feat: feed intelligence sources into analysis contexts

.env.example

@@ -348,6 +348,12 @@ SEARXNG_PUBLIC_INSTANCES_ENABLED=true
 # NEWS_STRATEGY_PROFILE=short
 # 新闻最大时效（天），搜索时限制结果在近期内，避免使用过时信息
 # NEWS_MAX_AGE_DAYS=3
+# 本地资讯/情报池保留天数；只清理资讯池 intelligence_items，不影响历史报告
+# NEWS_INTEL_RETENTION_DAYS=30
+# 单个 RSS/Atom 资讯源拉取超时（秒）
+# NEWS_INTEL_FETCH_TIMEOUT_SEC=8
+# 单次每个资讯源最多采集条数
+# NEWS_INTEL_MAX_ITEMS_PER_SOURCE=50
 # 乖离率阈值（%），偏离 MA5 超过此值提示不追高；强势趋势股自动放宽到 1.5 倍
 # BIAS_THRESHOLD=5.0
 

api/v1/endpoints/intelligence.py

@@ -0,0 +1,114 @@
+# -*- coding: utf-8 -*-
+"""Intelligence source API endpoints."""
+
+from __future__ import annotations
+
+import logging
+from typing import Optional
+
+from fastapi import APIRouter, HTTPException, Query
+
+from api.v1.schemas.common import ErrorResponse
+from api.v1.schemas.intelligence import (
+    IntelligenceFetchResponse,
+    IntelligenceItemListResponse,
+    IntelligenceSourceCreateRequest,
+    IntelligenceSourceItem,
+    IntelligenceSourceListResponse,
+    IntelligenceSourceTestResponse,
+)
+from src.services.intelligence_service import IntelligenceService, IntelligenceServiceError
+
+logger = logging.getLogger(__name__)
+router = APIRouter()
+
+
+def _bad_request(exc: Exception) -> HTTPException:
+    return HTTPException(status_code=400, detail={"error": "validation_error", "message": str(exc)})
+
+
+def _not_found(message: str) -> HTTPException:
+    return HTTPException(status_code=404, detail={"error": "not_found", "message": message})
+
+
+def _internal_error(message: str, exc: Exception) -> HTTPException:
+    logger.error("%s: %s", message, exc, exc_info=True)
+    return HTTPException(status_code=500, detail={"error": "internal_error", "message": f"{message}: {str(exc)}"})
+
+
+@router.post("/sources", response_model=IntelligenceSourceItem, responses={400: {"model": ErrorResponse}, 500: {"model": ErrorResponse}}, summary="Create intelligence source")
+def create_source(request: IntelligenceSourceCreateRequest) -> IntelligenceSourceItem:
+    try:
+        return IntelligenceSourceItem(**IntelligenceService().create_source(request.model_dump()))
+    except IntelligenceServiceError as exc:
+        raise _bad_request(exc)
+    except Exception as exc:
+        raise _internal_error("Create intelligence source failed", exc)
+
+
+@router.get("/sources", response_model=IntelligenceSourceListResponse, responses={500: {"model": ErrorResponse}}, summary="List intelligence sources")
+def list_sources(
+    enabled: Optional[bool] = Query(None),
+    source_type: Optional[str] = Query(None),
+    scope_type: Optional[str] = Query(None),
+    market: Optional[str] = Query(None),
+    page: int = Query(1, ge=1),
+    page_size: int = Query(50, ge=1, le=100),
+) -> IntelligenceSourceListResponse:
+    try:
+        return IntelligenceSourceListResponse(**IntelligenceService().list_sources(
+            enabled=enabled, source_type=source_type, scope_type=scope_type,
+            market=market, page=page, page_size=page_size,
+        ))
+    except Exception as exc:
+        raise _internal_error("List intelligence sources failed", exc)
+
+
+@router.post("/sources/test", response_model=IntelligenceSourceTestResponse, responses={400: {"model": ErrorResponse}, 500: {"model": ErrorResponse}}, summary="Dry-run an intelligence source payload")
+def test_source_payload(request: IntelligenceSourceCreateRequest) -> IntelligenceSourceTestResponse:
+    try:
+        return IntelligenceSourceTestResponse(**IntelligenceService().test_source(request.model_dump()))
+    except IntelligenceServiceError as exc:
+        raise _bad_request(exc)
+    except Exception as exc:
+        raise _internal_error("Test intelligence source failed", exc)
+
+
+@router.post("/sources/{source_id}/fetch", response_model=IntelligenceFetchResponse, responses={400: {"model": ErrorResponse}, 404: {"model": ErrorResponse}, 500: {"model": ErrorResponse}}, summary="Fetch one intelligence source")
+def fetch_source(source_id: int, dry_run: bool = Query(False)) -> IntelligenceFetchResponse:
+    try:
+        return IntelligenceFetchResponse(**IntelligenceService().fetch_source(source_id, dry_run=dry_run))
+    except IntelligenceServiceError as exc:
+        message = str(exc)
+        if "not found" in message.lower():
+            raise _not_found(message)
+        raise _bad_request(exc)
+    except Exception as exc:
+        raise _internal_error("Fetch intelligence source failed", exc)
+
+
+@router.post("/sources/fetch-enabled", response_model=IntelligenceFetchResponse, responses={500: {"model": ErrorResponse}}, summary="Fetch all enabled intelligence sources with fail-open semantics")
+def fetch_enabled_sources() -> IntelligenceFetchResponse:
+    try:
+        return IntelligenceFetchResponse(**IntelligenceService().fetch_enabled_sources())
+    except Exception as exc:
+        raise _internal_error("Fetch enabled intelligence sources failed", exc)
+
+
+@router.get("/items", response_model=IntelligenceItemListResponse, responses={500: {"model": ErrorResponse}}, summary="List persisted intelligence items")
+def list_items(
+    scope_type: Optional[str] = Query(None),
+    scope_value: Optional[str] = Query(None),
+    market: Optional[str] = Query(None),
+    query: Optional[str] = Query(None),
+    days: Optional[int] = Query(None, ge=1),
+    page: int = Query(1, ge=1),
+    page_size: int = Query(50, ge=1, le=100),
+) -> IntelligenceItemListResponse:
+    try:
+        return IntelligenceItemListResponse(**IntelligenceService().list_items(
+            scope_type=scope_type, scope_value=scope_value, market=market,
+            query=query, days=days, page=page, page_size=page_size,
+        ))
+    except Exception as exc:
+        raise _internal_error("List intelligence items failed", exc)

