Merge branch 'master' of http://192.168.0.200/knowledge-base-bj/deerflow-code into master-nj
This commit is contained in:
commit
819e537aa3
28 files changed
+2838
-355
No files matched your search
@@ -179,6 +179,33 @@ class ManualKnowledgeCreate(BaseModel):
|
||||
content: str = Field(..., min_length=1, max_length=2_000_000)
|
||||
|
||||
|
||||
class WikiPageCreate(BaseModel):
|
||||
slug: str = Field(..., min_length=1, max_length=512)
|
||||
title: str = Field(..., min_length=1, max_length=255)
|
||||
content: str = Field(..., min_length=1, max_length=2_000_000)
|
||||
summary: str | None = Field(default=None, max_length=200_000)
|
||||
page_type: str | None = Field(default="article", min_length=1, max_length=64)
|
||||
status: str | None = Field(default="published", min_length=1, max_length=64)
|
||||
aliases: list[str] | None = Field(default=None, max_length=100)
|
||||
parent_slug: str | None = Field(default=None, max_length=512)
|
||||
category_path: list[str] | None = Field(default=None, max_length=50)
|
||||
folder_id: str | None = Field(default=None, max_length=128)
|
||||
wiki_path: str | None = Field(default=None, max_length=1024)
|
||||
|
||||
|
||||
class WikiPageUpdate(BaseModel):
|
||||
title: str | None = Field(default=None, min_length=1, max_length=255)
|
||||
content: str | None = Field(default=None, min_length=1, max_length=2_000_000)
|
||||
summary: str | None = Field(default=None, max_length=200_000)
|
||||
page_type: str | None = Field(default=None, min_length=1, max_length=64)
|
||||
status: str | None = Field(default=None, min_length=1, max_length=64)
|
||||
aliases: list[str] | None = Field(default=None, max_length=100)
|
||||
parent_slug: str | None = Field(default=None, max_length=512)
|
||||
category_path: list[str] | None = Field(default=None, max_length=50)
|
||||
folder_id: str | None = Field(default=None, max_length=128)
|
||||
wiki_path: str | None = Field(default=None, max_length=1024)
|
||||
|
||||
|
||||
class ConversationDepositCreate(BaseModel):
|
||||
thread_id: str = Field(..., min_length=1, max_length=128)
|
||||
question: str = Field(default="", max_length=200_000)
|
||||
@@ -291,15 +318,17 @@ async def _get_weknora_admin_session(runtime: Any, *, force_refresh: bool = Fals
|
||||
return cached
|
||||
|
||||
try:
|
||||
async with httpx.AsyncClient(timeout=httpx.Timeout(20.0), follow_redirects=False) as client:
|
||||
async with httpx.AsyncClient(timeout=httpx.Timeout(20.0), follow_redirects=False, trust_env=False) as client:
|
||||
response = await client.post(
|
||||
f"{runtime.api_base_url}/api/v1/auth/login",
|
||||
json={"email": runtime.admin_email, "password": runtime.admin_password},
|
||||
headers={"Accept": "application/json"},
|
||||
)
|
||||
except httpx.TimeoutException as exc:
|
||||
logger.warning("WeKnora admin login timed out: base_url=%s", runtime.api_base_url)
|
||||
raise HTTPException(status_code=504, detail="WeKnora admin login timed out") from exc
|
||||
except httpx.HTTPError as exc:
|
||||
logger.warning("WeKnora admin login unavailable: base_url=%s error=%r", runtime.api_base_url, exc)
|
||||
raise HTTPException(status_code=502, detail="WeKnora admin login is unavailable") from exc
|
||||
|
||||
if response.status_code >= 400:
|
||||
@@ -412,13 +441,14 @@ def _request_kb_id(request: Request, body: Any) -> str:
|
||||
|
||||
async def _admin_json(runtime: Any, path: str, *, token: str, params: dict[str, Any] | None = None) -> dict[str, Any]:
|
||||
try:
|
||||
async with httpx.AsyncClient(timeout=httpx.Timeout(20.0), follow_redirects=False) as client:
|
||||
async with httpx.AsyncClient(timeout=httpx.Timeout(20.0), follow_redirects=False, trust_env=False) as client:
|
||||
response = await client.get(
|
||||
f"{runtime.api_base_url}{_encode_upstream_path(path)}",
|
||||
params=params,
|
||||
headers={"Accept": "application/json", "Authorization": f"Bearer {token}"},
|
||||
)
|
||||
except httpx.HTTPError as exc:
|
||||
logger.warning("WeKnora verification request failed: base_url=%s path=%s error=%r", runtime.api_base_url, path, exc)
|
||||
raise HTTPException(status_code=502, detail="WeKnora verification request failed") from exc
|
||||
if response.status_code == 404:
|
||||
raise HTTPException(status_code=404, detail="WeKnora resource was not found")
|
||||
@@ -502,6 +532,18 @@ async def _authorize_weknora_api_proxy(
|
||||
if len(tail) == 1:
|
||||
raise HTTPException(status_code=403, detail="Global WeKnora knowledge collection is not allowed")
|
||||
action = tail[1]
|
||||
if read_method and action == "batch":
|
||||
ids = [
|
||||
item.strip()
|
||||
for raw in request.query_params.getlist("ids")
|
||||
for item in raw.split(",")
|
||||
if item.strip()
|
||||
]
|
||||
if not ids:
|
||||
raise HTTPException(status_code=403, detail="Batch operation must include document ids")
|
||||
for document_id in ids:
|
||||
await _verify_proxy_document_belongs_to_kb(runtime, ctx, document_id, token)
|
||||
return
|
||||
if action in {"batch", "batch-delete", "batch-reparse", "folder", "move", "tags"}:
|
||||
if _request_kb_id(request, json_body) != ctx.weknora_id:
|
||||
raise HTTPException(status_code=403, detail="Batch operation must be scoped to the current knowledge base")
|
||||
@@ -783,6 +825,7 @@ def _weknora_embed_injection(
|
||||
"tenantId": tenant_id,
|
||||
"user": user,
|
||||
"tenant": tenant,
|
||||
"weknoraId": ctx.weknora_id,
|
||||
"publicPrefix": public_prefix,
|
||||
"apiPrefix": _with_public_prefix(public_prefix, _WEKNORA_EMBED_API_PREFIX),
|
||||
}
|
||||
@@ -878,6 +921,29 @@ html.deerflow-weknora-embed [class^="guide__"],
|
||||
html.deerflow-weknora-embed [class*=" guide__"] {{
|
||||
display: none !important;
|
||||
}}
|
||||
html.deerflow-source-detail-open [data-deerflow-source-detail-hidden="true"] {{
|
||||
display: none !important;
|
||||
}}
|
||||
html.deerflow-source-detail-open [data-deerflow-source-detail-panel="true"] {{
