fix(model-picker): dedupe user providers and enrich live OpenAI counts#13190
Closed
BigDon86 wants to merge 3 commits into
Closed
fix(model-picker): dedupe user providers and enrich live OpenAI counts#13190BigDon86 wants to merge 3 commits into
BigDon86 wants to merge 3 commits into
Conversation
added 3 commits
April 20, 2026 16:20
…talog counts Two related bugs in `hermes_cli.model_switch.list_authenticated_providers` caused the /model picker to show duplicate provider rows and zero-count OpenAI / OpenAI Direct rows. 1. Section 3 (user-defined providers from config.yaml `providers:`) appended rows without checking `seen_slugs`, producing duplicate entries when a config provider name overlapped with a built-in (e.g. "Nous Portal (27)" plus "nous (0)"). Now skips and registers the slug, matching Sections 1, 2, 2b, and 4. 2. Provider rows were built from curated/config-only counts while model selection used live catalogs, so OpenAI / OpenAI Direct displayed `total_models=0` until the user entered the row. A new enrichment pass at the end of `list_authenticated_providers` calls `provider_model_ids()` (and a direct `/v1/models` probe via `fetch_api_models` for user-defined providers that supply only `base_url` + `key_env`) so the CLI picker matches what the TUI gateway already produced. Explicit config `models:` lists are preserved — live data only fills empty rows. Closes NousResearch#7524 for the Section 3 dedupe path.
…providers The CLI picker calls get_compatible_custom_providers() which expands config.providers entries into legacy custom_providers shape (tagged with provider_key). Section 3 already emits a row for each user_provider, but Section 4 was processing the expanded copies separately and generating *different* slugs from the display name (e.g. user_provider key 'custom' named 'Ollama (local)' → Section 3 emits slug=custom, Section 4 emits slug=ollama-local), so the existing seen_slugs check missed them. Section 4 now skips entries whose provider_key is already in seen_slugs. The TUI gateway path was unaffected because it passes raw custom_providers (usually empty) rather than the expanded compatibility view.
Collaborator
|
Related: #7524 — same duplicate provider rows bug in /model picker. |
1 similar comment
Collaborator
|
Related: #7524 — same duplicate provider rows bug in /model picker. |
This was referenced Apr 22, 2026
Contributor
|
Closing as partially-redundant. The dedup half is already on main via The live OpenAI count enrichment half (calling Thanks for the detailed root-cause analysis. |
This was referenced Apr 23, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Problem
The CLI
/modelpicker had two regressions that were reproducible in the user's live environment:OpenAIandOpenAI Directshowed0models in the picker even thoughprovider_model_ids()returned the live catalog (122models).providers:config entries into compatibilitycustom_providersentries.Examples seen live before the fix included duplicate
Ollama (local)/Ollamastyle rows and duplicateOpenAI Directrows with inconsistent counts.Root cause
1. OpenAI/OpenAI Direct zero-count rows
list_authenticated_providers()built picker rows from curated/static/config-only counts, while the actual model-selection flow consulted live catalogs later at selection time. That meant providers with dynamic catalogs, especiallyopenaiandopenai-direct, rendered withtotal_models=0until the user selected them.2. Duplicate provider rows in CLI picker
The CLI path calls
get_compatible_custom_providers(), which expandsproviders:config entries into legacycustom_providersshape. This created two duplication paths:Because those slugs differed, the existing
seen_slugscheck did not catch the duplicate.Fix
_resolve_openai_catalog_credentials()and use live/v1/modelsprobing inprovider_model_ids()foropenaiandopenai-direct.list_authenticated_providers():provider_keywas already emitted in Section 3models:configured by the user instead of overwriting them with live probesVerification
Automated
Targeted picker/model tests on the rebased branch:
tests/hermes_cli/test_picker_section3_dedupe_and_enrichment.pytests/hermes_cli/test_model_validation.pytests/hermes_cli/test_model_switch_custom_providers.pytests/hermes_cli/test_user_providers_model_switch.pytests/hermes_cli/test_overlay_slug_resolution.pytests/hermes_cli/test_model_picker_viewport.pytests/hermes_cli/test_codex_cli_model_picker.pyResult:
112 passedLive verification in user environment
Observed picker rows after the fix:
Ollama—19OpenAI—122OpenAI Direct—122OpenRouter—31GitHub Copilot—14Anthropic—9No duplicate rows remained in the live picker.
Notes
This PR keeps the CLI picker behavior aligned with the TUI/gateway path, which was already enriching model options from live catalogs.