fix(model_switch): group custom_providers by endpoint in /model picker by davidvv · Pull Request #9210 · NousResearch/hermes-agent

davidvv · 2026-04-13T20:29:53Z

Fixes the /model command producing no output (or no picker) when the active configuration uses custom_providers with an Ollama or other OpenAI-compatible endpoint.

Root cause

list_authenticated_providers() had two problems in section 4 (custom_providers handling):

1. Missing current_base_url parameter
The function had no way to tell which custom entry corresponds to the active endpoint. As a result is_current was always False and the slug was a name-derived hash (e.g. ollama-glm-5-1) that did not match the config model.provider value (custom). Switching via the picker therefore sent the wrong provider slug to switch_model, causing credential resolution to fail.

2. One picker entry per config entry instead of grouping by endpoint
A single Ollama host with 4 model entries appeared as 4 separate single-model providers in the picker instead of one group with 4 models to choose from.

3. Silent TypeError at call site (same symptom on older gateway builds)
Both call sites in gateway/run.py already passed custom_providers=... as a keyword arg, but the old signature did not declare that parameter — resulting in a TypeError that was silently swallowed by the surrounding try/except, leaving providers = [] and the picker never opening.

Fix

Add current_base_url: str = parameter.
Replace section 4 with a grouping pass: entries are bucketed by (base_url, api_key). When a bucket matches the active endpoint (base_url == current_base_url) the slug is set to current_provider so that credential resolution in the existing switch_model pipeline works without any extra lookup.
Display name is derived from the first entry in the group, stripping per-model suffixes (Ollama — GLM 5.1 → Ollama). All model names from matching entries are merged into a single list.
Pass current_base_url through both list_authenticated_providers call sites in gateway/run.py.

Result

Before: 0 providers shown, picker never opened (TypeError swallowed).
After: 1 provider group Ollama with all 4 models, correctly marked as current, inline keyboard picker opens and model switching works.

list_authenticated_providers() had two issues with custom_providers: 1. Missing current_base_url parameter — the function had no way to tell which custom entry matches the active endpoint, so is_current was always False and the slug was a name-derived hash that did not match the config provider value (e.g. "custom"). Switching via the picker therefore resolved wrong credentials. 2. One entry per config item instead of grouping by endpoint — a single Ollama host with 4 model entries appeared as 4 separate providers in the picker, each with 1 model, instead of one group with 4 models. Fix: - Add current_base_url: str = "" parameter. - Replace section 4 with a grouping pass: entries are bucketed by (base_url, api_key). When a bucket matches the active endpoint (base_url == current_base_url) the slug is set to current_provider so that credential resolution in the existing switch pipeline works without any extra lookup. Display name is derived from the first entry in the group, stripping per-model suffixes ("Ollama — GLM" -> "Ollama"). Models from all entries in the group are merged into a single list. - Pass current_base_url through both list_authenticated_providers call sites in gateway/run.py (_handle_model_command picker + text list). Tested against a config with 4 Ollama cloud models on one endpoint: before: 0 providers shown (TypeError swallowed), picker never opened. after: 1 provider group "Ollama" with all 4 models, correctly marked as current, inline keyboard picker opens and switching works.

Copilot

Pull request overview

Fixes the /model interactive picker failing (often silently) when the active config uses custom_providers pointing at an OpenAI-compatible endpoint (e.g., Ollama), by correctly identifying the active endpoint and grouping custom provider entries by endpoint.

Changes:

Extend list_authenticated_providers() to accept current_base_url and use it to mark the active custom endpoint correctly.
Replace per-entry custom provider listing with grouping by (base_url, api_key) and merging models into a single picker entry per endpoint.
Pass current_base_url into list_authenticated_providers() from both /model call sites in gateway/run.py.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 4 comments.

