GoModel & LibreChat

Overview

LibreChat is a self-hosted chat UI that talks to OpenAI-compatible APIs through custom endpoints. Pointing one custom endpoint at GoModel gives every LibreChat user all your providers behind a single connection, and every chat is logged, priced, and attributed to the signed-in LibreChat user in GoModel. Flow: LibreChat chat -> LibreChat backend -> GoModel -> OpenAI/Anthropic/Gemini/... What you get:

Full request and response bodies for every chat in the GoModel audit log.
Cost per LibreChat user via header templates and labels.
All providers behind one custom endpoint; no per-provider setup in LibreChat.
Prometheus metrics and OpenTelemetry traces for chat traffic, once you enable those exporters (see Notes).
Failover, rate limits, budgets, and response caching applied to chat traffic, when you configure them in GoModel.

1. Run GoModel next to LibreChat

LibreChat reads custom endpoints from librechat.yaml. Create it next to your compose file:

librechat.yaml

version: 1.3.13
cache: true
endpoints:
  custom:
    - name: "GoModel"
      apiKey: "${GOMODEL_API_KEY}"
      baseURL: "http://gomodel:8080/v1"
      headers:
        X-LibreChat-User-Email: "{{LIBRECHAT_USER_EMAIL}}"
      models:
        default: ["openai/gpt-5-mini"]
        fetch: true
      titleConvo: true
      titleModel: "openai/gpt-5-mini"
      modelDisplayLabel: "GoModel"

Then run GoModel, LibreChat, and MongoDB on one Docker network so LibreChat can reach GoModel by container name:

docker-compose.yml

services:
  gomodel:
    image: enterpilot/gomodel
    environment:
      GOMODEL_MASTER_KEY: change-me
      OPENAI_API_KEY: sk-...
      ANTHROPIC_API_KEY: sk-ant-...
      TAGGING_HEADER_1: X-LibreChat-User-Email
      TAGGING_HEADER_1_DONOTPASS: "true"
    ports:
      - "127.0.0.1:8080:8080"
    volumes:
      - gomodel_data:/app/data

  librechat:
    image: ghcr.io/danny-avila/librechat:v0.8.7
    depends_on:
      - mongodb
    environment:
      HOST: 0.0.0.0
      MONGO_URI: mongodb://mongodb:27017/LibreChat
      GOMODEL_API_KEY: change-me
      ALLOW_REGISTRATION: "true"
      SEARCH: "false"
      JWT_SECRET: ${JWT_SECRET:?generate with openssl rand -hex 32}
      JWT_REFRESH_SECRET: ${JWT_REFRESH_SECRET:?generate with openssl rand -hex 32}
      CREDS_KEY: ${CREDS_KEY:?generate with openssl rand -hex 32}
      CREDS_IV: ${CREDS_IV:?generate with openssl rand -hex 16}
    ports:
      - "127.0.0.1:3080:3080"
    volumes:
      - ./librechat.yaml:/app/librechat.yaml
      - librechat_uploads:/app/uploads

  mongodb:
    image: mongo:8
    volumes:
      - mongo_data:/data/db

volumes:
  gomodel_data:
  librechat_uploads:
  mongo_data:

Setting	Why
`baseURL`	Sends the endpoint’s traffic to GoModel. Keep the `/v1` suffix.
`apiKey` / `GOMODEL_API_KEY`	A GoModel API key (or `GOMODEL_MASTER_KEY`), not a real provider key.
`headers`	Adds the signed-in user’s email to every model call via LibreChat’s `{{LIBRECHAT_USER_EMAIL}}` template, which GoModel turns into labels below.
`models.fetch`	Fills the model picker from GoModel’s `GET /v1/models` instead of the hardcoded `default` list.
`titleModel`	LibreChat titles conversations with `gpt-3.5-turbo` by default, which GoModel does not expose. Set a model id GoModel serves.
`TAGGING_HEADER_1`	Tells GoModel to record that header’s value as a label on every request.
`TAGGING_HEADER_1_DONOTPASS`	Keeps the label but strips the header before forwarding, so user emails never reach a provider on passthrough routes. Translated routes never forward client headers.
`ALLOW_REGISTRATION`	Without a mounted `.env`, LibreChat refuses new sign-ups by default. The first account to register becomes the admin, so keep the port loopback-bound (as above) until that account exists, and turn registration off again once your users have accounts.
`SEARCH`	Disables Meilisearch chat search so no extra container is needed.
`JWT_SECRET`, `CREDS_KEY`, …	LibreChat’s session and encryption secrets. The `${VAR:?...}` syntax makes Compose refuse to start until you supply real values — generate each with `openssl rand -hex 32` (`-hex 16` for `CREDS_IV`) and put them in an `.env` file next to the compose file.

Open http://localhost:3080, register the first account (it becomes the LibreChat admin), and the GoModel endpoint appears in the model picker with every model GoModel exposes — including Anthropic and Gemini models and any virtual models.