|
||||
position: fixed !important;
|
||||
inset: 0 !important;
|
||||
width: 100vw !important;
|
||||
height: 100vh !important;
|
||||
max-width: none !important;
|
||||
max-height: none !important;
|
||||
margin: 0 !important;
|
||||
transform: none !important;
|
||||
z-index: 2147483000 !important;
|
||||
background: var(--td-bg-color-page, #fff) !important;
|
||||
box-shadow: none !important;
|
||||
border-radius: 0 !important;
|
||||
}}
|
||||
html.deerflow-source-detail-open [data-deerflow-source-detail-panel="true"] [class*="drawer"],
|
||||
html.deerflow-source-detail-open [data-deerflow-source-detail-panel="true"] [class*="Drawer"],
|
||||
html.deerflow-source-detail-open [data-deerflow-source-detail-panel="true"] [class*="dialog"],
|
||||
html.deerflow-source-detail-open [data-deerflow-source-detail-panel="true"] [class*="Dialog"] {{
|
||||
max-width: none !important;
|
||||
}}
|
||||
</style>
|
||||
<script id="deerflow-weknora-embed-guard">
|
||||
(() => {{
|
||||
@@ -1019,6 +1085,143 @@ html.deerflow-weknora-embed [class*=" guide__"] {{
|
||||
}});
|
||||
}}
|
||||
|
||||
const launchParams = new URLSearchParams(location.search);
|
||||
const requestedWikiSlug = launchParams.get("wikiSlug") || launchParams.get("slug") || "";
|
||||
const requestedDocumentId =
|
||||
launchParams.get("documentId") ||
|
||||
launchParams.get("knowledgeId") ||
|
||||
launchParams.get("knowledge_id") ||
|
||||
"";
|
||||
const sourceDetailMode = Boolean(requestedDocumentId || requestedWikiSlug);
|
||||
if (sourceDetailMode) {{
|
||||
root.classList.add("deerflow-source-detail-request");
|
||||
}}
|
||||
let requestedWikiOpened = false;
|
||||
let requestedDocumentOpened = false;
|
||||
let sourceDetailAttempts = 0;
|
||||
const maxSourceDetailAttempts = 18;
|
||||
function escapeAttr(value) {{
|
||||
if (window.CSS && typeof window.CSS.escape === "function") return window.CSS.escape(value);
|
||||
return String(value).replace(/["\\\\]/g, "\\\\$&");
|
||||
}}
|
||||
function openRequestedWikiSlug() {{
|
||||
if (!requestedWikiSlug || requestedWikiOpened || root.classList.contains("deerflow-source-detail-open")) return;
|
||||
const slug = escapeAttr(requestedWikiSlug);
|
||||
const link =
|
||||
document.querySelector(`[data-slug="${{slug}}"]`) ||
|
||||
document.querySelector(`a[href*="${{encodeURIComponent(requestedWikiSlug)}}"]`) ||
|
||||
document.querySelector(`a[href*="${{requestedWikiSlug.replace(/"/g, "%22")}}"]`);
|
||||
if (!link || typeof link.click !== "function") return;
|
||||
requestedWikiOpened = true;
|
||||
link.click();
|
||||
}}
|
||||
function openRequestedDocumentId() {{
|
||||
if (!requestedDocumentId || requestedDocumentOpened || root.classList.contains("deerflow-source-detail-open")) return;
|
||||
const id = escapeAttr(requestedDocumentId);
|
||||
const encoded = encodeURIComponent(requestedDocumentId);
|
||||
const candidates = [
|
||||
`[data-id="${{id}}"]`,
|
||||
`[data-row-key="${{id}}"]`,
|
||||
`[data-knowledge-id="${{id}}"]`,
|
||||
`[data-document-id="${{id}}"]`,
|
||||
`a[href*="${{encoded}}"]`,
|
||||
`button[href*="${{encoded}}"]`
|
||||
];
|
||||
let node = null;
|
||||
for (const selector of candidates) {{
|
||||
node = document.querySelector(selector);
|
||||
if (node) break;
|
||||
}}
|
||||
if (!node) {{
|
||||
node = Array.from(document.querySelectorAll("[data-id],[data-row-key],[data-knowledge-id],[data-document-id]"))
|
||||
.find((item) => {{
|
||||
const values = [
|
||||
item.getAttribute("data-id"),
|
||||
item.getAttribute("data-row-key"),
|
||||
item.getAttribute("data-knowledge-id"),
|
||||
item.getAttribute("data-document-id")
|
||||
];
|
||||
return values.includes(requestedDocumentId);
|
||||
}}) || null;
|
||||
}}
|
||||
if (!node) return;
|
||||
const clickable = node.closest("button,a,[role='button'],.t-card,.document-card,.knowledge-card,.file-card,.t-list-item,.t-table__row") || node;
|
||||
if (!clickable || typeof clickable.click !== "function") return;
|
||||
requestedDocumentOpened = true;
|
||||
clickable.click();
|
||||
}}
|
||||
function sourceDetailCandidateScore(node) {{
|
||||
if (!node || node === document.body || node === document.documentElement) return 0;
|
||||
const rect = node.getBoundingClientRect();
|
||||
if (rect.width < 360 || rect.height < 260) return 0;
|
||||
const text = (node.textContent || "").slice(0, 1200);
|
||||
const className = String(node.className || "");
|
||||
let score = 0;
|
||||
if (/基本信息|上传时间|创建时间|摘要|文档内容|查看分块|预览|引用内容|Wiki 原文|全文/.test(text)) score += 4;
|
||||
if (/drawer|Drawer|dialog|Dialog|modal|Modal|detail|Detail|preview|Preview/.test(className)) score += 3;
|
||||
if (rect.right > window.innerWidth * 0.65 && rect.width >= window.innerWidth * 0.38) score += 2;
|
||||
if (getComputedStyle(node).position === "fixed") score += 2;
|
||||
return score;
|
||||
}}
|
||||
function findSourceDetailPanel() {{
|
||||
const selectors = [
|
||||
".t-drawer",
|
||||
".t-drawer__content",
|
||||
".t-dialog",
|
||||
".t-dialog__ctx",
|
||||
".document-detail",
|
||||
".knowledge-detail",
|
||||
".document-preview",
|
||||
"[class*='Drawer']",
|
||||
"[class*='drawer']",
|
||||
"[class*='Dialog']",
|
||||
"[class*='dialog']",
|
||||
"[class*='Detail']",
|
||||
"[class*='detail']",
|
||||
"[class*='Preview']",
|
||||
"[class*='preview']"
|
||||
].join(",");
|
||||
let best = null;
|
||||
let bestScore = 0;
|
||||
document.querySelectorAll(selectors).forEach((node) => {{
|
||||
const score = sourceDetailCandidateScore(node);
|
||||
if (score > bestScore) {{
|
||||
best = node;
|
||||
bestScore = score;
|
||||
}}
|
||||
}});
|
||||
return bestScore >= 5 ? best : null;
|
||||
}}
|
||||
function isolateSourceDetailPanel() {{
|
||||
if (!sourceDetailMode) return;
|
||||
if (root.classList.contains("deerflow-source-detail-open")) return;
|
||||
const panel = findSourceDetailPanel();
|
||||
if (!panel) return;
|
||||
