Skip to content

Agent Zero AI framework

Refactor model search & reorganize model_providers.yaml

keyboardstaff1b0c57a3e86d
diff --git a/conf/model_providers.yaml b/conf/model_providers.yaml
index de037c0f2..7eba45619 100644
--- a/conf/model_providers.yaml
+++ b/conf/model_providers.yaml
@@ -16,20 +16,21 @@
# kwargs: A dictionary of extra parameters to pass to LiteLLM.
# This is useful for `api_base`, `extra_headers`, etc.
#
-# Optional model search fields (used by the Model Configuration plugin):
-# models_endpoint: URL or path for the model listing API.
+# Optional model listing fields (used by the Model Configuration plugin):
+# models_list:
+# endpoint_url: URL or path for the model listing API.
# Absolute URL (https://...) is used directly.
# Relative path (/path) is appended to api_base or default_base.
-# models_format: Response parsing format: "openai" (default), "google", "ollama".
-# models_params: Extra query parameters for the listing request.
-# static_models: Hardcoded model list for providers without a listing API.
-# default_base: Default base URL for local/self-hosted providers.
+# format: Response parsing format: "openai" (default), "google", "ollama".
+# params: Extra query parameters for the listing request.
+# default_base: Default base URL for local/self-hosted providers.
chat:
a0_venice:
name: Agent Zero API
litellm_provider: openai
- models_endpoint: "https://api.venice.ai/api/v1/models"
+ models_list:
+ endpoint_url: "https://api.venice.ai/api/v1/models"
kwargs:
api_base: https://llm.agent-zero.ai/v1
venice_parameters:
@@ -37,24 +38,23 @@ chat:
anthropic:
name: Anthropic
litellm_provider: anthropic
- models_endpoint: "https://api.anthropic.com/v1/models"
- models_params:
- limit: "1000"
+ models_list:
+ endpoint_url: "https://api.anthropic.com/v1/models"
+ params:
+ limit: "1000"
cometapi:
name: CometAPI
litellm_provider: cometapi
+ models_list:
+ endpoint_url: "https://api.cometapi.com/v1/models"
deepseek:
name: DeepSeek
litellm_provider: deepseek
- models_endpoint: "https://api.deepseek.com/models"
+ models_list:
+ endpoint_url: "https://api.deepseek.com/models"
github_copilot:
name: GitHub Copilot
litellm_provider: github_copilot
- static_models:
- - "gpt-4.1"
- - "gpt-4o"
- - "gpt-5-mini"
- - "oswe-vscode-prime"
kwargs:
extra_headers:
"Editor-Version": "vscode/1.85.1"
@@ -63,54 +63,63 @@ chat:
google:
name: Google
litellm_provider: gemini
- models_endpoint: "/v1beta/models"
- models_format: "google"
- models_params:
- pageSize: "1000"
- default_base: "https://generativelanguage.googleapis.com"
+ models_list:
+ endpoint_url: "/v1beta/models"
+ format: "google"
+ params:
+ pageSize: "1000"
+ default_base: "https://generativelanguage.googleapis.com"
groq:
name: Groq
litellm_provider: groq
- models_endpoint: "https://api.groq.com/openai/v1/models"
+ models_list:
+ endpoint_url: "https://api.groq.com/openai/v1/models"
huggingface:
name: HuggingFace
litellm_provider: huggingface
lm_studio:
name: LM Studio
litellm_provider: lm_studio
- models_endpoint: "/v1/models"
- default_base: "http://host.docker.internal:1234"
+ models_list:
+ endpoint_url: "/v1/models"
+ default_base: "http://host.docker.internal:1234"
mistral:
name: Mistral AI
litellm_provider: mistral
- models_endpoint: "https://api.mistral.ai/v1/models"
+ models_list:
+ endpoint_url: "https://api.mistral.ai/v1/models"
moonshot:
name: Moonshot AI
litellm_provider: moonshot
- models_endpoint: "https://api.moonshot.cn/v1/models"
+ models_list:
+ endpoint_url: "https://api.moonshot.cn/v1/models"
ollama:
name: Ollama
litellm_provider: ollama
- models_endpoint: "/api/tags"