api/v1/router.py

@@ -21,6 +21,7 @@
     decision_signals,
     health,
     history,
+    intelligence,
     portfolio,
     stocks,
     system_config,
@@ -103,6 +104,12 @@
     tags=["AlphaSift"]
 )
 
+router.include_router(
+    intelligence.router,
+    prefix="/intelligence",
+    tags=["Intelligence"]
+)
+
 router.include_router(
     health.router,
     tags=["Health"]

api/v1/schemas/intelligence.py

@@ -0,0 +1,98 @@
+# -*- coding: utf-8 -*-
+"""Intelligence source API schemas."""
+
+from __future__ import annotations
+
+from typing import List, Literal, Optional
+
+from pydantic import BaseModel, Field
+
+SourceTypeValue = Literal["rss", "atom"]
+ScopeTypeValue = Literal["symbol", "market", "sector"]
+MarketValue = Literal["cn", "hk", "us", "global"]
+
+
+class IntelligenceSourceCreateRequest(BaseModel):
+    name: str = Field(..., min_length=1, max_length=100)
+    url: str = Field(..., min_length=1, max_length=1000)
+    source_type: SourceTypeValue = "rss"
+    enabled: bool = True
+    scope_type: ScopeTypeValue = "market"
+    scope_value: Optional[str] = Field(None, max_length=64)
+    market: MarketValue = "cn"
+    description: Optional[str] = None
+
+
+class IntelligenceSourceItem(BaseModel):
+    id: int
+    name: str
+    source_type: str
+    url: str
+    enabled: bool
+    scope_type: str
+    scope_value: Optional[str] = None
+    market: str
+    description: Optional[str] = None
+    last_status: Optional[str] = None
+    last_error: Optional[str] = None
+    last_fetched_at: Optional[str] = None
+    created_at: Optional[str] = None
+    updated_at: Optional[str] = None
+
+
+class IntelligenceSourceListResponse(BaseModel):
+    items: List[IntelligenceSourceItem] = Field(default_factory=list)
+    total: int
+    page: int
+    page_size: int
+
+
+class IntelligenceItem(BaseModel):
+    id: int
+    source_id: Optional[int] = None
+    source_name: Optional[str] = None
+    source_type: str
+    title: str
+    summary: Optional[str] = None
+    url: str
+    source: Optional[str] = None
+    published_at: Optional[str] = None
+    fetched_at: Optional[str] = None
+    scope_type: str
+    scope_value: Optional[str] = None
+    market: str
+
+
+class IntelligenceSampleItem(BaseModel):
+    title: str
+    summary: Optional[str] = None
+    url: str
+    source: Optional[str] = None
+    published_at: Optional[str] = None
+
+
+class IntelligenceItemListResponse(BaseModel):
+    items: List[IntelligenceItem] = Field(default_factory=list)
+    total: int
+    page: int
+    page_size: int
+
+
+class IntelligenceFetchResponse(BaseModel):
+    ok: bool
+    source_id: Optional[int] = None
+    source_count: Optional[int] = None
+    fetched_count: Optional[int] = None
+    saved_count: Optional[int] = None
+    retention_deleted: Optional[int] = None
+    dry_run: Optional[bool] = None
+    sample_items: List[IntelligenceSampleItem] = Field(default_factory=list)
+    results: Optional[List[dict]] = None
+    error: Optional[str] = None
+
+
+class IntelligenceSourceTestResponse(BaseModel):
+    ok: bool
+    source: dict
+    fetched_count: int
+    sample_items: List[IntelligenceSampleItem] = Field(default_factory=list)