panel.dataset.deerflowSourceDetailPanel = "true";
|
||||
document.querySelectorAll("[data-deerflow-source-detail-hidden='true']").forEach((node) => {{
|
||||
if (node.contains(panel) || panel.contains(node)) return;
|
||||
delete node.dataset.deerflowSourceDetailHidden;
|
||||
}});
|
||||
Array.from(document.body.children).forEach((child) => {{
|
||||
if (child === panel || child.contains(panel)) return;
|
||||
child.dataset.deerflowSourceDetailHidden = "true";
|
||||
}});
|
||||
root.classList.add("deerflow-source-detail-open");
|
||||
}}
|
||||
function tickSourceDetail() {{
|
||||
if (!sourceDetailMode || root.classList.contains("deerflow-source-detail-open")) return true;
|
||||
sourceDetailAttempts += 1;
|
||||
openRequestedDocumentId();
|
||||
openRequestedWikiSlug();
|
||||
isolateSourceDetailPanel();
|
||||
if (root.classList.contains("deerflow-source-detail-open")) return true;
|
||||
if (sourceDetailAttempts >= maxSourceDetailAttempts) {{
|
||||
root.classList.add("deerflow-source-detail-open");
|
||||
return true;
|
||||
}}
|
||||
return false;
|
||||
}}
|
||||
|
||||
const allowedPath = cfg.allowedPath;
|
||||
function allowedUrl(url) {{
|
||||
try {{
|
||||
@@ -1160,14 +1363,20 @@ html.deerflow-weknora-embed [class*=" guide__"] {{
|
||||
enforceLocation();
|
||||
purgeGuides();
|
||||
hideWeKnoraShell();
|
||||
tickSourceDetail();
|
||||
if (ready) return;
|
||||
if (!force && (location.pathname !== allowedPath || !hasDetailShell())) return;
|
||||
ready = true;
|
||||
root.classList.add("deerflow-weknora-ready");
|
||||
}}
|
||||
let observerTimer = 0;
|
||||
const readyObserver = new MutationObserver(() => {{
|
||||
markReady(false);
|
||||
refreshThemeSoon();
|
||||
if (observerTimer) return;
|
||||
observerTimer = window.setTimeout(() => {{
|
||||
observerTimer = 0;
|
||||
markReady(false);
|
||||
refreshThemeSoon();
|
||||
}}, 120);
|
||||
}});
|
||||
readyObserver.observe(document.documentElement, {{ childList: true, subtree: true, attributes: true, attributeFilter: ["class", "style", "data-theme", "theme-mode"] }});
|
||||
if (document.readyState === "complete") {{
|
||||
@@ -1175,12 +1384,23 @@ html.deerflow-weknora-embed [class*=" guide__"] {{
|
||||
}} else {{
|
||||
window.addEventListener("load", () => setTimeout(() => markReady(false), 180), {{ once: true }});
|
||||
}}
|
||||
setInterval(() => {{
|
||||
const maintenanceTimer = setInterval(() => {{
|
||||
enforceLocation();
|
||||
purgeGuides();
|
||||
hideWeKnoraShell();
|
||||
if (tickSourceDetail() && sourceDetailMode) {{
|
||||
clearInterval(maintenanceTimer);
|
||||
try {{ readyObserver.disconnect(); }} catch (_) {{}}
|
||||
}}
|
||||
}}, 500);
|
||||
setTimeout(() => markReady(true), 2200);
|
||||
setTimeout(() => {{
|
||||
markReady(true);
|
||||
if (sourceDetailMode && !root.classList.contains("deerflow-source-detail-open")) {{
|
||||
root.classList.add("deerflow-source-detail-open");
|
||||
clearInterval(maintenanceTimer);
|
||||
try {{ readyObserver.disconnect(); }} catch (_) {{}}
|
||||
}}
|
||||
}}, 9000);
|
||||
}})();
|
||||
</script>
|
||||
"""
|
||||
@@ -1236,7 +1456,7 @@ async def _proxy_weknora_request(
|
||||
|
||||
url = f"{target_base_url.rstrip('/')}{_encode_upstream_path(upstream_path)}"
|
||||
try:
|
||||
async with httpx.AsyncClient(timeout=httpx.Timeout(60.0), follow_redirects=False) as client:
|
||||
async with httpx.AsyncClient(timeout=httpx.Timeout(60.0), follow_redirects=False, trust_env=False) as client:
|
||||
response = await client.request(
|
||||
request.method,
|
||||
url,
|
||||
@@ -1244,10 +1464,13 @@ async def _proxy_weknora_request(
|
||||
headers=headers,
|
||||
content=body,
|
||||
)
|
||||
if is_api and response.status_code == 401:
|
||||
if is_api and response.status_code in {401, 403}:
|
||||
session = await _get_weknora_admin_session(runtime, force_refresh=True)
|
||||
headers["Authorization"] = f"Bearer {session.get('token')}"
|
||||
async with httpx.AsyncClient(timeout=httpx.Timeout(60.0), follow_redirects=False) as client:
|
||||
tenant_id = _tenant_id_from_session(session)
|
||||
if tenant_id:
|
||||
headers["X-Tenant-ID"] = tenant_id
|
||||
async with httpx.AsyncClient(timeout=httpx.Timeout(60.0), follow_redirects=False, trust_env=False) as client:
|
||||
response = await client.request(
|
||||
request.method,
|
||||
url,
|
||||
@@ -1258,6 +1481,7 @@ async def _proxy_weknora_request(
|
||||
except httpx.TimeoutException as exc:
|
||||
raise HTTPException(status_code=504, detail="WeKnora proxy request timed out") from exc
|
||||
except httpx.HTTPError as exc:
|
||||
logger.warning("WeKnora proxy request failed: url=%s error=%r", url, exc)
|
||||
raise HTTPException(status_code=502, detail="WeKnora proxy request failed") from exc
|
||||
|
||||
content = response.content
|
||||
@@ -1577,7 +1801,18 @@ def _wiki_page_view(page: dict[str, Any]) -> dict[str, Any] | None:
|
||||
"status": str(page.get("status") or ""),
|
||||
"summary": summary,
|
||||
"content": content,
|
||||
"aliases": list(page.get("aliases") or []) if isinstance(page.get("aliases"), list) else [],
|
||||
"parent_slug": str(page.get("parent_slug") or page.get("parentSlug") or ""),
|
||||
"category_path": list(page.get("category_path") or []) if isinstance(page.get("category_path"), list) else [],
|
||||
"folder_id": str(page.get("folder_id") or page.get("folderId") or ""),
|
||||
"wiki_path": str(page.get("wiki_path") or page.get("wikiPath") or ""),
|
||||
"depth": int(page.get("depth") or 0),
|
||||
"version": int(page.get("version") or 0),
|
||||
"created_at": str(page.get("created_at") or ""),
|
||||
"updated_at": str(page.get("updated_at") or ""),
|
||||
"source_refs": list(page.get("source_refs") or []) if isinstance(page.get("source_refs"), list) else [],
|
||||
"in_links": list(page.get("in_links") or []) if isinstance(page.get("in_links"), list) else [],
|
||||
"out_links": list(page.get("out_links") or []) if isinstance(page.get("out_links"), list) else [],
