fix(config): preserve list-format models in custom_providers normalize by drstrangerujn · Pull Request #13509 · NousResearch/hermes-agent

drstrangerujn · 2026-04-21T12:32:39Z

Problem

_normalize_custom_provider_entry in hermes_cli/config.py only preserves models when it's a dict. Configs that write models as a plain list of ids — a hand-edited shape that older Hermes versions also used — get the field silently dropped. The gateway calls get_compatible_custom_providers(cfg) before list_authenticated_providers, so the picker receives entries without models and /model shows the provider with (0) models.

The downstream picker code in hermes_cli/model_switch.py (around line 1150) already has a elif isinstance(cfg_models, list) branch, so the drop happens strictly in the normalize pass.

Repro

custom_providers:
- name: tencent-coding-plan
  base_url: https://api.lkeap.cloud.tencent.com/coding/v3
  api_key: sk-…
  api_mode: chat_completions
  models:
  - glm-5
  - kimi-k2.5
  - minimax-m2.5

/model shows tencent-coding-plan (0). Calling list_authenticated_providers() directly with the raw custom_providers list (skipping get_compatible_custom_providers / normalize) returns all three models, confirming the drop is in normalize.

Fix

Accept list-shaped models and convert to the empty-value dict shape the rest of the pipeline already handles ({model_id: {}}). Dict-format entries pass through unchanged.

Non-string / empty / whitespace-only entries in the list are filtered out so bad data doesn't propagate.

Tests

Added to tests/hermes_cli/test_provider_config_validation.py:

test_models_list_converted_to_dict
test_models_dict_preserved
test_models_list_filters_empty_and_non_string
test_models_empty_list_omitted

pytest tests/hermes_cli/test_provider_config_validation.py -q → 16 passed (12 existing + 4 new).

_normalize_custom_provider_entry silently drops the models field when it's a list. Hand-edited configs (and the shape used by older Hermes versions) still write models as a plain list of ids, so after the normalize pass the entry reaches list_authenticated_providers() with no models and /model shows the provider with (0) models — even though the underlying picker code handles lists fine. Convert list-format models into the empty-value dict shape the rest of the pipeline already expects. Dict-format entries keep passing through unchanged. Repro (before the fix): custom_providers: - name: acme base_url: https://api.example.com/v1 models: [foo, bar, baz] /model shows "acme (0)"; bypassing normalize in list_authenticated_providers returns three models, confirming the drop happens in normalize. Adds four unit tests covering list→dict conversion, dict pass-through, filtering of empty/non-string entries, and the empty-list case.

alt-glitch · 2026-04-22T01:18:01Z

Related to #11017 (same root cause reported). Also see #12045 (another PR fixing list-format models).

alt-glitch · 2026-04-22T01:18:52Z

Related to #11017 (same root cause reported). Also see #12045 (another PR fixing list-format models).

teknium1 · 2026-04-23T09:37:39Z

Merged via PR #14495 — your commit f11d1898 was cherry-picked onto current main with your authorship preserved in git log (a5b0c7e2 on main).

#14495

Thanks for the clean diagnosis and targeted fix — _normalize_custom_provider_entry silently dropping list-shaped models: was exactly the root cause. 4 new tests on the normalizer + E2E verified against the tencent-coding-plan repro.

@davidvv

#9210) Multiple custom_providers entries sharing the same base_url + api_key are now grouped into a single picker row. A local Ollama host with per-model display names ("Ollama — GLM 5.1", "Ollama — Qwen3-coder", "Ollama — Kimi K2", "Ollama — MiniMax M2.7") previously produced four near-duplicate picker rows that differed only by suffix; now it appears as one "Ollama" row with four models. Key changes: - Grouping key changed from slug-by-name to (base_url, api_key). Names frequently differ per model while the endpoint stays the same. - When the grouped endpoint matches current_base_url, the row's slug is set to current_provider so picker-driven switches route through the live credential pipeline (no re-resolution needed). - Per-model suffix is stripped from the display name ("Ollama — X" → "Ollama") via em-dash / " - " separators. - Two groups with different api_keys at the same base_url (or otherwise colliding on cleaned name) are disambiguated with a numeric suffix (custom:openai, custom:openai-2) so both stay visible. - current_base_url parameter plumbed through both gateway call sites. Existing #8216, #11499, #13509 regressions covered (dict/list shapes of models:, section-3/section-4 dedup, normalized list-format entries). Salvaged from @davidvv's PR #9210 — the underlying code had diverged ~1400 commits since that PR was opened, so this is a reconstruction of the same approach on current main rather than a clean cherry-pick. Authorship preserved via --author on this commit. Closes #9210

