaction (e.g., describing a photo)

Custom AI models from GWDG are visible and working (we use OpenAI ChatGPT but provided by GWDG not OpenAI) 

Librechat is connected to Completion and RAG (logs say it works)

Issues

1. RAG / File Upload

Despite setting the RAG_* environment variables correctly, only image uploads work.
Uploading PDFs or CSVs is not possible. Only an "upload pictures" buttons appear.

2. Embeddings Model appears in UI

We do not want the e5-mistral-7b-instruct embedding model to appear as a selectable chat model.
This is likely because the model is listed under embedding_models or not filtered out properly.
How can we prevent this?

3. Web Search doesn't work
   Although WEBSEARCH_ENABLED=true and the three keys are provided, web search fails silently or doesn't return anything.

4. We also defined multiple custom endpoints in librechat.yaml for the GWDG models, all working fine. But adding a rag: section in librechat.yaml breaks all model loading.

There seems to be no place in the .env or librechat.yaml where these are set. Any advice?

ENV:
====================================================================#

LibreChat Configuration

#=====================================================================#

Please refer to the reference documentation for assistance

with configuring your LibreChat environment.

https://www.librechat.ai/docs/configuration/dotenv

#=====================================================================#

VITE_CUSTOM_LOGIN_LOGO=/ZALF_Logo_DE.png

#==================================================#

Server Configuration

#==================================================#
CHAT_AI_API_KEY=xxx
HOST=xxx
PORT=xxx

MONGO_URI=xxx

DOMAIN_CLIENT=xxx
DOMAIN_SERVER=hxxx

NO_INDEX=true

Use the address that is at most n number of hops away from the Express application.

req.socket.remoteAddress is the first hop, and the rest are looked for in the X-Forwarded-For header from right to left.

A value of 0 means that the first untrusted address would be req.socket.remoteAddress, i.e. there is no reverse proxy.

Defaulted to 1.

TRUST_PROXY=1

#===============#

JSON Logging

#===============#

Use when process console logs in cloud deployment like GCP/AWS

CONSOLE_JSON=false

#===============#

Debug Logging

#===============#

DEBUG_LOGGING=true
DEBUG_CONSOLE=true

#=============#

Permissions

#=============#

UID=0
GID=0

#===============#

Configuration

#===============#

Use an absolute path, a relative path, or a URL

CONFIG_PATH="/alternative/path/to/librechat.yaml"

#===================================================#

Endpoints

#===================================================#

ENDPOINTS=custom

PROXY=

#============#

GWDG

#============#

CHAT_AI_API_KEY=xxx
OPENAI_BASE_URL=https://chat-ai.academiccloud.de/v1
DEBUG_OPENAI=true

EMBEDDING_API_KEY=xxx
EMBEDDINGS_PROVIDER=custom # ← NEU (falls nicht vorhanden)
EMBEDDINGS_MODEL=e5-mistral-7b-instruct # ← NEU (GWDG-Embeddings-Modell)
EMBEDDING_BASE_URL=https://chat-ai.academiccloud.de/v2
FILE_UPLOAD_MAX_SIZE=20 # ← optional größer als 10 MB

#============#

Plugins

#============#

PLUGIN_MODELS=gpt-4o,gpt-4o-mini,gpt-4,gpt-4-turbo-preview,gpt-4-0125-preview,gpt-4-1106-preview,gpt-4-0613,gpt-3.5-turbo,gpt-3.5-turbo-0125,gpt-3.5-turbo-1106,gpt-3.5-turbo-0613

DEBUG_PLUGINS=true

CREDS_KEY=xxx
CREDS_IV=xxx

#==================================================#

Search

#==================================================#

SEARCH=true
MEILI_NO_ANALYTICS=true
MEILI_HOST=http://0.0.0.0:7700
MEILI_MASTER_KEY=xxx