|
||||
}
|
||||
|
||||
|
||||
@@ -1627,31 +1862,6 @@ async def _find_wiki_page_for_chunk(
|
||||
if view and (view.get("content") or view.get("summary")):
|
||||
return view
|
||||
|
||||
candidate_refs = _candidate_chunk_ref_tokens(focus, chunk_id)
|
||||
query = str(focus.get("knowledge_title") or focus.get("title") or focus.get("file_name") or "").strip()
|
||||
try:
|
||||
page_list = await client.list_wiki_pages(knowledge_base_id, page=1, page_size=100, query=query)
|
||||
except WeKnoraError as exc:
|
||||
if exc.status_code != 404:
|
||||
logger.debug("Failed to list WeKnora wiki pages for source preview", exc_info=True)
|
||||
return None
|
||||
|
||||
for page in page_list.get("pages", []):
|
||||
if not isinstance(page, dict):
|
||||
continue
|
||||
source_refs = _source_ref_tokens(page.get("source_refs"))
|
||||
if not source_refs or not candidate_refs.intersection(source_refs):
|
||||
continue
|
||||
slug = str(page.get("slug") or "").strip()
|
||||
try:
|
||||
full_page = await client.get_wiki_page(knowledge_base_id, slug) if slug else page
|
||||
except WeKnoraError as exc:
|
||||
if exc.status_code != 404:
|
||||
logger.debug("Failed to hydrate WeKnora wiki page slug=%s", slug, exc_info=True)
|
||||
full_page = page
|
||||
view = _wiki_page_view(full_page)
|
||||
if view and (view.get("content") or view.get("summary")):
|
||||
return view
|
||||
return None
|
||||
|
||||
|
||||
@@ -2187,17 +2397,37 @@ async def list_wiki_pages(
|
||||
request: Request,
|
||||
mapping_id: str,
|
||||
page: int = Query(default=1, ge=1),
|
||||
page_size: int = Query(default=50, ge=1, le=100),
|
||||
page_size: int = Query(default=50, ge=1, le=500),
|
||||
query: str = Query(default="", max_length=255),
|
||||
) -> dict[str, Any]:
|
||||
row, _, _ = await _authorized_mapping(request, mapping_id, write=False)
|
||||
try:
|
||||
return await _client_or_503().list_wiki_pages(
|
||||
result = await _client_or_503().list_wiki_pages(
|
||||
str(row["weknora_id"]),
|
||||
page=page,
|
||||
page_size=page_size,
|
||||
query=query.strip(),
|
||||
)
|
||||
raw_pages = result.get("pages") if isinstance(result, dict) else []
|
||||
pages = [
|
||||
view
|
||||
for page_item in (raw_pages if isinstance(raw_pages, list) else [])
|
||||
if isinstance(page_item, dict)
|
||||
for view in [_wiki_page_view(page_item)]
|
||||
if view is not None
|
||||
]
|
||||
return {**result, "pages": pages}
|
||||
except WeKnoraError as exc:
|
||||
_raise_upstream(exc)
|
||||
|
||||
|
||||
@router.post("/api/llmwiki/knowledge-bases/{mapping_id}/wiki/pages")
|
||||
async def create_wiki_page(request: Request, mapping_id: str, body: WikiPageCreate) -> dict[str, Any]:
|
||||
row, _, _ = await _authorized_mapping(request, mapping_id, write=True)
|
||||
payload = body.model_dump(exclude_none=True)
|
||||
try:
|
||||
page = await _client_or_503().create_wiki_page(str(row["weknora_id"]), payload)
|
||||
return _wiki_page_view(page) or page
|
||||
except WeKnoraError as exc:
|
||||
_raise_upstream(exc)
|
||||
|
||||
@@ -2206,7 +2436,34 @@ async def list_wiki_pages(
|
||||
async def get_wiki_page(request: Request, mapping_id: str, slug: str) -> dict[str, Any]:
|
||||
row, _, _ = await _authorized_mapping(request, mapping_id, write=False)
|
||||
try:
|
||||
return await _client_or_503().get_wiki_page(str(row["weknora_id"]), slug)
|
||||
page = await _client_or_503().get_wiki_page(str(row["weknora_id"]), slug)
|
||||
return _wiki_page_view(page) or page
|
||||
except WeKnoraError as exc:
|
||||
_raise_upstream(exc)
|
||||
|
||||
|
||||
@router.patch("/api/llmwiki/knowledge-bases/{mapping_id}/wiki/pages/{slug:path}")
|
||||
async def update_wiki_page(request: Request, mapping_id: str, slug: str, body: WikiPageUpdate) -> dict[str, Any]:
|
||||
row, _, _ = await _authorized_mapping(request, mapping_id, write=True)
|
||||
changes = body.model_dump(exclude_none=True)
|
||||
if not changes:
|
||||
try:
|
||||
page = await _client_or_503().get_wiki_page(str(row["weknora_id"]), slug)
|
||||
return _wiki_page_view(page) or page
|
||||
except WeKnoraError as exc:
|
||||
_raise_upstream(exc)
|
||||
try:
|
||||
page = await _client_or_503().update_wiki_page(str(row["weknora_id"]), slug, changes)
|
||||
return _wiki_page_view(page) or page
|
||||
except WeKnoraError as exc:
|
||||
_raise_upstream(exc)
|
||||
|
||||
|
||||
@router.delete("/api/llmwiki/knowledge-bases/{mapping_id}/wiki/pages/{slug:path}", status_code=204)
|
||||
async def delete_wiki_page(request: Request, mapping_id: str, slug: str) -> None:
|
||||
row, _, _ = await _authorized_mapping(request, mapping_id, write=True)
|
||||
try:
|
||||
await _client_or_503().delete_wiki_page(str(row["weknora_id"]), slug)
|
||||
except WeKnoraError as exc:
|
||||
_raise_upstream(exc)
|
||||
|
||||
@@ -2407,6 +2664,20 @@ async def get_source_context(
|
||||
sanitize_deposit = _is_conversation_deposit_mapping(row)
|
||||
if sanitize_deposit and wiki_page is not None:
|
||||
wiki_page = _sanitize_conversation_deposit_value(wiki_page)
|
||||
focus_chunk = next((chunk for chunk in chunks if str(chunk.get("id") or "") == chunk_id), {})
|
||||
document_id = str(
|
||||
focus_chunk.get("knowledge_id")
|
||||
or focus_chunk.get("document_id")
|
||||
or focus_chunk.get("knowledgeId")
|
||||
or ""
|
||||
)
|
||||
document_title = str(
|
||||
focus_chunk.get("knowledge_title")
|
||||
or focus_chunk.get("title")
|
||||
or focus_chunk.get("file_name")
|
||||
or focus_chunk.get("filename")
|
||||
or ""
|
||||
)
|
||||
normalized_chunks = [
|
||||
{
|
||||
"id": str(chunk.get("id") or ""),
|
||||
@@ -2426,6 +2697,8 @@ async def get_source_context(
|
||||
"data": {
|
||||
"knowledge_base_id": row["id"],
|
||||
"knowledge_base_name": row.get("name") or "",
|
||||
"document_id": _sanitize_conversation_deposit_value(document_id) if sanitize_deposit else document_id,
|
||||
