Emit the existing model-setup-changed signal after successful preset selection or override clearing. The setup gate checks its original chat and retries the retained prompt automatically, preserving new composer drafts and ignoring failed saves. Validation: 67 focused gate, selector, model-config, and DOM tests passed; the 11 gate tests passed again after final cleanup. Live Docker WebUI checks verified automatic retry after selecting a ready preset and clearing an unconfigured override, with the draft and original chat target preserved. |
||
|---|---|---|
| .. | ||
| api | ||
| extensions | ||
| helpers | ||
| webui | ||
| AGENTS.md | ||
| default_config.yaml | ||
| hooks.py | ||
| mode_presets_fallback.yaml | ||
| plugin.yaml | ||
| provider_metadata.yaml | ||
| README.md | ||
Model Configuration
Manage the reusable model presets used for Agent Zero's main, utility, and embedding models.
Model Presets
Defaultis always present, appears first, and cannot be renamed or deleted.- Every runtime configuration resolves from a preset. There is no separate durable model configuration layer.
- Preset definitions are global and live in
usr/plugins/_model_config/presets.yaml. Plugin defaults inmode_presets_fallback.yamlare used only when a saved collection cannot be initialized. - Non-default presets may omit advanced fields or model slots; omitted values resolve from
Default, while provider-specifickwargsare replaced or cleared instead of leaking between providers. - API keys remain in the approved environment/settings flow and are never written into preset YAML.
- Models whose provider or name contains
anthropic, or whose name containsclaude, receive public-action-summary wording for the bundledthoughtsinstructions. Matching is case-insensitive and applies regardless of provider, including third-party and OpenAI-compatible routes. Shared prompt files and thethoughtsJSON field stay unchanged.
Scoped Selection
Global, project, project/profile, and agent-profile scopes store only a preset name in their standard plugin config.json:
{"model_preset": "Balance"}
For example, a project selection is stored at:
/a0/usr/projects/<project>/.a0proj/plugins/_model_config/config.json
The normal plugin resolution order selects the most specific available reference. A missing or deleted reference safely resolves to Default. Chats may store an explicit preset reference in chat_model_override; clearing it returns the chat to its scoped selection.
User Interfaces
- Agent Settings shows the global selection, its three resolved models, and actions for preset editing, API keys, and per-project/agent settings.
- The full plugin settings modal uses the generic scope selector and stores only the chosen preset at that scope.
- The closed chat switcher shows the effective preset plus the main model's short name; its menu supports a chat-only selection or returning to the scoped preset. Preset and profile menus shrink to fit short windows so the full scroll track stays reachable.
- The adjacent profile switcher shows the active agent's effective title and avatar and links to Create, Edit, and Manage agents.
- The preset editor exposes the shared API key for each selected provider and saves key changes separately from secret-free preset definitions.
- Project Settings selects from the same global preset definitions.
Cache Routing
Venice and Agent Zero API send a stable prompt_cache_key derived from the model, leading instructions, and tool definitions. This is a routing hint; cache availability and retention remain provider-controlled. Venice documents it in its Chat Completions API.
In a chat model's Advanced Settings → Additional parameters, set a0_prompt_cache_mode=prompt_cache_key or a0_prompt_cache_mode=cache_control to match a custom provider's supported format. auto keeps the existing provider detection; off disables automatic hints. Explicit keys, including extra_body keys, remain unchanged. Marker mode keeps marked requests on Chat Completions and uses the existing stable message/tool boundaries.
Initial Presets And Migration
The startup migration converts the prior global full model configuration into Default, preserves existing global presets, promotes distinct scoped full configurations and legacy project presets into uniquely named global presets, and rewrites scoped config files to selection-only JSON. Original files receive a .pre-unified-presets.bak backup before replacement; the migration is idempotent.
At every startup, legacy migration runs first. If usr/plugins/_model_config/presets.yaml then exists, initialization returns immediately without network access. Otherwise the plugin makes one bounded request for agent0ai/a0-presets/model_presets.yaml, validates the whole collection, removes secret fields, and saves it locally. A download, parse, or validation failure processes and saves plugin-local mode_presets_fallback.yaml through the same validation path, so initialization remains usable offline.
Key Files
helpers/model_config.pyowns preset validation, resolution, compatibility config shapes, and runtime model construction.api/model_presets.pyowns global preset editing, scoped selection, and reference repair after rename/delete/reset.extensions/python/startup_migration/_10_migrate_model_config.pyowns legacy conversion.extensions/python/startup_migration/_20_bootstrap_model_presets.pyowns missing-collection initialization and plugin-local fallback.webui/preset-overview.htmlis the shared Settings/plugin-settings summary widget.webui/main.htmlis the preset editor.
Plugin Metadata
- Name:
_model_config - Settings section:
agent - Per-project config:
true - Per-agent config:
true - Always enabled:
true