feat: added improved llm model selector

This commit is contained in:
DESKTOP-RTLN3BA\$punk 2026-02-20 14:28:01 -08:00
parent dc19b43967
commit a3cd598e01
23 changed files with 14733 additions and 126 deletions

View file

@ -230,10 +230,12 @@ def create_create_linear_issue_tool(
raise
logger.error(f"Error creating Linear issue: {e}", exc_info=True)
if isinstance(e, (ValueError, LinearAPIError)):
if isinstance(e, ValueError | LinearAPIError):
message = str(e)
else:
message = "Something went wrong while creating the issue. Please try again."
message = (
"Something went wrong while creating the issue. Please try again."
)
return {"status": "error", "message": message}
return create_linear_issue

View file

@ -238,7 +238,9 @@ def create_delete_linear_issue_tool(
if result.get("status") == "success":
result["deleted_from_kb"] = deleted_from_kb
if issue_identifier:
result["message"] = f"Issue {issue_identifier} archived successfully."
result["message"] = (
f"Issue {issue_identifier} archived successfully."
)
if deleted_from_kb:
result["message"] = (
f"{result.get('message', '')} Also removed from the knowledge base."
@ -253,10 +255,12 @@ def create_delete_linear_issue_tool(
raise
logger.error(f"Error deleting Linear issue: {e}", exc_info=True)
if isinstance(e, (ValueError, LinearAPIError)):
if isinstance(e, ValueError | LinearAPIError):
message = str(e)
else:
message = "Something went wrong while deleting the issue. Please try again."
message = (
"Something went wrong while deleting the issue. Please try again."
)
return {"status": "error", "message": message}
return delete_linear_issue

View file

@ -290,10 +290,12 @@ def create_update_linear_issue_tool(
raise
logger.error(f"Error updating Linear issue: {e}", exc_info=True)
if isinstance(e, (ValueError, LinearAPIError)):
if isinstance(e, ValueError | LinearAPIError):
message = str(e)
else:
message = "Something went wrong while updating the issue. Please try again."
message = (
"Something went wrong while updating the issue. Please try again."
)
return {"status": "error", "message": message}
return update_linear_issue

View file

@ -224,10 +224,12 @@ def create_create_notion_page_tool(
raise
logger.error(f"Error creating Notion page: {e}", exc_info=True)
if isinstance(e, (ValueError, NotionAPIError)):
if isinstance(e, ValueError | NotionAPIError):
message = str(e)
else:
message = "Something went wrong while creating the page. Please try again."
message = (
"Something went wrong while creating the page. Please try again."
)
return {"status": "error", "message": message}
return create_notion_page

View file

@ -262,10 +262,12 @@ def create_delete_notion_page_tool(
raise
logger.error(f"Error deleting Notion page: {e}", exc_info=True)
if isinstance(e, (ValueError, NotionAPIError)):
if isinstance(e, ValueError | NotionAPIError):
message = str(e)
else:
message = "Something went wrong while deleting the page. Please try again."
message = (
"Something went wrong while deleting the page. Please try again."
)
return {"status": "error", "message": message}
return delete_notion_page

View file

@ -261,10 +261,12 @@ def create_update_notion_page_tool(
raise
logger.error(f"Error updating Notion page: {e}", exc_info=True)
if isinstance(e, (ValueError, NotionAPIError)):
if isinstance(e, ValueError | NotionAPIError):
message = str(e)
else:
message = "Something went wrong while updating the page. Please try again."
message = (
"Something went wrong while updating the page. Please try again."
)
return {"status": "error", "message": message}
return update_notion_page

File diff suppressed because it is too large Load diff

View file

@ -30,6 +30,7 @@ class LinearAPIError(Exception):
without any additional prefix or wrapping.
"""
ORGANIZATION_QUERY = """
query {
organization {
@ -267,7 +268,10 @@ class LinearConnector:
if errors:
ext = errors[0].get("extensions", {})
code = ext.get("code", "")
if code == "INPUT_ERROR" and "too complex" in errors[0].get("message", "").lower():
if (
code == "INPUT_ERROR"
and "too complex" in errors[0].get("message", "").lower()
):
friendly = (
"Linear rejected the request because the workspace is too large "
"to fetch in one query. Please try again — if the problem persists, "