"document_title": _sanitize_conversation_deposit_value(document_title) if sanitize_deposit else document_title,
|
||||
"focus_chunk_id": chunk_id,
|
||||
"display_mode": "wiki" if wiki_page is not None else "chunks",
|
||||
"wiki_page": wiki_page,
|
||||
|
||||
+5
-2
@@ -12,7 +12,7 @@ from typing import Any, override
|
||||
from langchain.agents import AgentState
|
||||
from langchain.agents.middleware import AgentMiddleware
|
||||
from langchain.agents.middleware.types import ModelCallResult, ModelRequest, ModelResponse
|
||||
from langchain_core.messages import SystemMessage
|
||||
from langchain_core.messages import HumanMessage
|
||||
|
||||
from deerflow.runtime.references import (
|
||||
build_latest_reference_prompt,
|
||||
@@ -105,7 +105,10 @@ class RagCitationMiddleware(AgentMiddleware[AgentState]):
|
||||
)
|
||||
patched_messages = [
|
||||
*messages,
|
||||
SystemMessage(content=prompt),
|
||||
HumanMessage(
|
||||
content=prompt,
|
||||
additional_kwargs={"hide_from_ui": True, "lc_source": "rag_citation"},
|
||||
),
|
||||
]
|
||||
return request.override(messages=patched_messages)
|
||||
|
||||
|
||||
+6
-5
@@ -14,13 +14,13 @@ from typing import override
|
||||
from langchain.agents import AgentState
|
||||
from langchain.agents.middleware import AgentMiddleware
|
||||
from langchain.agents.middleware.types import ModelCallResult, ModelRequest, ModelResponse
|
||||
from langchain_core.messages import HumanMessage, SystemMessage
|
||||
from langchain_core.messages import HumanMessage
|
||||
|
||||
|
||||
class WritingSetupRoundCapMiddleware(AgentMiddleware[AgentState]):
|
||||
"""达到澄清轮数上限时,强制模型调用 ``setup_writing`` 提交。
|
||||
|
||||
软约束(注入 SystemMessage)+ 硬约束(``tool_choice`` 锁定工具)双保险:即使个别模型
|
||||
软约束(注入隐藏 HumanMessage)+ 硬约束(``tool_choice`` 锁定工具)双保险:即使个别模型
|
||||
无视提示,``tool_choice`` 也会逼它产出一次 ``setup_writing`` 调用。
|
||||
"""
|
||||
|
||||
@@ -33,8 +33,8 @@ class WritingSetupRoundCapMiddleware(AgentMiddleware[AgentState]):
|
||||
human_count = sum(1 for m in messages if isinstance(m, HumanMessage))
|
||||
return human_count >= self._max_rounds
|
||||
|
||||
def _reminder(self) -> SystemMessage:
|
||||
return SystemMessage(
|
||||
def _reminder(self) -> HumanMessage:
|
||||
return HumanMessage(
|
||||
content=(
|
||||
f"【系统强制】意图澄清已达 {self._max_rounds} 轮上限。"
|
||||
"现在必须立即调用 setup_writing 提交写作配置,禁止再向用户提任何问题。"
|
||||
@@ -42,7 +42,8 @@ class WritingSetupRoundCapMiddleware(AgentMiddleware[AgentState]):
|
||||
"word_count_target 默认 800、target_audience 默认「普通读者」、"
|
||||
"writing_mode 默认 loose、material_source 默认 knowledge_base、"
|
||||
"article_type 不确定就留空。"
|
||||
)
|
||||
),
|
||||
additional_kwargs={"hide_from_ui": True, "lc_source": "writing_setup_round_cap"},
|
||||
)
|
||||
|
||||
def _patch(self, request: ModelRequest) -> ModelRequest:
|
||||
|
||||
+37
-1
@@ -110,6 +110,7 @@ class WeKnoraClient:
|
||||
client = self._http_client or httpx.AsyncClient(
|
||||
timeout=httpx.Timeout(self.timeout_seconds),
|
||||
follow_redirects=False,
|
||||
trust_env=False,
|
||||
)
|
||||
try:
|
||||
response = await client.post(
|
||||
@@ -164,7 +165,11 @@ class WeKnoraClient:
|
||||
data: Mapping[str, Any] | None = None,
|
||||
) -> Any:
|
||||
owns_client = self._http_client is None
|
||||
client = self._http_client or httpx.AsyncClient(timeout=httpx.Timeout(self.timeout_seconds), follow_redirects=False)
|
||||
client = self._http_client or httpx.AsyncClient(
|
||||
timeout=httpx.Timeout(self.timeout_seconds),
|
||||
follow_redirects=False,
|
||||
trust_env=False,
|
||||
)
|
||||
try:
|
||||
for attempt in range(2):
|
||||
response = await client.request(
|
||||
@@ -392,6 +397,7 @@ class WeKnoraClient:
|
||||
client = self._http_client or httpx.AsyncClient(
|
||||
timeout=httpx.Timeout(self.timeout_seconds),
|
||||
follow_redirects=True,
|
||||
trust_env=False,
|
||||
)
|
||||
try:
|
||||
for attempt in range(2):
|
||||
@@ -486,6 +492,36 @@ class WeKnoraClient:
|
||||
raise WeKnoraError("WeKnora returned an invalid Wiki page")
|
||||
return dict(value)
|
||||
|
||||
async def create_wiki_page(self, knowledge_base_id: str, payload: Mapping[str, Any]) -> dict[str, Any]:
|
||||
value = _data(
|
||||
await self._request(
|
||||
"POST",
|
||||
f"/api/v1/knowledgebase/{quote(knowledge_base_id, safe='')}/wiki/pages",
|
||||
json=dict(payload),
|
||||
)
|
||||
)
|
||||
if not isinstance(value, dict):
|
||||
raise WeKnoraError("WeKnora returned an invalid Wiki page")
|
||||
return dict(value)
|
||||
|
||||
async def update_wiki_page(self, knowledge_base_id: str, slug: str, changes: Mapping[str, Any]) -> dict[str, Any]:
|
||||
safe_slug = "/".join(quote(part, safe="") for part in slug.split("/"))
|
||||
value = _data(
|
||||
await self._request(
|
||||
"PUT",
|
||||
f"/api/v1/knowledgebase/{quote(knowledge_base_id, safe='')}/wiki/pages/{safe_slug}",
|
||||
json=dict(changes),
|
||||
)
|
||||
)
|
||||
return dict(value) if isinstance(value, dict) else {}
|
||||
|
||||
async def delete_wiki_page(self, knowledge_base_id: str, slug: str) -> None:
|
||||
safe_slug = "/".join(quote(part, safe="") for part in slug.split("/"))
|
||||
await self._request(
|
||||
"DELETE",
|
||||
f"/api/v1/knowledgebase/{quote(knowledge_base_id, safe='')}/wiki/pages/{safe_slug}",
|
||||
)
|
||||
|
||||
async def search(self, query: str, knowledge_base_ids: Sequence[str]) -> list[dict[str, Any]]:
|
||||
ids = [str(value) for value in knowledge_base_ids if str(value)]
|
||||
if not ids:
|
||||
|
||||
@@ -0,0 +1,82 @@
|
||||
"""Utilities for provider-specific chat message ordering constraints."""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
from typing import Any
|
||||
|
||||
from langchain_core.messages import BaseMessage, SystemMessage
|
||||
|
||||
|
||||
_SYSTEM_SEPARATOR = "\n\n---\n\n"
|
||||
|
||||
|
||||
def normalize_chat_payload_messages(payload: dict[str, Any]) -> None:
|
||||
"""Ensure payload ``system`` messages are represented once at the beginning.
