feat: add governed public knowledge retrieval

This commit is contained in:
张胜宇
2026-09-10 18:42:44 +08:00
parent ba22a2220f
commit 511a8ca18f
20 changed files with 729 additions and 7 deletions
+2
View File
@@ -54,6 +54,8 @@ class Settings(BaseSettings):
model_router_config_ref: str = "local"
model_default_endpoint: str = ""
model_fallback_endpoint: str = ""
knowledge_embedding_endpoint_code: str = ""
knowledge_embedding_timeout_ms: int = Field(default=15000, gt=0)
sse_heartbeat_seconds: int = Field(default=15, gt=0)
sse_chunk_characters: int = Field(default=256, ge=1, le=4096)
sse_max_connection_seconds: int = Field(default=300, gt=0)
+43
View File
@@ -0,0 +1,43 @@
from pydantic import BaseModel, ConfigDict, Field, field_validator
ALLOWED_KNOWLEDGE_COLLECTIONS = frozenset({
"fin_faq_collection",
"fin_product_collection",
"fin_policy_collection",
})
class KnowledgeQuery(BaseModel):
model_config = ConfigDict(extra="forbid", frozen=True)
query: str = Field(min_length=1, max_length=2000)
intents: tuple[str, ...] = Field(min_length=1, max_length=4)
top_k: int = Field(default=5, ge=1, le=20)
@field_validator("query")
@classmethod
def query_must_not_be_blank(cls, value: str) -> str:
if not value.strip():
raise ValueError("query must not be blank")
return value
class KnowledgeHit(BaseModel):
model_config = ConfigDict(extra="forbid", frozen=True)
knowledge_id: str
collection: str
snippet: str
title: str | None = None
answer: str | None = None
score: float | None = Field(default=None, ge=0, le=1)
version: str | None = None
class KnowledgeSearchResult(BaseModel):
model_config = ConfigDict(extra="forbid", frozen=True)
hits: tuple[KnowledgeHit, ...] = ()
degraded: bool = False
degradation_reason: str | None = None
searched_collections: tuple[str, ...] = ()
+2 -1
View File
@@ -92,6 +92,7 @@ class JwtAuthenticator:
if claims.get("visitor") is True:
return RequestContext(
user_id=str(subject), trace_id=str(uuid4()), roles=("visitor",),
permissions=("agent:run",), data_scope="public",
# 访客仅可运行 Agent 与读取已发布的公共知识,绝不含个人数据权限。
permissions=("agent:run", "knowledge:query"), data_scope="public",
)
return RequestContext(user_id=str(claims["sub"]), trace_id=str(uuid4()))