docs/CHANGELOG.md

@@ -37,6 +37,8 @@ and this project adheres to [Semantic Versioning](https://semver.org/).
 - [修复] 发布说明生成查询 PR 作者失败时保留降级并输出包含 PR 编号和异常类型的 warning，便于排查 token、权限、网络或 GitHub API 异常。
 - [改进] DSA 数据源链路新增 Tencent 日 K 直连 fetcher、daily source health 短期熔断，并升级 AlphaSift 默认 pin/runtime bridge，默认启用 `DAILY_SOURCE=auto`、Sina snapshot 优先级和候选级 quote context。
 - [文档] 补充 AlphaSift 迁移与回退边界：明确 `ALPHASIFT_INSTALL_SPEC` 显式覆盖语义、`requirements.txt + DEFAULT_ALPHASIFT_INSTALL_SPEC` 与运行时兼容边界、以及回滚路径（关闭功能/完整 revert）说明，覆盖旧 pin 用户升级行为。
+- [新功能] #1707 新增合规 RSS/Atom 资讯源配置、拉取、去重、入库、查询、retention 与基础安全校验 API，作为个股/市场资讯情报池基线。
+- [改进] #1707 个股分析、Agent 分析和大盘复盘会 fail-open 读取本地资讯/情报池，并把来源链接作为新闻上下文和 evidence 输入。
 
 - [新功能] 个股分析历史成功保存后会从最终报告 best-effort 提取 `DecisionSignal` 决策信号，复用现有信号去重、计划质量计算和脱敏契约。
 

docs/INDEX.md

@@ -42,6 +42,7 @@
 | [Bot 命令与接入](bot-command.md) | Bot 命令、Webhook、平台接入和回调说明 |
 | [Bot 平台配置](bot/) | 飞书、钉钉、Discord 等 Bot 配置截图和补充说明 |
 | [实时告警中心](alerts.md) | EventMonitor 基线、Web 规则管理、通知结果、冷却状态和 Phase 边界 |
+| [资讯 / 情报源](intelligence-sources.md) | RSS/Atom 合规资讯源配置、测试、拉取、去重、存储、查询与安全边界 |
 | [分析上下文包契约、运行态消费与可见性](analysis-context-pack.md) | AnalysisContextPack 首版范围、字段质量状态、P1/P2 内部契约、P3 Prompt 摘要消费、P4 历史/API/Web 低敏可见性、P5 数据质量评分、P6 迁移回滚与源码锚点；完整指南补充 #1386 阶段感知分析、迁移与回滚入口 |
 | [图片识别 Prompt](image-extract-prompt.md) | 图片识别股票信息的 Prompt 与使用边界 |
 | [OpenClaw Skill 集成](openclaw-skill-integration.md) | OpenClaw / Skill 外部集成说明 |

