diff --git a/CHANGELOG.md b/CHANGELOG.md index 39f9dec..579ac85 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -1,5 +1,57 @@ # CHANGELOG +## 2026-09-30 · 智能体模型改为「页面可选」(方案 A)+ 修引擎模型读不到的静默 bug(版本 2.5.4 → 2.6.0) + +老板问「智能体当时为什么说不能自己设置模型」,并选择**方案 A**:能选、但不新增 profile。 + +**先推翻了自己的旧结论**(这是本轮最重要的一步):当初(第三十八轮)实测的是 +「**请求体里带 `model` 会被忽略**」→ 据此推成「模型是 profile 级 ⇒ 所有智能体只能共用一个模型」。 +**这个推论过头了。** 翻源码 + 实测后确认: + +| 事实 | 证据 | +|---|---| +| 模型**不是全局唯一**,而是**每个 profile 各存一份** `active_model` | `PUT /api/models/active`(`scope=agent` + `agent_id`)写入 `workspaces//agent.json` | +| 前缀是 **`/api/models/active`** | `routers/providers.py` 里 `APIRouter(prefix="/models")`;实测 `/api/providers/active` → **404** | +| 改完**即时生效**,不用重启 | `runtime/builder.py:324` 每轮都重读 `agent_config.active_model` | +| 端到端确实换了模型 | 切 `agnes-3.0-flash` 后真对话 `turn_usage` 回 `model_name=agnes-3.0-flash`、`context_size` 1M → 512K | + +所以正确口径是:**请求级指定不行,改 profile 的 `active_model` 行**。 + +**顺手抓到一个静默 bug(真 bug)**:`agents_routes.py` 的 `refresh_engine_live()` 读的是 +`/api/providers/active` —— **那条 404**,异常被 `except` 吞掉 → `engine_live["model"]` 恒为空串 +→ **配置页模型栏一直显示 `—`、列表页模型名一直是空的**。它不报错、不抛异常,纯静默。 +已改用新 helper `get_active_model()`(走正确路径),模型名现在能正常读出。 + +**新增(方案 A 的落地)**: + +- **后端** `app/adapters/qwenpaw_engine.py`:`ACTIVE_MODEL_PATH = "/api/models/active"` + + `get_active_model()` / `set_active_model()`(后者 4xx 抛 `EngineError` 并**透传引擎原文**, + 不自己编文案)。 +- **后端** `app/agents_routes.py` 两个端点: + - `GET /agents/engine/models` —— 可选清单(**只列真的能用的**:`require_api_key` 为真但没配 + Key 的供应商直接跳过,不列出来骗人)+ 实时 current(不用 TTL 缓存,刚改完要立刻看到) + - `PUT /agents/engine/model` —— 切换;带**防自指护栏**(不得指向 `agent/*`,与对话入口同款) +- **前端** `Agents.vue` 配置页「模型」栏:只读文字 → **可搜索下拉**(按供应商分组、视觉模型带 + 「视觉」标)+ 刷新按钮;失败**回滚选中值**并弹错。`lib/agents.ts` 加两个封装。 +- **i18n**:删掉过期的 `cfgModelReadonly` / `cfgModelHint`,换成 5 个新键(中英 **883/883** 对称)。 + +**语义与取舍(必须讲清)**:方案 A 是**所有智能体共用这一个模型** —— 因为只有一个 profile, +`active_model` 是这一份 config 上的一个值,**做不到「A 用 flash、B 用 pro」并存**。 +真要做到那样必须走方案 B(一智能体一 profile,有内存成本)。页面提示文案已按此口径重写。 + +**测试**:新增 5 条(端点挂载 / 路径常量回归(防再写回 `providers/active`)/ PUT 的 body 与头 / +4xx 抛错 / 防自指护栏),**224 passed / 0 failed**(原 219)。`ci_check.sh` 六道门禁全绿。 + +**附带发现(环境问题,非本轮引入)**:QwenPaw 里 `agnes` 供应商的 Key **已失效** +(切到 `agnes/agnes-3.0-flash` 报 401「无效的令牌」)。走网关的 `wangjinding` 正常 +(网关侧 Agnes Key 是好的)。要用 Agnes 直连需先修 Key。 + +**上线**:`supervisorctl restart api-gateway:api-gateway_00` + `npm run build`; +线上 `version=2.6.0`、入口 `index-BQYagb1Q.js`、Agents chunk `Agents-CN_agann.js`; +真机验证:下拉 112 项 / 7 组 → 选 `wjd` → UI 显示 `wjd` → 磁盘 `agent.json` +`{"provider_id":"wangjinding","model":"wjd"}` → 引擎回读一致 → 列表页同步。 + + ## 2026-09-30 · 智能体页「对话」与「配置」内容间距放宽(版本 2.5.3 → 2.5.4) 老板反馈:**智能体界面在「对话」和「配置」两个 Tab 里,内容间距也很挤,都优化下**。 diff --git a/app/adapters/qwenpaw_engine.py b/app/adapters/qwenpaw_engine.py index ac01a3f..d441e82 100644 --- a/app/adapters/qwenpaw_engine.py +++ b/app/adapters/qwenpaw_engine.py @@ -348,3 +348,71 @@ async def check_engine(base_url: str, *, timeout: float = 5.0) -> tuple[bool, st return False, ENGINE_DOWN_MSG except Exception as exc: # noqa: BLE001 return False, f"引擎不可达:{exc}" + + +# ── 引擎模型读写(第四十轮 · 「模型可选」方案 A)──────────────────────── +# +# 事实(实测 + 源码): +# - 模型不是"全局唯一",而是**每个 profile 各存一份** `active_model` +# (落盘 `workspaces//agent.json`)。