#==================================================#

RAG

#==================================================#

RAG_OPENAI_API_KEY=${EMBEDDING_API_KEY}
RAG_OPENAI_BASEURL=${EMBEDDING_BASE_URL}
EMBEDDINGS_PROVIDER=openai # so heißt die Variablen-Gruppe intern
EMBEDDINGS_MODEL=${EMBEDDINGS_MODEL}

#===================================================#

User System

#===================================================#

#========================#

Moderation

#========================#

OPENAI_MODERATION=false
OPENAI_MODERATION_API_KEY=

OPENAI_MODERATION_REVERSE_PROXY=

BAN_VIOLATIONS=true
BAN_DURATION=1000 * 60 * 60 * 2
BAN_INTERVAL=20

LOGIN_VIOLATION_SCORE=1
REGISTRATION_VIOLATION_SCORE=1
CONCURRENT_VIOLATION_SCORE=1
MESSAGE_VIOLATION_SCORE=1
NON_BROWSER_VIOLATION_SCORE=20

LOGIN_MAX=7
LOGIN_WINDOW=5
REGISTER_MAX=5
REGISTER_WINDOW=60

LIMIT_CONCURRENT_MESSAGES=true
CONCURRENT_MESSAGE_MAX=2

LIMIT_MESSAGE_IP=true
MESSAGE_IP_MAX=40
MESSAGE_IP_WINDOW=1

LIMIT_MESSAGE_USER=false
MESSAGE_USER_MAX=40
MESSAGE_USER_WINDOW=1

ILLEGAL_MODEL_REQ_SCORE=5

#========================#

Balance

#========================#

CHECK_BALANCE=false

START_BALANCE=20000 # note: the number of tokens that will be credited after registration.

#========================#

Registration and Login

#========================#

ALLOW_EMAIL_LOGIN=true
ALLOW_REGISTRATION=true
ALLOW_SOCIAL_LOGIN=true
ALLOW_SOCIAL_REGISTRATION=true
ALLOW_PASSWORD_RESET=false

ALLOW_ACCOUNT_DELETION=true # note: enabled by default if omitted/commented out

ALLOW_UNVERIFIED_EMAIL_LOGIN=true

SESSION_EXPIRY=1000 * 60 * 15
REFRESH_TOKEN_EXPIRY=(1000 * 60 * 60 * 24) * 7

JWT_SECRET=xxx
JWT_REFRESH_SECRET=xxx

OpenID / Entra ID

OPENID_ENABLED=true
OPENID_ISSUER=xxx
OPENID_CLIENT_ID=xx
OPENID_CLIENT_SECRET=xxx
OPENID_REDIRECT_URI=xxx
OPENID_SCOPES="openid profile email"

Set to determine which user info property returned from OpenID Provider to store as the User's username

OPENID_USERNAME_CLAIM=upn

Set to determine which user info property returned from OpenID Provider to store as the User's name

OPENID_NAME_CLAIM=name

OPENID_BUTTON_LABEL="Login mit Microsoft"
OPENID_IMAGE_URL="https://cdn-icons-png.flaticon.com/512/732/732221.png" # Beispiel-Icon

Set to true to automatically redirect to the OpenID provider when a user visits the login page

This will bypass the login form completely for users, only use this if OpenID is your only authentication method

OPENID_AUTO_REDIRECT=false

Set to true to use PKCE (Proof Key for Code Exchange) for OpenID authentication

OPENID_USE_PKCE=false
#Set to true to reuse openid tokens for authentication management instead of using the mongodb session and the custom refresh token.
OPENID_REUSE_TOKENS=false
#By default, signing key verification results are cached in order to prevent excessive HTTP requests to the JWKS endpoint.
#If a signing key matching the kid is found, this will be cached and the next time this kid is requested the signing key will be served from the cache.
#Default is true.
OPENID_JWKS_URL_CACHE_ENABLED=true
OPENID_JWKS_URL_CACHE_TIME=600000 # ms eq to 10 minutes leave empty to disable caching
#Set to true to trigger token exchange flow to acquire access token for the userinfo endpoint.
OPENID_ON_BEHALF_FLOW_FOR_USERINFRO_REQUIRED=
OPENID_ON_BEHALF_FLOW_USERINFRO_SCOPE = "user.read" # example for Scope Needed for Microsoft Graph API