docs/intelligence-sources.md

@@ -0,0 +1,55 @@
+# 资讯 / 情报源 MVP
+
+Issue #1707 的首版能力聚焦“合规资讯源采集、本地沉淀、可查询证据”，不把 RSS/Atom 混入按需搜索语义，也不默认新增独立舆情页。
+
+## 能力范围
+
+- 支持配置 RSS / Atom HTTP(S) 资讯源。
+- 保存资讯源配置、启用状态、作用域和最近一次拉取状态。
+- 拉取条目落库到 `intelligence_items`，保存标题、摘要、URL、来源、发布时间、拉取时间、市场与作用域。
+- 按 URL 去重；无 URL 条目使用 `no-url:intel:<hash>` 兜底键。
+- 支持 `symbol` / `market` / `sector` 作用域，以及 `cn` / `hk` / `us` / `global` 市场标记。
+- 拉取批处理采用 fail-open：单个源失败不会阻塞其他源或主分析链路。
+- 支持 retention 清理，避免资讯池无限增长。
+
+## 安全边界
+
+自定义 URL 会做基础校验：
+
+- 只允许绝对 `http` / `https` URL；
+- 禁止 URL 中携带 username/password；
+- 禁止 `localhost`、`.local`、回环地址、内网地址、链路本地地址、保留地址和组播地址；
+- 重定向后的最终 URL 也会再次校验；
+- 错误消息会脱敏常见 `token` / `key` / `secret` 查询参数。
+
+明确非目标：不做反爬、模拟登录、Cookie 抓取或非授权门户直抓。
+
+## 配置项
+
+```env
+NEWS_INTEL_RETENTION_DAYS=30
+NEWS_INTEL_FETCH_TIMEOUT_SEC=8
+NEWS_INTEL_MAX_ITEMS_PER_SOURCE=50
+```
+
+## API
+
+所有接口位于 `/api/v1/intelligence`。
+
+- `POST /sources`：创建资讯源。
+- `GET /sources`：查询资讯源。
+- `POST /sources/test`：测试 payload，不落库。
+- `POST /sources/{source_id}/fetch?dry_run=false`：拉取单个源。
+- `POST /sources/fetch-enabled`：fail-open 拉取全部启用源。
+- `GET /items?scope_type=market&market=cn&days=7`：查询资讯条目。
+
+## 后续接入建议
+
+首版基线之上，分析链路会 best-effort 读取本地资讯池：
+
+- 个股传统分析会优先读取 `symbol=<股票代码>` 的资讯，并补充同市场 `market` 级资讯；内容追加到既有 `news_context`，随 AnalysisContextPack 摘要和历史 `news_content` 保存。
+- Agent 分析同样通过 `news_context` 注入本地资讯证据，避免 Agent 必须重新搜索才能看到已沉淀新闻。
+- 大盘复盘会把同市场 `market` 级资讯合并到市场新闻列表，Prompt、结构化 payload 和报告 news 字段都能看到来源链接。
+- 本次能力仅新增本地资讯消费路径，不改模型名、provider/base URL、回退策略或运行时配置语义；兼容现有部署配置，回滚方式为清退本地资讯接入入口或移除本地资讯源配置/数据。
+
+后续 PR 可以继续完善报告 evidence 展示和 Web 设置/报告查看入口。

src/config.py

@@ -715,6 +715,9 @@ class Config:
     # === 新闻与分析筛选配置 ===
     news_max_age_days: int = 3   # 新闻最大时效（天）
     news_strategy_profile: str = "short"  # 新闻窗口策略档位：ultra_short/short/medium/long
+    news_intel_retention_days: int = 30  # 本地资讯池保留天数
+    news_intel_fetch_timeout_sec: float = 8.0  # 单个资讯源拉取超时
+    news_intel_max_items_per_source: int = 50  # 单次每个资讯源最多采集条数
     bias_threshold: float = 5.0  # 乖离率阈值（%），超过此值提示不追高
 
     # === Agent 模式配置 ===
@@ -1499,6 +1502,27 @@ def _load_from_env(cls) -> 'Config':
             news_strategy_profile=cls._parse_news_strategy_profile(
                 os.getenv('NEWS_STRATEGY_PROFILE', 'short')
             ),
+            news_intel_retention_days=parse_env_int(
+                os.getenv('NEWS_INTEL_RETENTION_DAYS'),
+                30,
+                field_name='NEWS_INTEL_RETENTION_DAYS',
+                minimum=1,
+                maximum=365,
+            ),
+            news_intel_fetch_timeout_sec=parse_env_float(
+                os.getenv('NEWS_INTEL_FETCH_TIMEOUT_SEC'),
+                8.0,
+                field_name='NEWS_INTEL_FETCH_TIMEOUT_SEC',
+                minimum=1.0,
+                maximum=30.0,
+            ),
+            news_intel_max_items_per_source=parse_env_int(
+                os.getenv('NEWS_INTEL_MAX_ITEMS_PER_SOURCE'),
+                50,
+                field_name='NEWS_INTEL_MAX_ITEMS_PER_SOURCE',
+                minimum=1,
+                maximum=200,
+            ),
             bias_threshold=parse_env_float(os.getenv('BIAS_THRESHOLD'), 5.0, field_name='BIAS_THRESHOLD', minimum=1.0),
             agent_litellm_model=agent_litellm_model,
             agent_mode=os.getenv('AGENT_MODE', 'false').lower() == 'true',