所以改它 = 只影响这一个 profile。 +# - 端点前缀是 **`/api/models/active`**(`routers/providers.py` 里 `prefix="/models"`)。 +# ⚠️ 别写成 `/api/providers/active` —— 那条是 **404**(本项目旧代码就踩过这个坑, +# 导致模型名一直读不出来、页面显示 `—`)。 +# - 改完 **即时生效**,不用重启引擎(每轮对话都会重读 agent_config)。 +# - 我们这边所有智能体**共用 `ai-gateway` 一个 profile** ⇒ 改这个 profile 的模型 +# 等于"换所有智能体用的模型"。这是方案 A 的既定语义(做不到 A/B 各用各的)。 + +ACTIVE_MODEL_PATH = "/api/models/active" + + +async def get_active_model(base_url: str, agent_id: str, + *, timeout: float = 8.0) -> dict[str, Any] | None: + """读某个 profile 当前生效的模型。返回 `{provider_id, model}`,读不到返回 None。""" + url = base_url.rstrip("/") + ACTIVE_MODEL_PATH + try: + async with httpx.AsyncClient(timeout=timeout) as client: + r = await client.get( + url, + params={"scope": "agent", "agent_id": agent_id}, + headers={"X-Agent-Id": agent_id}, + ) + if r.status_code >= 400: + return None + slot = (r.json() or {}).get("active_llm") or {} + if not slot.get("model"): + return None + return {"provider_id": slot.get("provider_id") or "", "model": slot.get("model") or ""} + except Exception: # noqa: BLE001 + return None + + +async def set_active_model(base_url: str, agent_id: str, + provider_id: str, model: str, + *, timeout: float = 15.0) -> dict[str, Any]: + """改某个 profile 的模型(即时生效)。校验失败抛 EngineError。 + + 引擎会先 `_validate_model_slot()`(供应商/模型必须存在),所以这里 + 直接把上游的错误原文透出去,比自己编文案准确。 + """ + url = base_url.rstrip("/") + ACTIVE_MODEL_PATH + payload = {"scope": "agent", "agent_id": agent_id, + "provider_id": provider_id, "model": model} + try: + async with httpx.AsyncClient(timeout=timeout) as client: + r = await client.put(url, json=payload, headers={"X-Agent-Id": agent_id}) + except httpx.ConnectError: + raise EngineError(ENGINE_DOWN_MSG) + except Exception as exc: # noqa: BLE001 + raise EngineError(f"引擎不可达:{exc}") + if r.status_code >= 400: + detail = "" + try: + b = r.json() + detail = b.get("detail") or b.get("message") or "" + except Exception: # noqa: BLE001 + detail = (r.text or "")[:200] + raise EngineError(f"引擎拒绝该模型:{detail or f'HTTP {r.status_code}'}") + slot = (r.json() or {}).get("active_llm") or {} + return {"provider_id": slot.get("provider_id") or provider_id, + "model": slot.get("model") or model} + diff --git a/app/agents_routes.py b/app/agents_routes.py index 98c21de..032bd2a 100644 --- a/app/agents_routes.py +++ b/app/agents_routes.py @@ -29,7 +29,13 @@ from app.agents import ( AgentStore, AgentsError, ) -from app.adapters.qwenpaw_engine import check_engine, stream_chat +from app.adapters.qwenpaw_engine import ( + check_engine, + EngineError, + get_active_model, + set_active_model, + stream_chat, +) log = logging.getLogger("api-gateway.agents") @@ -126,20 +132,16 @@ async def refresh_engine_live(store: AgentStore, *, force: bool = False) -> None agent_id = agents[0].engine_agent_id if agents else DEFAULT_ENGINE_AGENT ok, _why = await