- models_format: "ollama"
- default_base: "http://host.docker.internal:11434"
+ models_list:
+ endpoint_url: "/api/tags"
+ format: "ollama"
+ default_base: "http://host.docker.internal:11434"
openai:
name: OpenAI
litellm_provider: openai
- models_endpoint: "https://api.openai.com/v1/models"
+ models_list:
+ endpoint_url: "https://api.openai.com/v1/models"
azure:
name: OpenAI Azure
litellm_provider: azure
- models_endpoint: "/openai/models"
- models_params:
- api-version: "2024-10-21"
+ models_list:
+ endpoint_url: "/openai/models"
+ params:
+ api-version: "2024-10-21"
bedrock:
name: AWS Bedrock
litellm_provider: bedrock
openrouter:
name: OpenRouter
litellm_provider: openrouter
- models_endpoint: "https://openrouter.ai/api/v1/models"
+ models_list:
+ endpoint_url: "https://openrouter.ai/api/v1/models"
kwargs:
extra_headers:
"HTTP-Referer": "https://agent-zero.ai/"
@@ -121,7 +130,8 @@ chat:
sambanova:
name: Sambanova
litellm_provider: sambanova
- models_endpoint: "https://api.sambanova.ai/v1/models"
+ models_list:
+ endpoint_url: "https://api.sambanova.ai/v1/models"
venice:
name: Venice.ai
litellm_provider: openai
@@ -132,32 +142,20 @@ chat:
xai:
name: xAI
litellm_provider: xai
- models_endpoint: "https://api.x.ai/v1/models"
+ models_list:
+ endpoint_url: "https://api.x.ai/v1/models"
zai:
name: Z.AI
litellm_provider: openai
- static_models:
- - "glm-4-plus"
- - "glm-4-air-250414"
- - "glm-4-airx"
- - "glm-4-long"
- - "glm-4-flashx"
- - "glm-4-flash-250414"
- - "glm-4v-plus"
- - "glm-4v"
- - "glm-3-turbo"
+ models_list:
+ endpoint_url: "/models"
kwargs:
api_base: https://api.z.ai/api/paas/v4
zai_coding:
name: Z.AI Coding
litellm_provider: openai
- static_models:
- - "codegeex-4"
- - "glm-4-plus"
- - "glm-4-air-250414"
- - "glm-4-airx"
- - "glm-4-flashx"
- - "glm-4-flash-250414"
+ models_list:
+ endpoint_url: "/models"
kwargs:
api_base: https://api.z.ai/api/coding/paas/v4
other:
diff --git a/plugins/_model_config/api/model_search.py b/plugins/_model_config/api/model_search.py
index f2289e15e..fc5c8acc0 100644
--- a/plugins/_model_config/api/model_search.py
+++ b/plugins/_model_config/api/model_search.py
@@ -13,55 +13,51 @@ _LITELLM_EXCLUDE = frozenset({
class ModelSearch(ApiHandler):
async def process(self, input: dict, request: Request) -> dict | Response:
provider = input.get("provider", "")
- query = input.get("query", "").lower()
model_type = input.get("model_type", "chat")
user_api_base = input.get("api_base", "")
if not provider:
return {"models": []}
- cfg = self._get_search_config(model_type, provider)
+ cfg = self._get_provider_cfg(model_type, provider)
+ ml = self._get_models_list(cfg)
- static = cfg.get("static_models")
- if static:
- all_models = list(static)
- else:
- all_models = await self._fetch_models(provider, cfg, user_api_base) or []
+ all_models = await self._fetch_models(provider, cfg, ml, user_api_base) or []
- if not all_models:
- litellm_provider = (cfg or {}).get("litellm_provider", provider)
- if litellm_provider == provider:
- all_models = self._litellm_fallback(provider, cfg)
+ if not all_models:
+ litellm_provider = cfg.get("litellm_provider", provider)
+ if litellm_provider == provider:
+ all_models = self._litellm_fallback(provider, cfg)
- if query:
- all_models = [m for m in all_models if query in m.lower()]
-
- return {"models": sorted(all_models)[:50], "provider": provider}
+ return {"models": sorted(all_models), "provider": provider}
@staticmethod
- def _get_search_config(model_type: str, provider: str) -> dict:
- """Get provider config with search metadata, falling back to chat config."""