|
||||
|
||||
Several OpenAI-compatible gateways reject requests when a transient
|
||||
``system`` message appears after user/assistant/tool history. LangChain and
|
||||
DeerFlow middleware can legitimately create those transient messages, so the
|
||||
provider boundary folds every system payload into a single leading message
|
||||
while preserving all non-system messages in their original order.
|
||||
"""
|
||||
messages = payload.get("messages")
|
||||
if not isinstance(messages, list):
|
||||
return
|
||||
|
||||
normalized = _normalize_payload_message_list(messages)
|
||||
if normalized is not messages:
|
||||
payload["messages"] = normalized
|
||||
|
||||
|
||||
def normalize_langchain_messages(messages: list[BaseMessage]) -> list[BaseMessage]:
|
||||
"""Return LangChain messages with all system context merged at the front."""
|
||||
system_messages = [message for message in messages if isinstance(message, SystemMessage)]
|
||||
if not system_messages:
|
||||
return messages
|
||||
|
||||
first_non_system_index = next((idx for idx, message in enumerate(messages) if not isinstance(message, SystemMessage)), len(messages))
|
||||
if len(system_messages) == first_non_system_index:
|
||||
return messages
|
||||
|
||||
merged = system_messages[0].model_copy(
|
||||
update={
|
||||
"content": _merge_contents([message.content for message in system_messages]),
|
||||
}
|
||||
)
|
||||
return [merged, *(message for message in messages if not isinstance(message, SystemMessage))]
|
||||
|
||||
|
||||
def _normalize_payload_message_list(messages: list[Any]) -> list[Any]:
|
||||
system_messages = [message for message in messages if isinstance(message, dict) and message.get("role") == "system"]
|
||||
if not system_messages:
|
||||
return messages
|
||||
|
||||
first_non_system_index = next(
|
||||
(
|
||||
idx
|
||||
for idx, message in enumerate(messages)
|
||||
if not (isinstance(message, dict) and message.get("role") == "system")
|
||||
),
|
||||
len(messages),
|
||||
)
|
||||
if len(system_messages) == first_non_system_index:
|
||||
return messages
|
||||
|
||||
merged_system = dict(system_messages[0])
|
||||
merged_system["content"] = _merge_contents([message.get("content") for message in system_messages])
|
||||
return [merged_system, *(message for message in messages if not (isinstance(message, dict) and message.get("role") == "system"))]
|
||||
|
||||
|
||||
def _merge_contents(contents: list[Any]) -> Any:
|
||||
if all(isinstance(content, str) for content in contents):
|
||||
return _SYSTEM_SEPARATOR.join(content for content in contents if content)
|
||||
|
||||
blocks: list[Any] = []
|
||||
for idx, content in enumerate(contents):
|
||||
if idx > 0:
|
||||
blocks.append({"type": "text", "text": _SYSTEM_SEPARATOR})
|
||||
if isinstance(content, list):
|
||||
blocks.extend(content)
|
||||
elif content is not None:
|
||||
blocks.append({"type": "text", "text": str(content)})
|
||||
return blocks
|
||||
+7
-5
@@ -6,10 +6,12 @@ import uuid
|
||||
from collections.abc import Iterator
|
||||
|
||||
import httpx
|
||||
from langchain_core.messages import AIMessage, AIMessageChunk, HumanMessage, ToolMessage
|
||||
from langchain_core.outputs import ChatGenerationChunk, ChatResult
|
||||
from langchain_openai import ChatOpenAI
|
||||
|
||||
from langchain_core.messages import AIMessage, AIMessageChunk, HumanMessage, ToolMessage
|
||||
from langchain_core.outputs import ChatGenerationChunk, ChatResult
|
||||
from langchain_openai import ChatOpenAI
|
||||
|
||||
from deerflow.models.message_ordering import normalize_langchain_messages
|
||||
|
||||
|
||||
def _fix_messages(messages: list) -> list:
|
||||
"""Sanitize incoming messages for MindIE compatibility.
|
||||
@@ -55,7 +57,7 @@ def _fix_messages(messages: list) -> list:
|
||||
|
||||
fixed.append(msg.model_copy(update={"content": text}))
|
||||
|
||||
return fixed
|
||||
return normalize_langchain_messages(fixed)
|
||||
|
||||
|
||||
def _parse_xml_tool_call_to_dict(content: str) -> tuple[str, list[dict]]:
|
||||
|
||||
+13
-10
@@ -9,10 +9,12 @@ on all assistant messages when thinking mode is enabled.
|
||||
|
||||
from typing import Any
|
||||
|
||||
from langchain_core.language_models import LanguageModelInput
|
||||
from langchain_core.messages import AIMessage
|
||||
from langchain_deepseek import ChatDeepSeek
|
||||
|
||||
from langchain_core.language_models import LanguageModelInput
|
||||
from langchain_core.messages import AIMessage
|
||||
from langchain_deepseek import ChatDeepSeek
|
||||
|
||||
from deerflow.models.message_ordering import normalize_chat_payload_messages
|
||||
|
||||
|
||||
class PatchedChatDeepSeek(ChatDeepSeek):
|
||||
"""ChatDeepSeek with proper reasoning_content preservation.