@davidvv

NousResearch#9210) Multiple custom_providers entries sharing the same base_url + api_key are now grouped into a single picker row. A local Ollama host with per-model display names ("Ollama — GLM 5.1", "Ollama — Qwen3-coder", "Ollama — Kimi K2", "Ollama — MiniMax M2.7") previously produced four near-duplicate picker rows that differed only by suffix; now it appears as one "Ollama" row with four models. Key changes: - Grouping key changed from slug-by-name to (base_url, api_key). Names frequently differ per model while the endpoint stays the same. - When the grouped endpoint matches current_base_url, the row's slug is set to current_provider so picker-driven switches route through the live credential pipeline (no re-resolution needed). - Per-model suffix is stripped from the display name ("Ollama — X" → "Ollama") via em-dash / " - " separators. - Two groups with different api_keys at the same base_url (or otherwise colliding on cleaned name) are disambiguated with a numeric suffix (custom:openai, custom:openai-2) so both stay visible. - current_base_url parameter plumbed through both gateway call sites. Existing NousResearch#8216, NousResearch#11499, NousResearch#13509 regressions covered (dict/list shapes of models:, section-3/section-4 dedup, normalized list-format entries). Salvaged from @davidvv's PR NousResearch#9210 — the underlying code had diverged ~1400 commits since that PR was opened, so this is a reconstruction of the same approach on current main rather than a clean cherry-pick. Authorship preserved via --author on this commit. Closes NousResearch#9210

@davidvv

NousResearch#9210) Multiple custom_providers entries sharing the same base_url + api_key are now grouped into a single picker row. A local Ollama host with per-model display names ("Ollama — GLM 5.1", "Ollama — Qwen3-coder", "Ollama — Kimi K2", "Ollama — MiniMax M2.7") previously produced four near-duplicate picker rows that differed only by suffix; now it appears as one "Ollama" row with four models. Key changes: - Grouping key changed from slug-by-name to (base_url, api_key). Names frequently differ per model while the endpoint stays the same. - When the grouped endpoint matches current_base_url, the row's slug is set to current_provider so picker-driven switches route through the live credential pipeline (no re-resolution needed). - Per-model suffix is stripped from the display name ("Ollama — X" → "Ollama") via em-dash / " - " separators. - Two groups with different api_keys at the same base_url (or otherwise colliding on cleaned name) are disambiguated with a numeric suffix (custom:openai, custom:openai-2) so both stay visible. - current_base_url parameter plumbed through both gateway call sites. Existing NousResearch#8216, NousResearch#11499, NousResearch#13509 regressions covered (dict/list shapes of models:, section-3/section-4 dedup, normalized list-format entries). Salvaged from @davidvv's PR NousResearch#9210 — the underlying code had diverged ~1400 commits since that PR was opened, so this is a reconstruction of the same approach on current main rather than a clean cherry-pick. Authorship preserved via --author on this commit. Closes NousResearch#9210

@davidvv

NousResearch#9210) Multiple custom_providers entries sharing the same base_url + api_key are now grouped into a single picker row. A local Ollama host with per-model display names ("Ollama — GLM 5.1", "Ollama — Qwen3-coder", "Ollama — Kimi K2", "Ollama — MiniMax M2.7") previously produced four near-duplicate picker rows that differed only by suffix; now it appears as one "Ollama" row with four models. Key changes: - Grouping key changed from slug-by-name to (base_url, api_key). Names frequently differ per model while the endpoint stays the same. - When the grouped endpoint matches current_base_url, the row's slug is set to current_provider so picker-driven switches route through the live credential pipeline (no re-resolution needed). - Per-model suffix is stripped from the display name ("Ollama — X" → "Ollama") via em-dash / " - " separators. - Two groups with different api_keys at the same base_url (or otherwise colliding on cleaned name) are disambiguated with a numeric suffix (custom:openai, custom:openai-2) so both stay visible. - current_base_url parameter plumbed through both gateway call sites. Existing NousResearch#8216, NousResearch#11499, NousResearch#13509 regressions covered (dict/list shapes of models:, section-3/section-4 dedup, normalized list-format entries). Salvaged from @davidvv's PR NousResearch#9210 — the underlying code had diverged ~1400 commits since that PR was opened, so this is a reconstruction of the same approach on current main rather than a clean cherry-pick. Authorship preserved via --author on this commit. Closes NousResearch#9210