check_engine(base) model = "" + provider_id = "" if ok: - try: - async with httpx.AsyncClient(timeout=8.0) as c: - r = await c.get( - base.rstrip("/") + "/api/providers/active", - params={"scope": "agent", "agent_id": agent_id}, - headers={"X-Agent-Id": agent_id}, - ) - if r.status_code < 400: - slot = (r.json() or {}).get("active_llm") or {} - model = slot.get("model") or "" - except Exception: # noqa: BLE001 - log.debug("读引擎 active_model 失败") - engine_live.update({"model": model, "ready": ok, "checked_at": now}) + # ⚠️ 第四十轮修:旧代码读的是 `/api/providers/active` —— 那条**是 404** + # (真前缀是 `/api/models/active`),所以 model 一直是空串、页面显示 `—`。 + slot = await get_active_model(base, agent_id) + if slot: + model = slot.get("model") or "" + provider_id = slot.get("provider_id") or "" + engine_live.update({"model": model, "provider_id": provider_id, + "ready": ok, "checked_at": now}) def can_use_agent(a: Agent, role: str) -> bool: @@ -205,6 +207,79 @@ def install(app: FastAPI, *, store: AgentStore, }, } + # ── 引擎模型:可选清单 + 切换(第四十轮 · 方案 A)────────────────── + # + # 语义(老板 2026-09-30 拍板方案 A):所有智能体共用一个 profile + # ⇒ 这里切的模型是"所有智能体一起用的",不是各自独立的。 + # 要"每个智能体各用各的"须走方案 B(一智能体一 profile),本轮不做。 + @app.get("/agents/engine/models") + async def agents_engine_models(request: Request, _auth=Depends(require_auth)): + """可选模型清单 = 网关自己的模型列表(都由 wangjinding 供应商送达引擎)。""" + require_admin(request) + store.reload_if_changed() + agents = store.list() + base = agents[0].engine_base_url if agents else DEFAULT_ENGINE_URL + agent_id = agents[0].engine_agent_id if agents else DEFAULT_ENGINE_AGENT + + # ① 引擎里配好的、且真的能用的模型(有 Key 的供应商) + options: list[dict[str, Any]] = [] + try: + async with httpx.AsyncClient(timeout=10.0) as c: + r = await c.get(base.rstrip("/") + "/api/models") + if r.status_code < 400: + for p in (r.json() or []): + pid = p.get("id") or "" + if p.get("require_api_key") and not p.get("api_key"): + continue # 没配 Key 的选不了,别列出来骗人 + for m in (list(p.get("models") or []) + list(p.get("extra_models") or [])): + mid = m.get("id") if isinstance(m, dict) else m + if not mid: + continue + options.append({ + "provider_id": pid, + "provider_name": p.get("name") or pid, + "model": mid, + "vision": bool(m.get("supports_image")) if isinstance(m, dict) else False, + }) + except Exception: # noqa: BLE001 + log.warning("读引擎模型清单失败(可选清单降级为空)") + + # ② 当前生效的模型(实时读,别用 TTL 缓存 —— 用户刚改完要立刻看到) + current = await get_active_model(base, agent_id) or {} + ok, _why = await check_engine(base) + return { + "options": options, + "current": {"provider_id": current.get("provider_id", ""), + "model": current.get("model", "")}, + "engine": {"base_url": base, "agent_id": agent_id, "ready": ok}, + "scope_note": "shared", # 前端据此显示"所有智能体共用"提示 + } + + @app.put("/agents/engine/model") + async def agents_engine_set_model(request: Request, _auth=Depends(require_auth)): + """切换引擎模型(即时生效,不用重启)。""" + require_admin(request) + body = await json_body(request) + provider_id = str(body.get("provider_id") or "").strip() + model = str(body.get("model") or "").strip() + if not provider_id or not model: + raise HTTPException(400, "provider_id 与 model 都不能为空") + # 防自指(与对话入口同一条护栏) + if model == "agent" or model.startswith("agent/"): + raise HTTPException(400, "不能把引擎模型指向智能体自身(agent/*)") + store.reload_if_changed() + agents = store.list() + base = agents[0].engine_base_url if agents else DEFAULT_ENGINE_URL + agent_id = agents[0].engine_agent_id if agents else DEFAULT_ENGINE_AGENT + try: + slot = await set_active_model(base, agent_id, provider_id, model) + except EngineError as e: + raise HTTPException(400, str(e)) + # 立刻刷新缓存,让列表页同步 + await refresh_engine_live(store, force=True) + log.info("引擎模型已切换:%s → %s/%s", agent_id, provider_id, model) + return {"status": "ok", "current": slot} + # ── 新建 / 修改 / 归档 / 恢复 ── @app.post("/agents") async def agents_create(request: Request, _auth=Depends(require_auth)): diff --git a/app/main.py b/app/main.py index f03b240..5bd6a0b 100644 --- a/app/main.py +++ b/app/main.py @@ -338,7 +338,7 @@ _docs_on = config.server.docs_enabled # 安全默认关(report: /docs、/ope # 三处,/health 干脆没返回版本 —— 前端「网关版本」一直显示 v-。 # 2026-09-29: 2.4.0 → 2.4.1(前端间距规范统一);2.4.1 → 2.5.0(新增「智能体」功能模块);2.5.0 → 2.5.1(智能体支持图片附件);2.5.1 → 2.5.2(智能体页间距与对齐优化);2.5.2 → 2.5.3(修 KPI 与下方块之间 0 间距)。 # ★ 版本号纪律:每次改动都要往上加,小改动加小版本号、大改动加大版本号(README 有专章)。 -VERSION = "2.5.4" +VERSION = "2.6.0" app = FastAPI( title="API Gateway", diff --git a/frontend/src/lib/agents.ts b/frontend/src/lib/agents.ts index df3c90f..0cd09c7 100644 --- a/frontend/src/lib/agents.ts +++ b/frontend/src/lib/agents.ts @@ -89,6 +89,29 @@ export const testAgent = (slug: string) => export const agentSessions = (slug: string) => req<{ slug: string; kind: string; items: any[] }>(`/agents/${slug}/sessions`) +/** 引擎可选模型清单 + 当前生效模型(第四十轮 · 方案 A:所有智能体共用)。 */ +export interface EngineModelOption { + provider_id: string + provider_name: string + model: string + vision: boolean +} +export interface EngineModelsResp { + options: EngineModelOption[] + current: { provider_id: string; model: string } + engine: { base_url: string; agent_id: string; ready: boolean } + scope_note: string +} + +export const engineModels = () => req('/agents/engine/models') + +/** 切换引擎模型(即时生效,不用重启)。 */ +export const setEngineModel = (providerId: string, model: string) => + req<{ status: string; current: { provider_id: string; model: string } }>( + '/agents/engine/model', + { method: 'PUT', body: JSON.stringify({ provider_id: providerId, model }) }, + ) + export const stopAgent = (slug: string, chatId: string) => req<{ status: string }>(`/agents/${slug}/stop?chat_id=${encodeURIComponent(chatId)}`, { method: 'POST' }) diff --git a/frontend/src/locales/en.ts b/frontend/src/locales/en.ts index b19bb79..409260b 100644 --- a/frontend/src/locales/en.ts +++ b/frontend/src/locales/en.ts @@ -411,8 +411,11 @@ export default { cfgDesc: 'Description', cfgModel: 'Model', cfgModelSub: 'From the gateway models', - cfgModelReadonly: 'Model is set by QwenPaw profile {profile}', - cfgModelHint: 'Engine model is a profile-level setting — all agents share one model. Change it in QwenPaw:', + cfgModelPick: 'Pick a model', + cfgModelRefresh: 