File	Description
`hermes_cli/model_switch.py`	Adds `current_base_url` support and groups `custom_providers` by endpoint for the `/model` picker output.
`gateway/run.py`	Threads `current_base_url` through to provider listing so the picker can correctly mark/select the active custom endpoint.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

-        for entry in custom_providers:
+        from collections import OrderedDict
+        groups: dict = OrderedDict()
+        for i, entry in enumerate(custom_providers):


+                    "models": [],
+                    "source": "custom_providers",
+                    "api_url": b_url,
+                }


+                        for m in g["models"]:
+                            if m not in existing["models"]:
+                                existing["models"].append(m)
+                        existing["total_models"] = len(existing["models"])


    # --- 4. Saved custom providers from config ---
+    # Entries are grouped by (base_url, api_key) so a single Ollama host
+    # with several models appears as one provider group in the picker.
+    # When the entry's base_url matches the currently active endpoint the
+    # slug is set to current_provider so that switching via the picker
+    # resolves credentials correctly through the existing runtime provider
+    # logic (no extra credential lookup needed).
    if custom_providers and isinstance(custom_providers, list):
-        for entry in custom_providers:
+        from collections import OrderedDict
+        groups: dict = OrderedDict()
+        for i, entry in enumerate(custom_providers):
            if not isinstance(entry, dict):
                continue
-
-            display_name = (entry.get("name") or "").strip()
-            api_url = (
-                entry.get("base_url", "")
-                or entry.get("url", "")
-                or entry.get("api", "")
-                or ""
-            ).strip()
-            if not display_name or not api_url:
+            b_url = (
+                entry.get("base_url") or entry.get("url") or entry.get("api") or ""
+            ).rstrip("/")
+            a_key = entry.get("api_key") or ""
+            model_id = (entry.get("model") or "").strip()
+            if not b_url:
                continue
-
-            slug = custom_provider_slug(display_name)
-            if slug in seen_slugs:
-                continue
-
-            models_list = []
-            default_model = (entry.get("model") or "").strip()
-            if default_model:
-                models_list.append(default_model)
-
-            results.append({
-                "slug": slug,
-                "name": display_name,
-                "is_current": slug == current_provider,
-                "is_user_defined": True,
-                "models": models_list,
-                "total_models": len(models_list),
-                "source": "user-config",
-                "api_url": api_url,
-            })
-            seen_slugs.add(slug)
+            if current_base_url and b_url == current_base_url.rstrip("/"):
+                slug = current_provider or "custom"
+            else:
+                slug = custom_provider_slug(entry.get("name") or b_url)
+            group_key = (b_url, a_key)
+            if group_key not in groups:
+                raw_name = (entry.get("name") or "").strip()
+                # Strip per-model suffix ("Ollama — GLM 5.1" -> "Ollama")
+                display_name = raw_name.split("—")[0].strip() if "—" in raw_name else raw_name


- Remove unused enumerate() in section 4 loop - Normalise source value to "user-config" (consistent with existing taxonomy); update docstring to document this - Apply max_models cap after merging models in the slug-collision path (total_models keeps the full deduplicated count) - Add three regression tests: test_list_authenticated_providers_groups_same_endpoint test_list_authenticated_providers_current_endpoint_slug_and_is_current test_list_authenticated_providers_max_models_cap_after_merge

@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

alt-glitch · 2026-04-23T10:16:52Z

Superseded by #14505 which salvaged this fix onto current main.

@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

Copilot AI review requested due to automatic review settings April 13, 2026 20:29

Copilot started reviewing on behalf of davidvv April 13, 2026 20:30 View session

Copilot AI reviewed Apr 13, 2026

View reviewed changes

teknium1 mentioned this pull request Apr 23, 2026

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

Merged

teknium1 closed this in #14505 Apr 23, 2026

alt-glitch added type/bug Something isn't working P2 Medium — degraded but workaround exists comp/cli CLI entry point, hermes_cli/, setup wizard comp/gateway Gateway runner, session dispatch, delivery labels Apr 23, 2026

kagura-agent mentioned this pull request Apr 23, 2026

fix(model-picker): deduplicate providers across user-config and custom sections #12321

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

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

fix(model_switch): group custom_providers by endpoint in /model picker#9210
davidvv wants to merge 2 commits into
NousResearch:mainfrom
davidvv:fix/model-picker-custom-providers

davidvv commented Apr 13, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

alt-glitch commented Apr 23, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Conversation

davidvv commented Apr 13, 2026

Root cause

Fix

Result

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

alt-glitch commented Apr 23, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants