pod
Other formats agents might prefer:
markdownjsonllms.txt

Agent? You probably want markdown or json, or Pod over MCP.

Reported issues for GoModel

Pod holds 12 of 12 problems reported by people outside the maintainer team. Issues filed by the project's own owners, members and collaborators are excluded entirely — a maintainer's release checklist is not a warning to a prospective user.

Back to GoModel.

Most discussed

Proposal: scoped model-access policies (allow-list which models a consumer subtree may call)

Auth keys currently authenticate and attribute (user_path/labels), but any active key can call every model the gateway serves; the only allow-list is global (CONFIGURED_PROVIDER_MODELS_MODE=allowlist). Multi-team / multi-workload deployments need "this consumer subtree may only call these models" — the same shape budgets and rate-limit rules already solve for spend and throughput.

Proposal: a modelaccess policy resource mirroring budgets/rate-limits:

  • `{scope: user_path|label, subject

Read the thread · 2026-08-21 · open · outside contributor · 3 comments

Feature: add Docker image with npx included

The default docker image just containes the gomodel binary, therefore it cannot run stdio mcp server as npx is not included in the binary. Could you create alternative images in the CI/CD flow with npx included, maybe with a different tag (-full e.g.)?

Read the thread · 2026-08-19 · open · external user · 2 comments

Feature: Add function to try models

It would be very useful if you could add a simple function to actually try the models with a simple chat request and a prompt to see if they actually work, maybe see the response time / speed (TPS), or maybe even compare 2 or multiple models' replies side-by-side.

I use GoModel by exposing only virtual models so this way it would be much easier to test the models before configuring them for the virtual model.

Read the thread · 2026-08-17 · closed · external user · 4 comments

Add Kilo Gateway (kilo.ai) as a native OpenAI-compatible provider

Feature Request: Add Kilo Gateway (kilo.ai) as a native OpenAI-compatible provider

Provider Overview Kilo Gateway (kilo.ai/gateway) is a universal AI inference API that exposes 500+ models (Anthropic, OpenAI, Google, xAI, Mistral, MiniMax, Z.ai, etc.) through a single OpenAI-compatible endpoint at https://api.kilo.ai/api/gateway. It uses standard Bearer authentication and supports the provider/model naming convention (similar to OpenRouter).

  • Docs: h

Read the thread · 2026-07-15 · closed · outside contributor · 3 comments

Codex + DeepSeek fails: include field rejected in chat-translated Responses requests

Problem

When using Codex with DeepSeek through GoModel via the Responses API, requests fail with:

{
  "error": {
    "code": null,
    "message": "responses field \"include\" is only supported by native Responses providers; use an OpenAI-compatible provider or passthrough for this request",
    "param": null,
    "type": "invalid_request_error"
  }
}

Root Cause

  1. Codex requirement: Codex (0.122.0+) enforces wire_api = "responses" in its configuration and automaticall

Read the thread · 2026-07-14 · closed · external user · 5 comments

MCP proxy in GoModel

Thumbs up if you’d love to see this implemented.

Read the thread · 2026-07-06 · closed · outside contributor · 4 comments

Memory usage spikes to ~3GB and never releases on coding-agent workload (restart required)

I run 2 GoModel instances with identical configuration (Postgres, Redis, and no semantic caching):

  • Instance A (data processing)

    • ~40k requests/day
    • ~30M tokens/day
    • Memory remains stable and stays < ~150MB

  • Instance B (coding agents)

    • ~1500 requests/day
    • ~20M tokens/day
    • When handling coding-agent workflows, GoModel memory grows very fast, sometimes reaching ~3GB
    • After the spike, memory does not decrease automatically; must **restart the contain

Read the thread · 2026-07-05 · closed · external user · 6 comments

Add a -health CLI flag for container healthchecks

The runtime image is built from distroless, which ships only the /gomodel binary - no shell, no wget/curl. This makes it impossible to write a working Docker HEALTHCHECK even though the gateway already exposes a GET /health endpoint.

Specifically:

  • CMD-SHELL healthchecks fail because /bin/sh doesn't exist in the image.
  • Exec-form healthchecks (CMD ["wget", ...] / CMD ["curl", ...]) fail because neither binary is present.
  • docker exec for debugging (ls, cat, etc.) is a

Read the thread · 2026-06-16 · closed · external user · 3 comments

Most recent

Feature: Integrate with llm-d

Is your feature request related to a problem? Please describe. Operators running self-hosted vLLM fleets through llm-d cannot currently use GoModel as the policy, compatibility, and observability gateway while retaining llm-d-aware endpoint selection. Modeling each serving replica as a static provider loses llm-d scheduling signals such as queue state, load, and prefix-cache locality.

Describe the solution you'd like Add a supported llm-d integration mode for Kubernetes deployments. GoM

Read the thread · 2026-08-06 · closed · outside contributor · 1 comment

Feature: Add /v1/images/edits endpoint

Is your feature request related to a problem? Please describe. GoModel does not implement the OpenAI-compatible /v1/images/edits endpoint (editing/inpainting an existing image given an image plus an optional mask and a prompt). No route, handler, or provider adapter exists anywhere in the codebase.

Describe the solution you'd like Add a /v1/images/edits endpoint accepting the standard OpenAI multipart request shape (source image, optional mask image, prompt, model, size, n) and retu

Read the thread · 2026-08-03 · closed · outside contributor · 0 comments

Validation with ENV and config

I have an OpenCode token I pass as an ENV var and that automatically provisions the provider upon detection.

I also have a config file I pass with my vLLM providers. I generated some bad yaml and the issue was hidden from me until I removed the OpenCode provider, at which point the server started crashing (no providers configured) and I was able to narrow down why my local models were not showing up.

I'm not sure what to do here validation wise, but it would be helpful to have a boot log line

Read the thread · 2026-07-08 · closed · external user · 1 comment

Endpoint so users can get their own usage (limit/status)

This would enable us to build plugins for tools like OpenCode

Read the thread · 2026-07-03 · closed · external user · 0 comments

The remaining reports are on the project's issue tracker.