View file

@ -26,6 +26,7 @@ from .jira_add_connector_route import router as jira_add_connector_router
from .linear_add_connector_route import router as linear_add_connector_router
from .logs_routes import router as logs_router
from .luma_add_connector_route import router as luma_add_connector_router
from .model_list_routes import router as model_list_router
from .new_chat_routes import router as new_chat_router
from .new_llm_config_routes import router as new_llm_config_router
from .notes_routes import router as notes_router
@ -68,6 +69,7 @@ router.include_router(jira_add_connector_router)
router.include_router(confluence_add_connector_router)
router.include_router(clickup_add_connector_router)
router.include_router(new_llm_config_router) # LLM configs with prompt configuration
router.include_router(model_list_router) # Dynamic LLM model catalogue from OpenRouter
router.include_router(logs_router)
router.include_router(circleback_webhook_router) # Circleback meeting webhooks
router.include_router(surfsense_docs_router) # Surfsense documentation for citations

View file

@ -0,0 +1,44 @@
"""
API route for fetching the available LLM models catalogue.
Serves a dynamically-updated list sourced from the OpenRouter public API,
with a local JSON fallback when the API is unreachable.
"""
import logging
from fastapi import APIRouter, Depends, HTTPException
from pydantic import BaseModel
from app.db import User
from app.services.model_list_service import get_model_list
from app.users import current_active_user
router = APIRouter()
logger = logging.getLogger(__name__)
class ModelListItem(BaseModel):
value: str
label: str
provider: str
context_window: str | None = None
@router.get("/models", response_model=list[ModelListItem])
async def list_available_models(
user: User = Depends(current_active_user),
):
"""
Return all available LLM models grouped by provider.
The list is sourced from the OpenRouter public API and cached for 1 hour.
If the API is unreachable, a local fallback file is used instead.
"""
try:
return await get_model_list()
except Exception as e:
logger.exception("Failed to fetch model list")
raise HTTPException(
status_code=500, detail=f"Failed to fetch model list: {e!s}"
) from e

View file

@ -110,13 +110,15 @@ class LinearToolMetadataService:
teams = await self._fetch_teams_context(linear_client)
except Exception as e:
return {"error": f"Failed to fetch Linear context: {e!s}"}
workspaces.append({
"id": workspace.id,
"name": workspace.name,
"organization_name": workspace.organization_name,
"teams": teams,
"priorities": priorities,
})
workspaces.append(
{
"id": workspace.id,
"name": workspace.name,
"organization_name": workspace.organization_name,
"teams": teams,
"priorities": priorities,
}
)
return {"workspaces": workspaces}
@ -307,16 +309,10 @@ class LinearToolMetadataService:
Document.document_type == DocumentType.LINEAR_CONNECTOR,
SearchSourceConnector.user_id == user_id,
or_(
func.lower(
Document.document_metadata.op("->>")(
"issue_title"
)
)
func.lower(Document.document_metadata.op("->>")("issue_title"))
== ref_lower,
func.lower(
Document.document_metadata.op("->>")(
"issue_identifier"
)
Document.document_metadata.op("->>")("issue_identifier")
)
== ref_lower,
func.lower(Document.title) == ref_lower,

View file

@ -162,7 +162,10 @@ async def validate_llm_config(
async def get_search_space_llm_instance(
session: AsyncSession, search_space_id: int, role: str, disable_streaming: bool = False
session: AsyncSession,
search_space_id: int,
role: str,
disable_streaming: bool = False,
) -> ChatLiteLLM | ChatLiteLLMRouter | None:
"""
Get a ChatLiteLLM instance for a specific search space and role.
@ -384,16 +387,24 @@ async def get_document_summary_llm(
) -> ChatLiteLLM | ChatLiteLLMRouter | None:
"""Get the search space's document summary LLM instance."""
return await get_search_space_llm_instance(
session, search_space_id, LLMRole.DOCUMENT_SUMMARY, disable_streaming=disable_streaming
session,
search_space_id,
LLMRole.DOCUMENT_SUMMARY,
disable_streaming=disable_streaming,
)
# Backward-compatible alias (LLM preferences are now per-search-space, not per-user)
async def get_user_long_context_llm(
session: AsyncSession, user_id: str, search_space_id: int, disable_streaming: bool = False
session: AsyncSession,
user_id: str,
search_space_id: int,
disable_streaming: bool = False,
) -> ChatLiteLLM | ChatLiteLLMRouter | None:
"""
Deprecated: Use get_document_summary_llm instead.
The user_id parameter is ignored as LLM preferences are now per-search-space.
"""
return await get_document_summary_llm(session, search_space_id, disable_streaming=disable_streaming)
return await get_document_summary_llm(
session, search_space_id, disable_streaming=disable_streaming
)