|
||||
@@ -65,9 +67,10 @@ class PatchedChatDeepSeek(ChatDeepSeek):
|
||||
ai_messages = [m for m in original_messages if isinstance(m, AIMessage)]
|
||||
assistant_payloads = [(i, m) for i, m in enumerate(payload_messages) if m.get("role") == "assistant"]
|
||||
|
||||
for (idx, payload_msg), ai_msg in zip(assistant_payloads, ai_messages):
|
||||
reasoning_content = ai_msg.additional_kwargs.get("reasoning_content")
|
||||
if reasoning_content is not None:
|
||||
payload_messages[idx]["reasoning_content"] = reasoning_content
|
||||
|
||||
return payload
|
||||
for (idx, payload_msg), ai_msg in zip(assistant_payloads, ai_messages):
|
||||
reasoning_content = ai_msg.additional_kwargs.get("reasoning_content")
|
||||
if reasoning_content is not None:
|
||||
payload_messages[idx]["reasoning_content"] = reasoning_content
|
||||
|
||||
normalize_chat_payload_messages(payload)
|
||||
return payload
|
||||
+11
-8
@@ -20,11 +20,13 @@ from langchain_core.language_models import LanguageModelInput
|
||||
from langchain_core.messages import AIMessage, AIMessageChunk
|
||||
from langchain_core.outputs import ChatGeneration, ChatGenerationChunk, ChatResult
|
||||
from langchain_openai import ChatOpenAI
|
||||
from langchain_openai.chat_models.base import (
|
||||
_convert_delta_to_message_chunk,
|
||||
_create_usage_metadata,
|
||||
)
|
||||
|
||||
from langchain_openai.chat_models.base import (
|
||||
_convert_delta_to_message_chunk,
|
||||
_create_usage_metadata,
|
||||
)
|
||||
|
||||
from deerflow.models.message_ordering import normalize_chat_payload_messages
|
||||
|
||||
_THINK_TAG_RE = re.compile(r"<think>\s*(.*?)\s*</think>", re.DOTALL)
|
||||
|
||||
|
||||
@@ -112,9 +114,10 @@ class PatchedChatMiniMax(ChatOpenAI):
|
||||
**extra_body,
|
||||
"reasoning_split": True,
|
||||
}
|
||||
else:
|
||||
payload["extra_body"] = {"reasoning_split": True}
|
||||
return payload
|
||||
else:
|
||||
payload["extra_body"] = {"reasoning_split": True}
|
||||
normalize_chat_payload_messages(payload)
|
||||
return payload
|
||||
|
||||
def _convert_chunk_to_generation_chunk(
|
||||
self,
|
||||
|
||||
+15
-12
@@ -29,9 +29,11 @@ from typing import Any
|
||||
|
||||
from langchain_core.language_models import LanguageModelInput
|
||||
from langchain_core.messages import AIMessage, AIMessageChunk
|
||||
from langchain_core.outputs import ChatGeneration, ChatGenerationChunk, ChatResult
|
||||
from langchain_openai import ChatOpenAI
|
||||
|
||||
from langchain_core.outputs import ChatGeneration, ChatGenerationChunk, ChatResult
|
||||
from langchain_openai import ChatOpenAI
|
||||
|
||||
from deerflow.models.message_ordering import normalize_chat_payload_messages
|
||||
|
||||
|
||||
class PatchedChatOpenAI(ChatOpenAI):
|
||||
"""ChatOpenAI with ``thought_signature`` preservation for Gemini thinking via OpenAI gateway.
|
||||
@@ -87,15 +89,16 @@ class PatchedChatOpenAI(ChatOpenAI):
|
||||
if payload_msg.get("role") == "assistant" and isinstance(orig_msg, AIMessage):
|
||||
_restore_tool_call_signatures(payload_msg, orig_msg)
|
||||
_restore_reasoning_content(payload_msg, orig_msg)
|
||||
else:
|
||||
# Fallback: match assistant-role entries positionally against AIMessages.
|
||||
ai_messages = [m for m in original_messages if isinstance(m, AIMessage)]
|
||||
assistant_payloads = [(i, m) for i, m in enumerate(payload_messages) if m.get("role") == "assistant"]
|
||||
for (_, payload_msg), ai_msg in zip(assistant_payloads, ai_messages):
|
||||
_restore_tool_call_signatures(payload_msg, ai_msg)
|
||||
_restore_reasoning_content(payload_msg, ai_msg)
|
||||
|
||||
return payload
|
||||
else:
|
||||
# Fallback: match assistant-role entries positionally against AIMessages.
|
||||
ai_messages = [m for m in original_messages if isinstance(m, AIMessage)]
|
||||
assistant_payloads = [(i, m) for i, m in enumerate(payload_messages) if m.get("role") == "assistant"]
|
||||
for (_, payload_msg), ai_msg in zip(assistant_payloads, ai_messages):
|
||||
_restore_tool_call_signatures(payload_msg, ai_msg)
|
||||
_restore_reasoning_content(payload_msg, ai_msg)
|
||||
|
||||
normalize_chat_payload_messages(payload)
|
||||
return payload
|
||||
|
||||
def _convert_chunk_to_generation_chunk(
|
||||
self,
|
||||
|
||||
+13
-10
@@ -32,9 +32,11 @@ from langchain_core.messages import (
|
||||
)
|
||||
from langchain_core.messages.tool import tool_call_chunk
|
||||
from langchain_core.outputs import ChatGeneration, ChatGenerationChunk, ChatResult
|
||||
from langchain_openai import ChatOpenAI
|
||||
from langchain_openai.chat_models.base import _create_usage_metadata
|
||||
|
||||
from langchain_openai import ChatOpenAI
|
||||
from langchain_openai.chat_models.base import _create_usage_metadata
|
||||
|
||||
from deerflow.models.message_ordering import normalize_chat_payload_messages
|
||||
|
||||
|
||||
def _normalize_vllm_chat_template_kwargs(payload: dict[str, Any]) -> None:
|
||||
"""Map DeerFlow's legacy ``thinking`` toggle to vLLM/Qwen's ``enable_thinking``.
|
||||
@@ -182,13 +184,14 @@ class VllmChatModel(ChatOpenAI):
|
||||
for payload_msg, orig_msg in zip(payload_messages, original_messages):
|
||||
if payload_msg.get("role") == "assistant" and isinstance(orig_msg, AIMessage):
|
||||
_restore_reasoning_field(payload_msg, orig_msg)
|
||||
else:
|
||||
ai_messages = [message for message in original_messages if isinstance(message, AIMessage)]
|
||||
assistant_payloads = [message for message in payload_messages if message.get("role") == "assistant"]
|
||||
for payload_msg, ai_msg in zip(assistant_payloads, ai_messages):
|
||||
_restore_reasoning_field(payload_msg, ai_msg)
|
||||
|
||||
return payload
|
||||
else:
|
||||
ai_messages = [message for message in original_messages if isinstance(message, AIMessage)]
|
||||
assistant_payloads = [message for message in payload_messages if message.get("role") == "assistant"]
|
||||
for payload_msg, ai_msg in zip(assistant_payloads, ai_messages):
|
||||
_restore_reasoning_field(payload_msg, ai_msg)
|
||||
|
||||
normalize_chat_payload_messages(payload)
|
||||
return payload
|
||||
|
||||
def _create_chat_result(self, response: dict | openai.BaseModel, generation_info: dict | None = None) -> ChatResult:
|
||||
"""Preserve vLLM reasoning on non-streaming responses."""