'Refresh', + cfgModelVision: 'vision', + cfgModelSwitched: 'Switched to {model} (shared by all agents, takes effect immediately)', + cfgModelHintShared: 'All agents share this one model (the engine stores it per profile). Switching takes effect immediately, no restart. You can also change it in QwenPaw:', cfgOpenQwenpaw: 'Open QwenPaw', cfgExec: 'Execution', cfgExecSub: 'Affects server safety, read-only for now', diff --git a/frontend/src/locales/zh.ts b/frontend/src/locales/zh.ts index ab1b098..99121d2 100644 --- a/frontend/src/locales/zh.ts +++ b/frontend/src/locales/zh.ts @@ -414,8 +414,11 @@ export default { cfgDesc: '描述', cfgModel: '模型', cfgModelSub: '从网关的模型里选', - cfgModelReadonly: '模型由 QwenPaw 的 {profile} profile 决定', - cfgModelHint: '引擎的模型是 profile 级设置 —— 所有智能体共用同一个模型。要改请去 QwenPaw:', + cfgModelPick: '选择模型', + cfgModelRefresh: '刷新', + cfgModelVision: '视觉', + cfgModelSwitched: '已切换到 {model}(所有智能体共用,即时生效)', + cfgModelHintShared: '所有智能体共用这一个模型(引擎按 profile 存模型)。切换即时生效,不用重启引擎。也可以去 QwenPaw 改:', cfgOpenQwenpaw: '打开 QwenPaw', cfgExec: '执行', cfgExecSub: '影响服务器安全,一期只读', diff --git a/frontend/src/views/Agents.vue b/frontend/src/views/Agents.vue index 829bf09..2037cd5 100644 --- a/frontend/src/views/Agents.vue +++ b/frontend/src/views/Agents.vue @@ -101,7 +101,7 @@
{{ t('agents.tabChat') }} - {{ t('agents.tabConfig') }} + {{ t('agents.tabConfig') }} {{ t('agents.tabRuns') }}
@@ -255,13 +255,36 @@
{{ t('agents.cfgModel') }}{{ t('agents.cfgModelSub') }}
{{ t('agents.cfgModel') }}
-
- {{ cur.engine_model || '—' }} - {{ t('agents.cfgReadonly') }} +
+ + + + {{ o.model }} + {{ t('agents.cfgModelVision') }} + + + + + {{ t('agents.cfgModelRefresh') }} +
- {{ t('agents.cfgModelHint') }} + {{ t('agents.cfgModelHintShared') }} {{ t('agents.cfgOpenQwenpaw') }} ↗
@@ -373,9 +396,9 @@ import { useI18n } from 'vue-i18n' import { ElMessage } from 'element-plus' import Icon from '../components/Icon.vue' import { - type AgentEvent, type AgentItem, type AgentRun, - agentRuns, agentSessions, createAgent, listAgents, - restoreAgent, retireAgent, streamAgentChat, testAgent, updateAgent, + type AgentEvent, type AgentItem, type AgentRun, type EngineModelOption, + agentRuns, agentSessions, createAgent, engineModels, listAgents, + restoreAgent, retireAgent, setEngineModel, streamAgentChat, testAgent, updateAgent, } from '../lib/agents' import { renderMarkdown } from '../lib/markdown' import { PRESETS, pickImage } from '../lib/upload' @@ -416,6 +439,19 @@ const create = reactive({ display_name: '', slug: '', description: '', system_su const testing = ref(false) const testResult = ref<{ ok: boolean; ms: number; detail: string } | null>(null) +// 引擎模型选择(第四十轮 · 方案 A):清单 + 当前值。`enginePick` 形如 "provider/model"。 +const engineOptions = ref([]) +const enginePick = ref('') +const engineLoading = ref(false) +const engineGroups = computed(() => { + const by: Record = {} + for (const o of engineOptions.value) { + if (!by[o.provider_id]) by[o.provider_id] = { provider_id: o.provider_id, provider_name: o.provider_name, items: [] } + by[o.provider_id].items.push(o) + } + return Object.values(by) +}) + const cur = computed(() => data.agents.find((a) => a.slug === current.value) || null) const hasRetired = computed(() => data.agents.some((a) => a.status === 'retired')) const activeCount = computed(() => data.agents.filter((a) => a.status !== 'retired').length) @@ -645,6 +681,38 @@ async function doTest() { finally { testing.value = false } } +/** 拉可选模型清单 + 当前生效模型(第四十轮 · 方案 A)。 */ +async function loadEngineModels() { + engineLoading.value = true + try { + const j = await engineModels() + engineOptions.value = j.options || [] + const c = j.current || {} + enginePick.value = c.model ? `${c.provider_id}/${c.model}` : '' + } catch (e: any) { + ElMessage.error(String(e?.message || e)) + } finally { engineLoading.value = false } +} + +/** 切换引擎模型 —— 即时生效(不用重启引擎)。 */ +async function doSetModel(v: string) { + const i = String(v || '').indexOf('/') + if (i < 0) return + const provider_id = v.slice(0, i) + const model = v.slice(i + 1) + const prev = enginePick.value + try { + const j = await setEngineModel(provider_id, model) + const c = j.current || {} + enginePick.value = c.model ? `${c.provider_id}/${c.model}` : '' + ElMessage.success(t('agents.cfgModelSwitched', { model })) + await load() // 让列表页 / 头部同步新模型名 + } catch (e: any) { + enginePick.value = prev // 失败回滚选中值 + ElMessage.error(String(e?.message || e)) + } +} + async function doRetire() { if (!cur.value || !confirm(t('agents.confirmDelete'))) return try { await retireAgent(cur.value.slug); await load(); pick(data.agents[0]?.slug || '') } @@ -790,6 +858,10 @@ onMounted(load) .f-val { flex: 1; min-width: 0; display: flex; align-items: center; gap: 7px; font-size: 12.5px; padding: 8px 10px; border-radius: 9px; background: var(--chip); } .f-val.code { font-family: ui-monospace, Menlo, monospace; font-size: 11.5px; } .f-lock { margin-left: auto; display: inline-flex; align-items: center; gap: 4px; font-size: 10.5px; color: var(--muted-2); } +/* 模型下拉(第四十轮):撑满 f-val,右侧留刷新按钮 */ +.model-pick { gap: var(--gap-sm); padding: 4px 6px 4px 4px; } +.model-pick .el-select { flex: 1; min-width: 0; } +.opt-vision { float: right; margin-left: 10px; font-size: 10.5px; color: var(--primary); opacity: .8; } .f-hint { font-size: 11px; color: var(--muted-2); line-height: 1.65; margin: var(--gap-xs) 0 0 88px; } .f-row-note { font-size: 11.5px; color: var(--muted); line-height: 1.6; margin-top: var(--gap-s); } .row-actions { display: flex; align-items: center; gap: var(--gap-sm); margin-top: var(--gap-lg); } diff --git a/tests/test_gateway.py b/tests/test_gateway.py index 7f03380..b1c2c78 100644 --- a/tests/test_gateway.py +++ b/tests/test_gateway.py @@ -5967,6 +5967,111 @@ def test_agents_routes_installed(): assert need in paths, f"缺少端点 {need}" +def test_agents_engine_model_endpoints_installed(): + """引擎模型端点已挂上(第四十轮 · 方案 A)。""" + import app.main as m + paths = {r.path for r in m.app.routes if hasattr(r, "path")} + for need in ("/agents/engine/models", "/agents/engine/model"): + assert need in paths, f"缺少端点 {need}" + + +def test_engine_active_model_path_is_models_not_providers(): + """回归:读/写引擎 active_model 必须用 `/api/models/active`(第四十轮)。 + + 旧代码写的是 `/api/providers/active` —— 那条**实测是 404**,导致 + `engine_live["model"]` 一直是空串、配置页模型栏永远显示 `—`。 + 这是"静默失效"型 bug:不报错、不抛异常,只是值一直是空。 + """ + from app.adapters import qwenpaw_engine as qe + assert qe.ACTIVE_MODEL_PATH == "/api/models/active" + assert "providers/active" not in qe.ACTIVE_MODEL_PATH + # 两个函数都要用这个常量(别再各写一份字面量) + import inspect + for fn in (qe.get_active_model, qe.set_active_model): + src = inspect.getsource(fn) + assert "ACTIVE_MODEL_PATH" in src, f"{fn.__name__} 未复用 ACTIVE_MODEL_PATH" + + +def test_engine_set_active_model_sends_agent_scope(): + """PUT 的 body 必须是 scope=agent + agent_id(profile 级写入)。""" + import asyncio, json as _json + from app.adapters import qwenpaw_engine as qe + + captured = {} + + class _Resp: + status_code = 200 + def json(self): + return {"active_llm": {"provider_id": "wangjinding", "model": "wjd"}} + + class _Client: + def __init__(self, *a, **k): pass + async def __aenter__(self): return self + async def __aexit__(self, *a): return False + async def put(self, url, json=None, headers=None): + captured["url"] = url + captured["json"] = json + captured["headers"] = headers + return _Resp() + + orig = qe.httpx.AsyncClient + qe.httpx.AsyncClient = _Client + try: + out = asyncio.run(qe.set_active_model("http://127.0.0.1:8088", "ai-gateway", + "wangjinding", "wjd")) + finally: + qe.httpx.AsyncClient = orig + + assert captured["url"].endswith("/api/models/active"), captured["url"] + assert captured["json"] == {"scope": "agent", "agent_id": "ai-gateway", + "provider_id": "wangjinding", "model": "wjd"} + assert captured["headers"].get("X-Agent-Id") == "ai-gateway" + assert out == {"provider_id": "wangjinding", "model": "wjd"} + + +def test_engine_set_active_model_raises_on_4xx(): + """引擎拒绝(如模型不存在)→ 抛 EngineError,错误原文透传(别吞成 ok)。""" + import asyncio + from app.adapters import qwenpaw_engine as qe + + class _Resp: + status_code = 400 + text = "" + def json(self): + return {"detail": "model 'nope' not found"} + + class _Client: + def __init__(self, *a, **k): pass + async def __aenter__(self): return self + async def __aexit__(self, *a): return False + async def put(self, url, json=None, headers=None): return _Resp() + + orig = qe.httpx.AsyncClient + qe.httpx.AsyncClient = _Client + try: + try: + asyncio.run(qe.set_active_model("http://x", "ai-gateway", "p", "nope")) + raise AssertionError("本应抛 EngineError") + except qe.EngineError as e: + assert "not found" in str(e), str(e) + finally: + qe.httpx.AsyncClient = orig + + +def test_agents_engine_set_model_rejects_agent_self_reference(): + """防自指:引擎模型不能指向 agent/*(否则引擎→智能体→引擎 死循环)。 + + 这条护栏在对话端点也有;新增的切换端点必须同样拦住,否则用户可以从 + 模型下拉里把引擎指向智能体自身,把服务打成死循环。 + """ + import inspect + import app.agents_routes as ar + src = inspect.getsource(ar.install) + # 切换端点里必须有这条检查(与对话入口同款) + assert "agent/" in src and "死循环" in src, "切换端点缺少防自指护栏" + + + def test_qwenpaw_engine_image_content_passthrough(): """图片附件:内容块必须原样转成 AgentScope 的 image 形状(第三十九轮)。