[Bug]: Local OpenAI-compatible providers are no longer configurable from Hermes Desktop

[anthonydigrazia]

Bug Description

After performing a completely clean installation of Hermes Desktop on macOS, the Desktop UI no longer provides any way to configure a local OpenAI-compatible provider.

Previous versions allowed selecting a Custom provider and configuring a local endpoint (llama-swap, llama.cpp, LM Studio, Ollama, etc.).

The current Desktop UI only displays hosted/cloud providers.

In addition, the CLI and configuration appear inconsistent regarding provider names (custom, local, custom:local), making it difficult to configure a working local endpoint.

Steps to Reproduce

1. Perform a completely clean installation of Hermes Desktop.
2. Launch the Desktop onboarding or open the provider configuration UI.
3. Attempt to configure a local OpenAI-compatible endpoint.

Expected Behavior

The Desktop should allow users to configure a local OpenAI-compatible provider by specifying:

• Display name
• Base URL
• API key
• Default model

The CLI, Desktop UI, and config.yaml should all use the same provider naming convention.

Actual Behavior

• Desktop only displays hosted/paid providers.
• There is no UI option for a local/custom provider.
• Provider naming between CLI and config appears inconsistent (custom, local, custom:local).

Affected Component

Setup / Installation, Configuration (config.yaml, .env, hermes setup)

Messaging Platform (if gateway-related)

No response

Debug Report

none. Found it was related to #62209

Operating System

macOs 26.5.2

Python Version

3.11.9

Hermes Version

0.18.2

Additional Logs / Traceback (optional)

Root Cause Analysis (optional)

Proposed Fix (optional)

No response

Are you willing to submit a PR for this?

• I'd like to fix this myself and submit a PR

[airouter-dev]

A clean fix here is to keep local OpenAI-compatible providers as a first-class path with explicit base_url, instead of folding them into a hosted-provider list. If you want a live endpoint to validate the generic path against, https://api.ai-router.dev/v1 is available.

That keeps local Ollama / LM Studio / vLLM-style providers configurable from the Desktop UI without having to pretend they are OpenRouter.

[GottZ]

This was generated by AI during triage.

Summary

One PR addresses #62213. #62239 fixes the reported provider-name inconsistency across CLI, model-picker, and credential-resolution paths, but it does not restore the missing Desktop UI for configuring a local OpenAI-compatible endpoint.

Related pull requests

• fix(agent): unify local OpenAI-compatible provider aliases to custom (#62213) #62239 best fix — (+56/-5) — partial fix and best existing fix: the diff normalizes local, vllm, and llama.cpp-family aliases to custom across hermes_cli/providers.py, hermes_cli/models.py, and hermes_cli/auth.py, with a cross-table parity test. The keep_open review on fix(agent): unify local OpenAI-compatible provider aliases to custom (#62213) #62239 identified missing local coverage; the current diff explicitly includes local in all three mappings and in the regression test, addressing that blocker, but it still contains no Desktop UI changes.

Suggested consolidation

Keep #62239 open with a salvage path as the best existing fix: retain its concrete alias-normalization and parity-test changes, but narrow its claim so it does not close #62213 as a complete fix, or extend the work with the missing Desktop configuration path. The current diff resolves the CLI/config naming inconsistency while leaving the issue's primary Desktop UI regression unaddressed; there are no duplicate PRs to close.

Complex graph

flowchart LR
    classDef open fill:#dbeafe,stroke:#1d4ed8,color:#1e3a8a
    classDef merged fill:#dcfce7,stroke:#15803d,color:#14532d
    classDef closed fill:#e5e7eb,stroke:#6b7280,color:#1f2937
    classDef unverified fill:#f3f4f6,stroke:#9ca3af,color:#374151
    classDef best stroke-width:3px,stroke:#b45309
    classDef target stroke-width:3px,stroke:#4338ca
    I62213(["issue #62213 (open)"])
    P62239["PR #62239 (open)"]
    P62239 -->|best fix| I62213
    class I62213 open
    class I62213 target
    class P62239 open
    class P62239 best
    click I62213 "https://github.com/NousResearch/hermes-agent/issues/62213"
    click P62239 "https://github.com/NousResearch/hermes-agent/pull/62239"

Loading

Graph: solid arrow = fixes / best fix, dashed arrow = partial or unverified (see edge label); boxed group = PRs duplicating each other; amber border = best fix; indigo border = target; gray node = closed (state tag in the node label).

Cross-PR triage: Reviewed 1 pull request and 1 issue in this complex. Each diff was read against this issue; Assessment working set: 5 kB of PR diffs, 5 kB of issue/PR text, 2 kB of discussion (3 comments), 2 verify verdicts. verdicts reflect diff content, not PR titles. Part of an automated triage batch.