+ def _get_provider_cfg(model_type: str, provider: str) -> dict:
+ """Get provider config, falling back to chat config for models_list."""
cfg = get_provider_config(model_type, provider) or {}
- if model_type != "chat" and not cfg.get("models_endpoint") and not cfg.get("static_models"):
+ if model_type != "chat" and not cfg.get("models_list"):
chat_cfg = get_provider_config("chat", provider) or {}
- merged = dict(cfg)
- for field in ("models_endpoint", "models_format", "models_params",
- "static_models", "default_base"):
- if field in chat_cfg and field not in merged:
- merged[field] = chat_cfg[field]
- return merged
+ if chat_cfg.get("models_list"):
+ merged = dict(cfg)
+ merged["models_list"] = chat_cfg["models_list"]
+ return merged
return cfg
- async def _fetch_models(self, provider: str, cfg: dict, user_api_base: str = "") -> list[str] | None:
+ @staticmethod
+ def _get_models_list(cfg: dict) -> dict:
+ """Extract models_list sub-config."""
+ return cfg.get("models_list") or {}
+
+ async def _fetch_models(self, provider: str, cfg: dict, ml: dict, user_api_base: str = "") -> list[str] | None:
api_key = models.get_api_key(provider)
api_base = user_api_base or (cfg or {}).get("kwargs", {}).get("api_base", "")
- url, fmt = self._resolve_url(cfg, api_base)
+ url, fmt = self._resolve_url(ml, api_base)
if not url:
return None
headers = self._build_headers(provider, api_key, cfg)
- params = dict(cfg.get("models_params", {}) or {})
+ params = dict(ml.get("params", {}) or {})
# Google uses query-param auth
if provider == "google" and api_key and api_key != "None":
@@ -80,10 +76,10 @@ class ModelSearch(ApiHandler):
return None
@staticmethod
- def _resolve_url(cfg: dict, api_base: str) -> tuple[str | None, str]:
- fmt = cfg.get("models_format", "openai")
- endpoint = cfg.get("models_endpoint", "")
- default_base = cfg.get("default_base", "")
+ def _resolve_url(ml: dict, api_base: str) -> tuple[str | None, str]:
+ fmt = ml.get("format", "openai")
+ endpoint = ml.get("endpoint_url", "")
+ default_base = ml.get("default_base", "")
if endpoint.startswith("http"):
return endpoint, fmt
diff --git a/plugins/_model_config/webui/config.html b/plugins/_model_config/webui/config.html
index c831f0015..d5fd544df 100644
--- a/plugins/_model_config/webui/config.html
+++ b/plugins/_model_config/webui/config.html
@@ -89,7 +89,8 @@
</div>
<div class="field-control" style="position:relative;"
x-data="{ results: [], open: false, searching: false,
- doSearch() { this.searching = true; $store.modelConfig.searchModels(preset.chat.provider, preset.chat.name, 'chat', preset.chat.api_base).then(r => { this.results = r; this.open = true; }).finally(() => this.searching = false); }
+ doSearch() { this.searching = true; $store.modelConfig.searchModels(preset.chat.provider, preset.chat.name, 'chat', preset.chat.api_base).then(r => { this.results = r; this.open = true; }).finally(() => this.searching = false); },
+ grouped() { return $store.modelConfig.groupResults(this.results, preset.chat.name); }
}"
@click.outside="open = false">
<input type="text" x-model="preset.chat.name" style="padding-right:32px;"
@@ -101,7 +102,11 @@
<span class="material-symbols-outlined model-search-spinner" :style="!searching && 'opacity:0'">progress_activity</span>
</span>
<div class="model-search-results" x-show="open && results.length > 0" x-transition.opacity>
- <template x-for="m in results" :key="m">
+ <template x-for="m in grouped().matched" :key="'m_'+m">
+ <div class="model-search-item matched" @click="preset.chat.name = m; open = false;" x-text="m"></div>
+ </template>
+ <div class="model-search-separator" x-show="grouped().matched.length > 0 && grouped().rest.length > 0"></div>
+ <template x-for="m in grouped().rest" :key="'r_'+m">
<div class="model-search-item" @click="preset.chat.name = m; open = false;" x-text="m"></div>
</template>
</div>
@@ -163,7 +168,8 @@
</div>
<div class="field-control" style="position:relative;"
x-data="{ results: [], open: false, searching: false,