View file

@ -0,0 +1,167 @@
"""
Service for fetching and caching the available LLM model list.
Uses the OpenRouter public API as the primary source, with a local
fallback JSON file when the API is unreachable.
"""
import json
import logging
import time
from pathlib import Path
import httpx
logger = logging.getLogger(__name__)
OPENROUTER_API_URL = "https://openrouter.ai/api/v1/models"
FALLBACK_FILE = Path(__file__).parent.parent / "config" / "model_list_fallback.json"
CACHE_TTL_SECONDS = 86400 # 24 hours
# In-memory cache
_cache: list[dict] | None = None
_cache_timestamp: float = 0
# Maps OpenRouter provider slug → our LiteLLMProvider enum value.
# Only providers where the model-name part (after the slash) can be
# used directly with the native provider's litellm prefix are listed.
#
# Excluded slugs and why:
# "deepseek" - Native API only accepts "deepseek-chat" / "deepseek-reasoner";
# OpenRouter uses different names (deepseek-v3.2, deepseek-r1, ...).
# "qwen" - Most OpenRouter Qwen entries are open-source models (qwen3-32b, ...)
# that are NOT available on the Dashscope API.
# "ai21" - OpenRouter name "jamba-large-1.7" != AI21 API name "jamba-1.5-large".
# "microsoft" - OpenRouter "microsoft/" = open-source Phi/WizardLM, NOT Azure
# OpenAI deployments (which require deployment names, not model ids).
OPENROUTER_SLUG_TO_PROVIDER: dict[str, str] = {
"openai": "OPENAI",
"anthropic": "ANTHROPIC",
"google": "GOOGLE",
"mistralai": "MISTRAL",
"cohere": "COHERE",
"x-ai": "XAI",
"perplexity": "PERPLEXITY",
}
def _format_context_length(length: int | None) -> str | None:
"""Convert a raw token count to a human-readable string (e.g. 128K, 1M)."""
if not length:
return None
if length >= 1_000_000:
return f"{length / 1_000_000:g}M"
if length >= 1_000:
return f"{length / 1_000:g}K"
return str(length)
async def _fetch_from_openrouter() -> list[dict] | None:
"""Try fetching the model catalogue from the OpenRouter public API."""
try:
async with httpx.AsyncClient(timeout=15) as client:
response = await client.get(OPENROUTER_API_URL)
response.raise_for_status()
data = response.json()
return data.get("data", [])
except Exception as e:
logger.warning("Failed to fetch from OpenRouter API: %s", e)
return None
def _load_fallback() -> list[dict]:
"""Load the local fallback model list."""
try:
with open(FALLBACK_FILE, encoding="utf-8") as f:
data = json.load(f)
return data.get("data", [])
except Exception as e:
logger.error("Failed to load fallback model list: %s", e)
return []
def _is_text_output_model(model: dict) -> bool:
"""Return True if the model's output is text-only (no audio/image generation)."""
output_mods = model.get("architecture", {}).get("output_modalities", [])
return output_mods == ["text"]
def _process_models(raw_models: list[dict]) -> list[dict]:
"""
Transform raw OpenRouter model entries into a flat list of
{value, label, provider, context_window} dicts.
Only text-output models are included (audio/image generators are skipped).
Each OpenRouter model is emitted once for OPENROUTER (full id) and,
when the slug maps to a native provider, once more with just the
model-name portion.
"""
processed: list[dict] = []
for model in raw_models:
model_id: str = model.get("id", "")
name: str = model.get("name", "")
context_length = model.get("context_length")
if "/" not in model_id:
continue
if not _is_text_output_model(model):
continue
provider_slug, model_name = model_id.split("/", 1)
context_window = _format_context_length(context_length)
# 1) Always emit for OPENROUTER (value = full OpenRouter id)
processed.append(
{
"value": model_id,
"label": name,
"provider": "OPENROUTER",
"context_window": context_window,
}
)
# 2) Emit for the native provider when we have a mapping
native_provider = OPENROUTER_SLUG_TO_PROVIDER.get(provider_slug)
if native_provider:
# Google's Gemini API only serves gemini-* models.
# Open-source models like gemma-* are NOT available through it.
if native_provider == "GOOGLE" and not model_name.startswith("gemini-"):
continue
processed.append(
{
"value": model_name,
"label": name,
"provider": native_provider,
"context_window": context_window,
}
)
return processed
async def get_model_list() -> list[dict]:
"""
Return the processed model list, using in-memory cache when fresh.
Tries the OpenRouter API first, falls back to the local JSON file.
"""
global _cache, _cache_timestamp
if _cache is not None and (time.time() - _cache_timestamp) < CACHE_TTL_SECONDS:
return _cache
raw_models = await _fetch_from_openrouter()
if raw_models is None:
logger.info("Using fallback model list")
raw_models = _load_fallback()
processed = _process_models(raw_models)
_cache = processed
_cache_timestamp = time.time()
return processed

