@nocobase/app-plugin-agents 0.1.0-beta.1 → 0.1.0-beta.10

package/dist/shared/presets.d.ts

@@ -1,30 +0,0 @@
-/**
- * Agent presets: roles people can start a new agent from. A preset only fills the new agent's form (its description,
- * instructions and recommended business actions); once created, the agent is like any other and every field can be
- * changed. The application registers its presets on the server (`agents.presets.register`); this plugin ships none.
- *
- * `GET agents/presets` (`agents.agents` read) answers `{ data: AgentPreset[], meta: { total } }`.
- */
-import type { AgentType } from './agents.js';
-import type { I18nText } from './i18n.js';
-export declare const PRESETS_ROUTE = "agents/presets";
-export interface AgentPreset {
-    /** Stable, such as `reviewer`. */
-    readonly key: string;
-    /** English; the panel shows `nameText` in the viewer's language when it has it. */
-    readonly name: string;
-    /** What the agent is good at, as the agent's description; English, translated through `descriptionText`. */
-    readonly description: string;
-    readonly nameText: I18nText;
-    readonly descriptionText: I18nText;
-    /** The agent's instructions, in English: they are prompt text, not product copy. */
-    readonly instructions: string;
-    /** The type the new-agent dialog starts it as; the person may still pick the other. `runner` when absent. */
-    readonly type?: AgentType;
-    /**
-     * The business actions to tick, among those the application offers now (`GET agents/actions`): a preset never
-     * recommends an action nobody provides.
-     */
-    readonly actions: readonly string[];
-}
-//# sourceMappingURL=presets.d.ts.map

package/dist/shared/presets.d.ts.map

@@ -1 +0,0 @@
-{"version":3,"file":"presets.d.ts","sourceRoot":"","sources":["../../shared/presets.ts"],"names":[],"mappings":"AAAA;;;;;;GAMG;AACH,OAAO,KAAK,EAAE,SAAS,EAAE,MAAM,aAAa,CAAC;AAC7C,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,WAAW,CAAC;AAE1C,eAAO,MAAM,aAAa,mBAAmB,CAAC;AAE9C,MAAM,WAAW,WAAW;IAC1B,kCAAkC;IAClC,QAAQ,CAAC,GAAG,EAAE,MAAM,CAAC;IACrB,mFAAmF;IACnF,QAAQ,CAAC,IAAI,EAAE,MAAM,CAAC;IACtB,4GAA4G;IAC5G,QAAQ,CAAC,WAAW,EAAE,MAAM,CAAC;IAC7B,QAAQ,CAAC,QAAQ,EAAE,QAAQ,CAAC;IAC5B,QAAQ,CAAC,eAAe,EAAE,QAAQ,CAAC;IACnC,oFAAoF;IACpF,QAAQ,CAAC,YAAY,EAAE,MAAM,CAAC;IAC9B,6GAA6G;IAC7G,QAAQ,CAAC,IAAI,CAAC,EAAE,SAAS,CAAC;IAC1B;;;OAGG;IACH,QAAQ,CAAC,OAAO,EAAE,SAAS,MAAM,EAAE,CAAC;CACrC"}

package/dist/shared/presets.js

@@ -1,2 +0,0 @@
-export const PRESETS_ROUTE = 'agents/presets';
-//# sourceMappingURL=presets.js.map

package/dist/shared/presets.js.map

@@ -1 +0,0 @@
-{"version":3,"file":"presets.js","sourceRoot":"","sources":["../../shared/presets.ts"],"names":[],"mappings":"AAUA,MAAM,CAAC,MAAM,aAAa,GAAG,gBAAgB,CAAC"}

package/skills/nocobase-app-plugin-agents/references/online-agents.md

@@ -1,31 +0,0 @@
-# Online agents and model services
-
-An online agent talks to a model of a model service and is run by the App itself, on every instance (`agents.server.enabled`), through the same claim as runner runs: it needs no runner and answers in seconds. Its tools are a sandboxed in-process shell (`bash`, no real processes or network) carrying the App's CLI, and `skill`, which reads one of the run's skills. The README's "What it does" (Online runs) and "Online agents" sections are the reference.
-
-## Model services
-
-- A service is a provider type (`MODEL_PROVIDERS` in `@nocobase/app-plugin-agents/shared/models`: OpenAI, Anthropic, Google, DeepSeek, Alibaba, Moonshot, Cohere, Ollama, and OpenAI-compatible endpoints), a name, a base URL, a write-only API key sealed with the secrets service, an on/off switch and the models it offers. OpenCode Zen and Go serve each model family through its own API under one base URL: DeepSeek, GLM, Kimi and MiMo through a DeepSeek or OpenAI-compatible service, Claude, MiniMax and Qwen through an Anthropic one, GPT and Grok through an OpenAI one; a call to an OpenCode base URL sends `x-opencode-session` automatically, one session id per conversation. Each model has a kind: `chat` (what agents talk to), `embedding` or `rerank`.
-- Services are added on the Models page (`modelsRoute`, behind `agents.services`) or through the API (`service create|discover|check` on the CLI). They are never configured in `config.yml`.
-- The system default chat model (Models page, `PUT /api/agents/defaultModels/chat`) is what an online agent that lists no model answers with. With none set, the first chat model an enabled service offers is used.
-- An online agent lists ordered entries (service, model, effort); its first is the default. A conversation may pick another of the agent's entries. A failing model fails the run; nothing falls back to another entry.
-- `GET /api/agents/models` (anyone signed in) lists what agents may use; `POST /api/agents/checkModel` (who reads agents) tests one with a short call and never throws.
-
-## The model gateway
-
-`online.gateway` on `agentsToken` serves the App's own model calls outside runs: `catalog(kind)`, `check(ref)`, `embed({ model, values, source })`, `rerank({ model, query, documents, topN?, source })` and `generate({ model, system?, prompt, maxOutputTokens?, source })`. Each names who asks (`source`), is recorded in `agModelUsage` and priced like runs, and throws `ModelError` with a code (`config`, `auth`, `quota`, `rateLimit`, `network`, `contextOverflow`, `filtered`, `badResponse`, `aborted`, `unknown`). Use it rather than calling a provider SDK, so keys, usage and prices stay in one place.
-
-`online.skills.register(skill)` adds a skill every online run gets besides its own, typically the App's CLI skill.
-
-## Commands in an online run
-
-The run's shell carries the App's CLI, built from the command manifest for the run (the actions of the waking person that the agent is configured with). Each command is one request through the App with the run's token, so the route checks the permission again. Commands that send or save files are refused. A route therefore needs `runToken` in its `security` and an `x-cli` `action` the gate allows before an online agent can call it.
-
-Run limits: `agents.server.maxSteps` model calls (then `stepLimit`); a cancel aborts within a second; at shutdown an instance hands its runs back (`runnerOffline`).
-
-## Consultations
-
-An online run may ask another online agent a question (`ask_agent`), synchronously, in a read-only child run (`parentRunId`) bounded by `agents.server.consult.timeoutMs` and `tokenBudget`, at most `CONSULT_MAX_DEPTH` deep. The gate keeps a consultation to reading actions plus those `ActionGate.consultable` allows. See the README's "Consultations" section.
-
-## Vectors
-
-`vectors.collection(spec)` gives the App semantic search over items it names, embedded with the model it chooses, in the store `agents.vectors` configures (`sqlite-vec` by default, `pgvector` for several instances or Alpine). `status()` answers availability and a `VectorUnavailableCode` when it is not available; then search answers null and the caller keeps its keyword search. See the README's "Vectors" section.