- doSearch() { this.searching = true; $store.modelConfig.searchModels(preset.utility.provider || preset.chat.provider, preset.utility.name, 'chat', preset.utility.api_base || preset.chat.api_base).then(r => { this.results = r; this.open = true; }).finally(() => this.searching = false); }
+ doSearch() { this.searching = true; $store.modelConfig.searchModels(preset.utility.provider || preset.chat.provider, preset.utility.name, 'chat', preset.utility.api_base || preset.chat.api_base).then(r => { this.results = r; this.open = true; }).finally(() => this.searching = false); },
+ grouped() { return $store.modelConfig.groupResults(this.results, preset.utility.name); }
}"
@click.outside="open = false">
<input type="text" x-model="preset.utility.name" style="padding-right:32px;"
@@ -175,7 +181,11 @@
<span class="material-symbols-outlined model-search-spinner" :style="!searching && 'opacity:0'">progress_activity</span>
</span>
<div class="model-search-results" x-show="open && results.length > 0" x-transition.opacity>
- <template x-for="m in results" :key="m">
+ <template x-for="m in grouped().matched" :key="'m_'+m">
+ <div class="model-search-item matched" @click="preset.utility.name = m; open = false;" x-text="m"></div>
+ </template>
+ <div class="model-search-separator" x-show="grouped().matched.length > 0 && grouped().rest.length > 0"></div>
+ <template x-for="m in grouped().rest" :key="'r_'+m">
<div class="model-search-item" @click="preset.utility.name = m; open = false;" x-text="m"></div>
</template>
</div>
@@ -262,7 +272,8 @@
</div>
<div class="field-control" style="position:relative;"
x-data="{ results: [], open: false, searching: false,
- doSearch() { this.searching = true; $store.modelConfig.searchModels(config[section.key].provider, config[section.key].name, $store.modelConfig.getSearchType(section.key), config[section.key].api_base).then(r => { this.results = r; this.open = true; }).finally(() => this.searching = false); }
+ doSearch() { this.searching = true; $store.modelConfig.searchModels(config[section.key].provider, config[section.key].name, $store.modelConfig.getSearchType(section.key), config[section.key].api_base).then(r => { this.results = r; this.open = true; }).finally(() => this.searching = false); },
+ grouped() { return $store.modelConfig.groupResults(this.results, config[section.key].name); }
}"
@click.outside="open = false">
<input type="text" x-model="config[section.key].name" style="padding-right:32px;"
@@ -274,7 +285,11 @@
<span class="material-symbols-outlined model-search-spinner" :style="!searching && 'opacity:0'">progress_activity</span>
</span>
<div class="model-search-results" x-show="open && results.length > 0" x-transition.opacity>
- <template x-for="m in results" :key="m">
+ <template x-for="m in grouped().matched" :key="'m_'+m">
+ <div class="model-search-item matched" @click="config[section.key].name = m; open = false;" x-text="m"></div>
+ </template>
+ <div class="model-search-separator" x-show="grouped().matched.length > 0 && grouped().rest.length > 0"></div>
+ <template x-for="m in grouped().rest" :key="'r_'+m">
<div class="model-search-item" @click="config[section.key].name = m; open = false;" x-text="m"></div>
</template>
</div>
@@ -608,6 +623,15 @@
cursor: default;
font-style: italic;
}
+ .model-search-item.matched {
+ font-weight: 500;
+ }
+ .model-search-separator {
+ height: 1px;
+ margin: 4px 8px;
+ background: var(--color-border);
+ opacity: 0.5;
+ }
.model-search-item.disabled:hover {
background: transparent;
}
diff --git a/plugins/_model_config/webui/model-config-store.js b/plugins/_model_config/webui/model-config-store.js
index bd1be368e..b91de42b8 100644
--- a/plugins/_model_config/webui/model-config-store.js
+++ b/plugins/_model_config/webui/model-config-store.js
@@ -168,6 +168,18 @@ export const store = createStore("modelConfig", {
}
},
+ groupResults(models, query) {
+ const q = (query || '').trim().toLowerCase();
+ if (!q) return { matched: [], rest: models };
+ const matched = [];
+ const rest = [];
+ for (const m of models) {
+ if (m.toLowerCase().includes(q)) matched.push(m);
+ else rest.push(m);
+ }
+ return { matched, rest };
+ },
+
// Model Switcher
async loadSwitcherState(contextId) {
const result = { allowed: false, presets: [], override: null };