View file

@ -57,7 +57,9 @@ class NotionKBSyncService:
)
logger.debug(f"Fetching page content from Notion for page {page_id}")
blocks, _ = await notion_connector.get_page_content(page_id, page_title=None)
blocks, _ = await notion_connector.get_page_content(
page_id, page_title=None
)
from app.utils.notion_utils import extract_all_block_ids, process_blocks
@ -100,11 +102,16 @@ class NotionKBSyncService:
full_content = fetched_content
content_verified = False
logger.debug(f"Final content length: {len(full_content)} chars, verified={content_verified}")
logger.debug(
f"Final content length: {len(full_content)} chars, verified={content_verified}"
)
logger.debug("Generating summary and embeddings")
user_llm = await get_user_long_context_llm(
self.db_session, user_id, search_space_id, disable_streaming=True # disable streaming to avoid leaking into the chat
self.db_session,
user_id,
search_space_id,
disable_streaming=True, # disable streaming to avoid leaking into the chat
)
if user_llm:

View file

@ -0,0 +1,262 @@
import Link from "next/link";
import { UPTIME_REPORT_URL } from "@/lib/env-config";
type UptimeStatus = "up" | "down";
interface LocationStat {
uptime_status: UptimeStatus;
response_time: number | null;
last_check: number;
}
interface UptimeMonitor {
id: string;
name: string;
type: string;
target: string;
last_check: number;
uptime_status: UptimeStatus;
monitor_status: string;
uptime: number;
locations?: Record<string, LocationStat>;
}
interface UptimeMonitorsApiResponse {
monitors?: unknown[];
}
const HETRIXTOOLS_API_BASE = "https://api.hetrixtools.com/v3";
function formatTimestamp(timestamp: number) {
if (!Number.isFinite(timestamp) || timestamp <= 0) return "n/a";
return new Date(timestamp * 1000).toLocaleString();
}
function formatLocationName(location: string) {
return location
.replaceAll("_", " ")
.split(" ")
.map((part) => part.charAt(0).toUpperCase() + part.slice(1))
.join(" ");
}
function toNumber(value: unknown, fallback = 0) {
if (typeof value === "number" && Number.isFinite(value)) return value;
if (typeof value === "string") {
const parsed = Number.parseFloat(value);
if (Number.isFinite(parsed)) return parsed;
}
return fallback;
}
function normalizeUptimeStatus(value: unknown): UptimeStatus {
return value === "down" ? "down" : "up";
}
function normalizeMonitor(rawMonitor: unknown): UptimeMonitor | null {
if (!rawMonitor || typeof rawMonitor !== "object") return null;
const monitor = rawMonitor as Record<string, unknown>;
const rawLocations =
monitor.locations && typeof monitor.locations === "object"
? (monitor.locations as Record<string, unknown>)
: {};
const locations: Record<string, LocationStat> = {};
for (const [locationName, rawLocation] of Object.entries(rawLocations)) {
if (!rawLocation || typeof rawLocation !== "object") continue;
const location = rawLocation as Record<string, unknown>;
locations[locationName] = {
uptime_status: normalizeUptimeStatus(location.uptime_status),
response_time:
location.response_time === null ? null : toNumber(location.response_time, 0),
last_check: toNumber(location.last_check, 0),
};
}
return {
id: String(monitor.id ?? ""),
name: String(monitor.name ?? "Unnamed monitor"),
type: String(monitor.type ?? ""),
target: String(monitor.target ?? ""),
last_check: toNumber(monitor.last_check, 0),