|
||||
|
||||
+3
@@ -133,6 +133,9 @@ async def llmwiki_search_tool(
|
||||
{
|
||||
"knowledge_base_id": mapping["id"],
|
||||
"chunk_id": visible_result.get("chunk_id") or "",
|
||||
"knowledge_id": visible_result.get("knowledge_id") or "",
|
||||
"document_id": visible_result.get("knowledge_id") or "",
|
||||
"wiki_slug": visible_result.get("wiki_slug") or visible_result.get("slug") or "",
|
||||
}
|
||||
)
|
||||
citations.append(
|
||||
|
||||
@@ -0,0 +1,84 @@
|
||||
import pytest
|
||||
from langchain_core.messages import AIMessage, HumanMessage, SystemMessage
|
||||
|
||||
from deerflow.agents.middlewares.writing_setup_round_cap_middleware import WritingSetupRoundCapMiddleware
|
||||
from deerflow.models.message_ordering import normalize_chat_payload_messages, normalize_langchain_messages
|
||||
from deerflow.models.patched_openai import PatchedChatOpenAI
|
||||
|
||||
|
||||
def _multi_turn_payload_with_late_systems(turns: int):
|
||||
messages = [{"role": "system", "content": "base system"}]
|
||||
for idx in range(turns):
|
||||
messages.append({"role": "user", "content": f"user {idx}"})
|
||||
messages.append({"role": "assistant", "content": f"assistant {idx}"})
|
||||
if idx >= 9 and (idx + 1) % 10 == 0:
|
||||
messages.append({"role": "system", "content": f"late summary after {idx + 1} turns"})
|
||||
messages.append({"role": "system", "content": "tail forced instruction"})
|
||||
return {"messages": messages}
|
||||
|
||||
|
||||
@pytest.mark.parametrize("turns", [12, 25, 50])
|
||||
def test_normalize_chat_payload_messages_merges_late_systems_after_many_turns(turns):
|
||||
payload = _multi_turn_payload_with_late_systems(turns)
|
||||
|
||||
normalize_chat_payload_messages(payload)
|
||||
|
||||
roles = [message["role"] for message in payload["messages"]]
|
||||
assert roles[0] == "system"
|
||||
assert roles.count("system") == 1
|
||||
assert roles[1:5] == ["user", "assistant", "user", "assistant"]
|
||||
assert "base system" in payload["messages"][0]["content"]
|
||||
assert "late summary after 10 turns" in payload["messages"][0]["content"]
|
||||
assert "tail forced instruction" in payload["messages"][0]["content"]
|
||||
|
||||
|
||||
def test_normalize_langchain_messages_moves_late_systems_before_history():
|
||||
messages = [SystemMessage(content="base system")]
|
||||
for idx in range(12):
|
||||
messages.append(HumanMessage(content=f"user {idx}"))
|
||||
messages.append(AIMessage(content=f"assistant {idx}"))
|
||||
if idx == 10:
|
||||
messages.append(SystemMessage(content="late compaction summary"))
|
||||
|
||||
normalized = normalize_langchain_messages(messages)
|
||||
|
||||
assert isinstance(normalized[0], SystemMessage)
|
||||
assert sum(isinstance(message, SystemMessage) for message in normalized) == 1
|
||||
assert normalized[0].content == "base system\n\n---\n\nlate compaction summary"
|
||||
assert isinstance(normalized[1], HumanMessage)
|
||||
assert normalized[1].content == "user 0"
|
||||
|
||||
|
||||
@pytest.mark.parametrize("turns", [12, 25, 50])
|
||||
def test_patched_openai_payload_has_only_leading_system_after_many_turns(turns):
|
||||
model = PatchedChatOpenAI(
|
||||
model="test-model",
|
||||
api_key="test-key",
|
||||
base_url="http://127.0.0.1:9/v1",
|
||||
)
|
||||
messages = [SystemMessage(content="base system")]
|
||||
for idx in range(turns):
|
||||
messages.append(HumanMessage(content=f"user {idx}"))
|
||||
messages.append(AIMessage(content=f"assistant {idx}"))
|
||||
if idx >= 9 and (idx + 1) % 10 == 0:
|
||||
messages.append(SystemMessage(content=f"late system after {idx + 1} turns"))
|
||||
messages.append(SystemMessage(content="tail system"))
|
||||
|
||||
payload = model._get_request_payload(messages)
|
||||
|
||||
roles = [message["role"] for message in payload["messages"]]
|
||||
assert roles[0] == "system"
|
||||
assert roles.count("system") == 1
|
||||
assert "base system" in payload["messages"][0]["content"]
|
||||
assert "late system after 10 turns" in payload["messages"][0]["content"]
|
||||
assert "tail system" in payload["messages"][0]["content"]
|
||||
|
||||
|
||||
def test_writing_round_cap_reminder_is_not_late_system_message():
|
||||
middleware = WritingSetupRoundCapMiddleware(max_rounds=3)
|
||||
|
||||
reminder = middleware._reminder()
|
||||
|
||||
assert isinstance(reminder, HumanMessage)
|
||||
assert not isinstance(reminder, SystemMessage)
|
||||
assert reminder.additional_kwargs["hide_from_ui"] is True
|
||||
@@ -551,6 +551,39 @@ async def test_weknora_client_imports_url_manual_content_and_lists_wiki_pages()
|
||||
assert pages["pages"][0]["slug"] == "guide"
|
||||
|
||||
|
||||
@pytest.mark.asyncio
|
||||
async def test_weknora_client_creates_updates_and_deletes_wiki_pages() -> None:
|
||||
seen: list[tuple[str, str]] = []
|
||||
|
||||
async def handler(request: httpx.Request) -> httpx.Response:
|
||||
seen.append((request.method, request.url.path))
|
||||
if request.method == "POST":
|
||||
assert request.url.path == "/api/v1/knowledgebase/kb-one/wiki/pages"
|
||||
assert json.loads(request.content) == {"slug": "folder/guide", "title": "Guide", "content": "# Guide"}
|
||||
return httpx.Response(200, json={"success": True, "data": {"slug": "folder/guide", "title": "Guide"}})
|
||||
assert request.url.path == "/api/v1/knowledgebase/kb-one/wiki/pages/folder/guide"
|
||||
if request.method == "PUT":
|
||||
assert json.loads(request.content) == {"title": "Guide", "content": "# Guide"}
|
||||
return httpx.Response(200, json={"success": True, "data": {"slug": "folder/guide", "title": "Guide"}})
|
||||
if request.method == "DELETE":
|
||||
return httpx.Response(204)
|
||||
return httpx.Response(405)
|
||||
|
||||
async with httpx.AsyncClient(transport=httpx.MockTransport(handler)) as http_client:
|
||||
client = WeKnoraClient(base_url="http://weknora.local:8080", http_client=http_client)
|
||||
created = await client.create_wiki_page("kb-one", {"slug": "folder/guide", "title": "Guide", "content": "# Guide"})
|
||||
updated = await client.update_wiki_page("kb-one", "folder/guide", {"title": "Guide", "content": "# Guide"})
|
||||
await client.delete_wiki_page("kb-one", "folder/guide")
|
||||
|
||||
assert created["slug"] == "folder/guide"
|
||||
assert updated["slug"] == "folder/guide"
|
||||
assert seen == [
|
||||
("POST", "/api/v1/knowledgebase/kb-one/wiki/pages"),
|
||||
("PUT", "/api/v1/knowledgebase/kb-one/wiki/pages/folder/guide"),
|
||||
("DELETE", "/api/v1/knowledgebase/kb-one/wiki/pages/folder/guide"),
|
||||
]
|
||||
|
||||
|
||||
@pytest.mark.asyncio
|
||||
async def test_weknora_client_reads_document_chunks_and_wiki_graph() -> None:
|
||||
async def handler(request: httpx.Request) -> httpx.Response:
|
||||
|
||||
Reference in new issue
Block a user