Set to true to use the OpenID Connect end session endpoint for logout

OPENID_USE_END_SESSION_ENDPOINT=

VITE_OPENID_ENABLED=true
VITE_OPENID_BUTTON_LABEL="Login mit Microsoft"
VITE_OPENID_IMAGE_URL="https://cdn-icons-png.flaticon.com/512/732/732221.png"
#========================#

Email Password Reset

#========================#

EMAIL_SERVICE=
EMAIL_HOST=
EMAIL_PORT=25
EMAIL_ENCRYPTION=
EMAIL_ENCRYPTION_HOSTNAME=
EMAIL_ALLOW_SELFSIGNED=
EMAIL_USERNAME=
EMAIL_PASSWORD=
EMAIL_FROM_NAME=
EMAIL_FROM=[email protected]

#========================#

Shared Links

#========================#

ALLOW_SHARED_LINKS=true
ALLOW_SHARED_LINKS_PUBLIC=true

#===================================================#

UI

#===================================================#

APP_TITLE=LibreChat KiGate ZALF e.V.
CUSTOM_FOOTER="My custom footer"
HELP_AND_FAQ_URL=https://librechat.ai

SHOW_BIRTHDAY_ICON=true

Google tag manager id

#ANALYTICS_GTM_ID=user provided google tag manager id

#===============#

REDIS Options

#===============#

REDIS_URI=xxx

USE_REDIS=true

USE_REDIS_CLUSTER=true

REDIS_CA=/path/to/ca.crt

#======================#

Web Search

#======================#

WEBSEARCH_ENABLED=true
WEBSEARCH_PROVIDER=serper

Search Provider (Required)

SERPER_API_KEY="xxx"

Scraper (Required)

FIRECRAWL_API_KEY="xxx"

Optional: Custom Firecrawl API URL

FIRECRAWL_API_URL=your_firecrawl_api_url

Reranker (Required)

JINA_API_KEY="xxx"

Librechat.YAML:

-------------------------------------------------

LibreChat Haupt­konfiguration

-------------------------------------------------

version: 1.2.1
cache: true

-------------------------------------------------

Endpoints

-------------------------------------------------

endpoints:

1) Custom-Endpoints (GWDG /v2 – „inference-service“-Header)

custom:
- name: Chat AI openai-gpt41
apiKey: ${CHAT_AI_API_KEY}
baseURL: https://chat-ai.academiccloud.de/v2
headers: { inference-service: openai-gpt41 }
models: { default: [openai-gpt41], fetch: false }
stream: false
vision: false
titleConvo: true
titleModel: current_model

- name: Chat AI openai-gpt41-mini
  apiKey: ${CHAT_AI_API_KEY}
  baseURL: https://chat-ai.academiccloud.de/v2
  headers: { inference-service: openai-gpt41-mini }
  models:  { default: [openai-gpt41-mini], fetch: false }
  stream: false
  vision: false
  titleConvo: true
  titleModel: current_model

- name: Chat AI openai-gpt4o-mini
  apiKey: ${CHAT_AI_API_KEY}
  baseURL: https://chat-ai.academiccloud.de/v2
  headers: { inference-service: openai-gpt4o-mini }
  models:  { default: [openai-gpt4o-mini], fetch: false }
  stream: false
  vision: false
  titleConvo: true
  titleModel: current_model

- name: Chat AI openai-gpt4o
  apiKey: ${CHAT_AI_API_KEY}
  baseURL: https://chat-ai.academiccloud.de/v2
  headers: { inference-service: openai-gpt4o }
  models:  { default: [openai-gpt4o], fetch: false }
  stream: false
  vision: false
  titleConvo: true
  titleModel: current_model