Create a dedicated API key for LibreChat in the GoModel dashboard (API Keys -> Create API Key) instead of using the master key. Keys carry labels, budgets, and rate limits, so chat spend shows up separately in the dashboard.

2. Chat, and see every call in GoModel

Pick any model — openai/gpt-5-mini, anthropic/claude-haiku-4-5-20251001, gemini/gemini-2.5-flash — and chat. Streaming responses, and the title generation LibreChat runs after the first exchange, all pass through GoModel. Each call appears in the audit log with the full prompt, response, tokens, cost, and the header identifying the user.

Verified with LibreChat v0.8.7: model fetch, chat and title generation against OpenAI, Anthropic, and Gemini models, per-user labels, and RAG file embedding, all routed through GoModel.

Per-user cost tracking

With the headers template from step 1, every request carries X-LibreChat-User-Email and GoModel records its value as a label. The dashboard’s usage-by-label view then shows requests, tokens, and cost per person. If emails should not appear in usage data, use "{{LIBRECHAT_USER_ID}}" in the template instead, keeping TAGGING_HEADER_1_DONOTPASS. Keep TAGGING_HEADER_1_DONOTPASS set: regular routed requests never forward client headers, but passthrough requests would otherwise carry the email header to the provider. See Labelling for prefixes and other header options.

Curate the model picker

GoModel can expose hundreds of models, and fetch: true lists all of them. To offer a fixed menu instead, set fetch: false and list what your users should see:

models:
  default:
    - "openai/gpt-5-mini"
    - "anthropic/claude-haiku-4-5-20251001"
    - "gemini/gemini-2.5-flash"
  fetch: false

Or expose fewer providers in GoModel; fetch: true follows GET /v1/models.

Embeddings and RAG

LibreChat’s file uploads are embedded by its separate RAG API, which speaks the OpenAI embeddings API and can point at GoModel too. Add to the compose file:

  rag_api:
    image: ghcr.io/danny-avila/librechat-rag-api-dev-lite:latest
    depends_on:
      - vectordb
    environment:
      DB_HOST: vectordb
      POSTGRES_DB: librechat_rag
      POSTGRES_USER: librechat
      POSTGRES_PASSWORD: change-me
      RAG_OPENAI_BASEURL: http://gomodel:8080/v1
      RAG_OPENAI_API_KEY: change-me
      EMBEDDINGS_PROVIDER: openai
      EMBEDDINGS_MODEL: openai/text-embedding-3-small

  vectordb:
    image: pgvector/pgvector:pg15
    environment:
      POSTGRES_DB: librechat_rag
      POSTGRES_USER: librechat
      POSTGRES_PASSWORD: change-me
    volumes:
      - pg_data:/var/lib/postgresql/data

Add RAG_API_URL: http://rag_api:8000 to the librechat service and pg_data: to the volumes list. Every uploaded document is chunked and embedded through GoModel’s /v1/embeddings, so embedding usage is logged and priced like any other call. Chat routing works fine without this section.

Troubleshooting

Symptom	Fix
”Registration is not allowed.”	Set `ALLOW_REGISTRATION: "true"` on the `librechat` service (step 1) and recreate the container.
Endpoint missing from the model picker	`librechat.yaml` is not mounted at `/app/librechat.yaml`, or has a syntax error. Check the LibreChat startup logs for “Custom config file loaded”.
Model picker shows only the `default` list	`fetch: true` could not reach GoModel or the key was rejected. Use the Docker service name (`gomodel`) in `baseURL`, not `localhost`, and check GoModel logs for `401`.
”Outdated Config version” warning at startup	Bump the `version:` field in `librechat.yaml` to the version the warning suggests.
Conversations stay titled “New Chat”, with `404` errors in GoModel logs	`titleModel` is unset, so LibreChat asks for `gpt-3.5-turbo`. Set `titleModel` to a model GoModel exposes.
No per-user labels in usage	The `headers` block is missing from the endpoint, or `TAGGING_HEADER_1` is not set in GoModel. Both are needed.
File uploads fail	`RAG_API_URL` is not set or the RAG API is not running. See Embeddings and RAG.
Same answer repeats for identical prompts	GoModel’s response cache replays identical requests. This is usually what you want; disable the cache if not.

Notes

GoModel failover and budgets apply to chat traffic automatically; a per-key budget caps what the whole LibreChat instance can spend.
LibreChat’s agents, web search, and image generation features have their own provider settings and were not part of the verification run.
For dashboards and alerts, see Prometheus metrics and OpenTelemetry.

​Overview

​1. Run GoModel next to LibreChat

​2. Chat, and see every call in GoModel

​Per-user cost tracking

​Curate the model picker

​Embeddings and RAG

​Troubleshooting

​Notes

Overview

1. Run GoModel next to LibreChat

2. Chat, and see every call in GoModel

Per-user cost tracking

Curate the model picker

Embeddings and RAG

Troubleshooting

Notes