@davidvv

NousResearch#9210) Multiple custom_providers entries sharing the same base_url + api_key are now grouped into a single picker row. A local Ollama host with per-model display names ("Ollama — GLM 5.1", "Ollama — Qwen3-coder", "Ollama — Kimi K2", "Ollama — MiniMax M2.7") previously produced four near-duplicate picker rows that differed only by suffix; now it appears as one "Ollama" row with four models. Key changes: - Grouping key changed from slug-by-name to (base_url, api_key). Names frequently differ per model while the endpoint stays the same. - When the grouped endpoint matches current_base_url, the row's slug is set to current_provider so picker-driven switches route through the live credential pipeline (no re-resolution needed). - Per-model suffix is stripped from the display name ("Ollama — X" → "Ollama") via em-dash / " - " separators. - Two groups with different api_keys at the same base_url (or otherwise colliding on cleaned name) are disambiguated with a numeric suffix (custom:openai, custom:openai-2) so both stay visible. - current_base_url parameter plumbed through both gateway call sites. Existing NousResearch#8216, NousResearch#11499, NousResearch#13509 regressions covered (dict/list shapes of models:, section-3/section-4 dedup, normalized list-format entries). Salvaged from @davidvv's PR NousResearch#9210 — the underlying code had diverged ~1400 commits since that PR was opened, so this is a reconstruction of the same approach on current main rather than a clean cherry-pick. Authorship preserved via --author on this commit. Closes NousResearch#9210

@davidvv

NousResearch#9210) Multiple custom_providers entries sharing the same base_url + api_key are now grouped into a single picker row. A local Ollama host with per-model display names ("Ollama — GLM 5.1", "Ollama — Qwen3-coder", "Ollama — Kimi K2", "Ollama — MiniMax M2.7") previously produced four near-duplicate picker rows that differed only by suffix; now it appears as one "Ollama" row with four models. Key changes: - Grouping key changed from slug-by-name to (base_url, api_key). Names frequently differ per model while the endpoint stays the same. - When the grouped endpoint matches current_base_url, the row's slug is set to current_provider so picker-driven switches route through the live credential pipeline (no re-resolution needed). - Per-model suffix is stripped from the display name ("Ollama — X" → "Ollama") via em-dash / " - " separators. - Two groups with different api_keys at the same base_url (or otherwise colliding on cleaned name) are disambiguated with a numeric suffix (custom:openai, custom:openai-2) so both stay visible. - current_base_url parameter plumbed through both gateway call sites. Existing NousResearch#8216, NousResearch#11499, NousResearch#13509 regressions covered (dict/list shapes of models:, section-3/section-4 dedup, normalized list-format entries). Salvaged from @davidvv's PR NousResearch#9210 — the underlying code had diverged ~1400 commits since that PR was opened, so this is a reconstruction of the same approach on current main rather than a clean cherry-pick. Authorship preserved via --author on this commit. Closes NousResearch#9210

@davidvv

NousResearch#9210) Multiple custom_providers entries sharing the same base_url + api_key are now grouped into a single picker row. A local Ollama host with per-model display names ("Ollama — GLM 5.1", "Ollama — Qwen3-coder", "Ollama — Kimi K2", "Ollama — MiniMax M2.7") previously produced four near-duplicate picker rows that differed only by suffix; now it appears as one "Ollama" row with four models. Key changes: - Grouping key changed from slug-by-name to (base_url, api_key). Names frequently differ per model while the endpoint stays the same. - When the grouped endpoint matches current_base_url, the row's slug is set to current_provider so picker-driven switches route through the live credential pipeline (no re-resolution needed). - Per-model suffix is stripped from the display name ("Ollama — X" → "Ollama") via em-dash / " - " separators. - Two groups with different api_keys at the same base_url (or otherwise colliding on cleaned name) are disambiguated with a numeric suffix (custom:openai, custom:openai-2) so both stay visible. - current_base_url parameter plumbed through both gateway call sites. Existing NousResearch#8216, NousResearch#11499, NousResearch#13509 regressions covered (dict/list shapes of models:, section-3/section-4 dedup, normalized list-format entries). Salvaged from @davidvv's PR NousResearch#9210 — the underlying code had diverged ~1400 commits since that PR was opened, so this is a reconstruction of the same approach on current main rather than a clean cherry-pick. Authorship preserved via --author on this commit. Closes NousResearch#9210

alt-glitch added type/bug Something isn't working comp/cli CLI entry point, hermes_cli/, setup wizard area/config Config system, migrations, profiles labels Apr 22, 2026

teknium1 closed this in #14495 Apr 23, 2026

teknium1 mentioned this pull request Apr 23, 2026

fix(model_switch): group custom_providers by endpoint in /model picker (#9210) #14505

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

fix(config): preserve list-format models in custom_providers normalize#13509

fix(config): preserve list-format models in custom_providers normalize#13509
drstrangerujn wants to merge 1 commit into
NousResearch:mainfrom
drstrangerujn:fix/custom-providers-models-list

drstrangerujn commented Apr 21, 2026 •

edited

Loading

Uh oh!

alt-glitch commented Apr 22, 2026

Uh oh!

alt-glitch commented Apr 22, 2026

Uh oh!

teknium1 commented Apr 23, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Conversation

drstrangerujn commented Apr 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Problem

Repro

Fix

Tests

Uh oh!

alt-glitch commented Apr 22, 2026

Uh oh!

alt-glitch commented Apr 22, 2026

Uh oh!

teknium1 commented Apr 23, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

drstrangerujn commented Apr 21, 2026 •

edited

Loading