- name: Chat AI openai-o3
  apiKey: ${CHAT_AI_API_KEY}
  baseURL: https://chat-ai.academiccloud.de/v2
  headers: { inference-service: openai-o3 }
  models:  { default: [openai-o3], fetch: false }
  stream: false
  vision: false
  titleConvo: true
  titleModel: current_model

- name: Chat AI openai-o1-mini
  apiKey: ${CHAT_AI_API_KEY}
  baseURL: https://chat-ai.academiccloud.de/v2
  headers: { inference-service: openai-o1-mini }
  models:  { default: [openai-o1-mini], fetch: false }
  vision: false
  stream: false
  titleConvo: true
  titleModel: current_model

- name: Chat AI openai-o3-mini
  apiKey: ${CHAT_AI_API_KEY}
  baseURL: https://chat-ai.academiccloud.de/v2
  headers: { inference-service: openai-o3-mini }
  models:  { default: [openai-o3-mini], fetch: false }
  vision: false
  stream: false
  titleConvo: true
  titleModel: current_model

- name: Chat AI openai-o4-mini
  apiKey: ${CHAT_AI_API_KEY}
  baseURL: https://chat-ai.academiccloud.de/v2
  headers: { inference-service: openai-o4-mini }
  models:  { default: [openai-o4-mini], fetch: false }
  vision: false
  stream: false
  titleConvo: true
  titleModel: current_model

-------------------------------------------------

Interface-Optionen (unverändert zu vorher)

-------------------------------------------------

interface:
customWelcome: "Willkommen auf ZALF KI / Welcome to ZALF AI"
privacyPolicy: { externalUrl: https://librechat.ai/privacy-policy, openNewTab: true }
termsOfService:
externalUrl: https://librechat.ai/tos
openNewTab: true
modalAcceptance: true
modalTitle: "Terms of Service for LibreChat"
modalContent: |
(… Text gekürzt …)
endpointsMenu: true
modelSelect: true
parameters: true
sidePanel: true
presets: true
prompts: true
bookmarks: true
multiConvo: true
agents: true

-------------------------------------------------

registration:
socialLogins: [github, google, discord, openid, facebook, apple]

actions:
allowedDomains: [swapi.dev, librechat.ai, google.com]

fileConfig:
serverFileSizeLimit: 50 # maximale Größe pro Anfrage (MB)
endpoints:
default:
fileLimit: 5 # max. 5 Dateien pro Nachricht
fileSizeLimit: 20 # max. 20 MB pro Datei
totalSizeLimit: 50 # max. 50 MB gesamt
supportedMimeTypes:
- application/pdf
- text/plain
- text/csv
- application/json

Docker compose:

version: "3.9"

services:
############################################################

LibreChat-Backend

############################################################
api:
container_name: librechat
# Wenn du lieber das offizielle Image nutzen willst, auskommentieren:
image: ghcr.io/danny-avila/librechat-dev:latest

build:

# context: .
# dockerfile: Dockerfile
user: "${UID}:${GID}"
restart: unless-stopped
depends_on:
  - mongodb
  - rag_api
ports:
  - "${PORT:-3080}:${PORT:-3080}"
env_file:
  - .env                    # holt API-Key, PORT, RAG_* usw.
environment:               # NUR service-spezifische Werte
  HOST: 0.0.0.0
  MONGO_URI: xxx
  MEILI_HOST: http://meilisearch:7700
  RAG_API_URL: http://rag_api:${RAG_PORT:-8000}
  OPENAI_API_KEY: ${CHAT_AI_API_KEY}
  OPENAI_BASE_URL: ${OPENAI_BASE_URL}

volumes:
  - ./images:/app/client/public/images
  - ./uploads:/app/uploads
  - ./logs:/app/api/logs
  - ./librechat.yaml:/app/librechat.yaml
extra_hosts:
  - "host.docker.internal:host-gateway"