uptime_status: normalizeUptimeStatus(monitor.uptime_status),
monitor_status: String(monitor.monitor_status ?? "unknown"),
uptime: toNumber(monitor.uptime, 0),
locations,
};
}
async function fetchUptimeMonitors(): Promise<{
monitors: UptimeMonitor[];
error?: string;
}> {
const apiKey = process.env.HETRIXTOOLS_API_KEY;
const monitorId = process.env.HETRIXTOOLS_MONITOR_ID;
if (!apiKey) {
return {
monitors: [],
error:
"Missing HETRIXTOOLS_API_KEY. Add it to your server environment to enable custom uptime UI.",
};
}
const query = monitorId
? `id=${encodeURIComponent(monitorId)}`
: "per_page=20&page=1&order_by=last_check&order=desc";
try {
const response = await fetch(`${HETRIXTOOLS_API_BASE}/uptime-monitors?${query}`, {
method: "GET",
headers: {
Authorization: `Bearer ${apiKey}`,
},
next: { revalidate: 60 },
});
if (!response.ok) {
return {
monitors: [],
error: `HetrixTools API request failed (${response.status}).`,
};
}
const data = (await response.json()) as UptimeMonitorsApiResponse;
const monitors = (data.monitors ?? [])
.map((monitor) => normalizeMonitor(monitor))
.filter((monitor): monitor is UptimeMonitor => monitor !== null);
return { monitors };
} catch {
return {
monitors: [],
error: "Could not reach HetrixTools API from the server.",
};
}
}
export default async function UptimePage() {
const { monitors, error } = await fetchUptimeMonitors();
return (
<section className="min-h-screen pt-24 pb-16">
<div className="mx-auto flex w-full max-w-6xl flex-col gap-6 px-4 sm:px-6 lg:px-8">
<div className="rounded-2xl border border-neutral-200/70 bg-white/80 p-6 shadow-sm backdrop-blur-sm dark:border-neutral-800 dark:bg-neutral-950/70">
<p className="text-sm font-medium uppercase tracking-wide text-emerald-600 dark:text-emerald-400">
System Status
</p>
<h1 className="mt-2 text-3xl font-bold tracking-tight text-neutral-900 dark:text-neutral-100">
SurfSense uptime dashboard
</h1>
<div className="mt-4 flex flex-wrap items-center gap-3 text-sm">
<Link
href={UPTIME_REPORT_URL}
target="_blank"
rel="noopener noreferrer"
className="inline-flex items-center rounded-full border border-neutral-300 px-4 py-2 font-medium text-neutral-700 transition-colors hover:bg-neutral-100 dark:border-neutral-700 dark:text-neutral-200 dark:hover:bg-neutral-900"
>
Open original report
</Link>
<span className="text-xs text-neutral-500 dark:text-neutral-400">
Source: HetrixTools v3 API (`/uptime-monitors`).
</span>
</div>
</div>
{error ? (
<div className="rounded-2xl border border-amber-300 bg-amber-50 p-5 text-amber-900 dark:border-amber-700/70 dark:bg-amber-950/30 dark:text-amber-200">
<p className="font-semibold">Unable to load custom uptime data</p>
<p className="mt-1 text-sm">{error}</p>
</div>
) : monitors.length === 0 ? (
<div className="rounded-2xl border border-neutral-200/70 bg-white p-5 text-neutral-700 shadow-sm dark:border-neutral-800 dark:bg-neutral-950 dark:text-neutral-300">
No uptime monitors returned by HetrixTools API.
</div>
) : (
<div className="grid gap-4">
{monitors.map((monitor) => {
const locations = Object.entries(monitor.locations ?? {});
const isUp = monitor.uptime_status === "up";
return (
<div
key={monitor.id}
className="rounded-2xl border border-neutral-200/70 bg-white p-5 shadow-sm dark:border-neutral-800 dark:bg-neutral-950"