############################################################

Datenbanken

############################################################
mongodb:
container_name: chat-mongodb
image: mongo
restart: unless-stopped
user: "${UID}:${GID}"
volumes:
- ./data-node:/data/db
command: ["mongod", "--noauth"]

meilisearch:
container_name: chat-meilisearch
image: getmeili/meilisearch:v1.12.3
restart: unless-stopped
user: "${UID}:${GID}"
env_file:
- .env # liest MEILI_MASTER_KEY
environment:
MEILI_NO_ANALYTICS: "true"
volumes:
- ./meili_data_v1.12:/meili_data

vectordb:
container_name: vectordb
image: ankane/pgvector:latest
restart: unless-stopped
environment:
POSTGRES_DB: xxx
POSTGRES_USER: xxx
POSTGRES_PASSWORD: xxx
volumes:
- pgdata2:/var/lib/postgresql/data

############################################################

RAG-API (volle Version, kann GWDG/OpenAI ansprechen)

############################################################
rag_api:
container_name: rag_api
image: ghcr.io/danny-avila/librechat-rag-api-dev:latest
restart: unless-stopped
depends_on:
- vectordb

env_file:
  - .env               # alle anderen Variablen holt er hier

environment:          # <– ab hier KEINE Bindestriche!
  DB_HOST: vectordb
  RAG_PORT: ${RAG_PORT:-8000}
  DB_PORT: 5432
  DB_NAME: xxx
  DB_USERNAME: xx
  DB_PASSWORD: x


  # Schlüssel & Basis-URL fürs Embedding
  OPENAI_API_KEY: ${EMBEDDING_API_KEY}
  OPENAI_BASE_URL: ${OPENAI_BASE_URL}

  # gewähltes Embedding-Modell
  OPENAI_EMBEDDING_MODEL: e5-mistral-7b-instruct
  OPENAI_EMBEDDING_ENCODING: text
  OPENAI_EXTRA_HEADERS: '{"inference-service":"e5-mistral-7b-instruct"}'

volumes:
pgdata2:

HELP?!?!
Thank you in advance!!!
Best Teclablue

Version Information

root@srv-pr-ki2:/home/sos/LibreChat# docker images | grep librechat
ghcr.io/danny-avila/librechat-dev latest 7ffcb513e668 4 days ago 1.11GB
librechat_api latest 554be69c408a 4 days ago 972MB
ghcr.io/danny-avila/librechat v0.7.9-rc1 3d0aff522ece 5 days ago 1.11GB
ghcr.io/danny-avila/librechat-rag-api-dev latest eb981aff60a5 2 weeks ago 7.86GB
ghcr.io/danny-avila/librechat-dev 4af29672793f 7 weeks ago 1.03GB
ghcr.io/danny-avila/librechat latest 17cfdb1df083 2 months ago 1.02GB
ghcr.io/danny-avila/librechat-rag-api-dev-lite latest e7da8e6aca0f 2 months ago 1.31GB
ghcr.io/danny-avila/librechat v0.7.1 d16209914198 14 months ago 1.42GB

Steps to Reproduce

see above

What browsers are you seeing the problem on?

Firefox

Relevant log output

Which logs are needed?

Screenshots

No response

Code of Conduct

• I agree to follow this project's Code of Conduct

[onestardao]

yo just saw this — brutal combo. actually helped someone last week run into exactly the same issues (custom backend, weird ENV stack, embedding fallback logic silently skipped…)

the real problem? LibreChat doesn’t enforce a semantic-safe bootstrap order.

so what happens is:

upload module tries to bind before RAG port is stable

embeddings check env first, but model not exposed = silent failure

fallback logic doesn’t trigger because DISABLE_RAG isn’t granular enough

boom, broken chain, no one talks to anyone, no logs

we ran into this mess so often that we just built a semantic bootstrap validator — it auto-detects these traps before startup. MIT license, and even Tesseract.js creator gave us a thumbs up.

if you want the link, just say the word. it’ll save you hours of hair-pulling.