>
<div className="flex flex-wrap items-start justify-between gap-4">
<div>
<p className="text-lg font-semibold text-neutral-900 dark:text-neutral-100">
{monitor.name}
</p>
<p className="mt-1 text-sm text-neutral-600 dark:text-neutral-400">
{monitor.target || "No target shown"}
</p>
</div>
<div
className={`rounded-full px-3 py-1 text-xs font-semibold ${
isUp
? "bg-emerald-100 text-emerald-800 dark:bg-emerald-900/40 dark:text-emerald-300"
: "bg-red-100 text-red-800 dark:bg-red-900/40 dark:text-red-300"
}`}
>
{isUp ? "Operational" : "Outage"}
</div>
</div>
<div className="mt-4 grid gap-3 sm:grid-cols-3">
<div className="rounded-xl border border-neutral-200/70 p-3 dark:border-neutral-800">
<p className="text-xs text-neutral-500 dark:text-neutral-400">Uptime</p>
<p className="mt-1 text-lg font-semibold text-neutral-900 dark:text-neutral-100">
{monitor.uptime.toFixed(4)}%
</p>
</div>
<div className="rounded-xl border border-neutral-200/70 p-3 dark:border-neutral-800">
<p className="text-xs text-neutral-500 dark:text-neutral-400">
Last check
</p>
<p className="mt-1 text-sm font-medium text-neutral-900 dark:text-neutral-100">
{formatTimestamp(monitor.last_check)}
</p>
</div>
<div className="rounded-xl border border-neutral-200/70 p-3 dark:border-neutral-800">
<p className="text-xs text-neutral-500 dark:text-neutral-400">
Monitor status
</p>
<p className="mt-1 text-sm font-medium capitalize text-neutral-900 dark:text-neutral-100">
{monitor.monitor_status.replaceAll("_", " ")}
</p>
</div>
</div>
{locations.length > 0 && (
<div className="mt-4">
<p className="text-xs font-semibold uppercase tracking-wide text-neutral-500 dark:text-neutral-400">
Locations
</p>
<div className="mt-2 grid gap-2 sm:grid-cols-2 lg:grid-cols-3">
{locations.map(([locationName, locationData]) => (
<div
key={`${monitor.id}-${locationName}`}
className="rounded-xl border border-neutral-200/70 p-3 dark:border-neutral-800"
>
<p className="text-sm font-medium text-neutral-900 dark:text-neutral-100">
{formatLocationName(locationName)}
</p>
<p className="mt-1 text-xs text-neutral-600 dark:text-neutral-400">
{locationData.uptime_status === "up" ? "Up" : "Down"} ·{" "}
{locationData.response_time ?? "n/a"} ms
</p>
</div>
))}
</div>
</div>
)}
</div>
);
})}
</div>
)}
</div>
</section>
);
}

View file

@ -1,4 +1,6 @@
import { atomWithQuery } from "jotai-tanstack-query";
import type { LLMModel } from "@/contracts/enums/llm-models";
import { LLM_MODELS } from "@/contracts/enums/llm-models";
import { newLLMConfigApiService } from "@/lib/apis/new-llm-config-api.service";
import { cacheKeys } from "@/lib/query-client/cache-keys";
import { activeSearchSpaceIdAtom } from "../search-spaces/search-space-query.atoms";
@ -62,3 +64,33 @@ export const defaultSystemInstructionsAtom = atomWithQuery(() => {
},
};
});
/**
* Query atom for the dynamic LLM model catalogue.
* Fetched from the backend (which proxies OpenRouter's public API).
* Falls back to the static hardcoded list on error.
*/
export const modelListAtom = atomWithQuery(() => {
return {
queryKey: cacheKeys.newLLMConfigs.modelList(),
staleTime: 60 * 60 * 1000, // 1 hour - models don't change often
placeholderData: LLM_MODELS,
queryFn: async (): Promise<LLMModel[]> => {
const data = await newLLMConfigApiService.getModels();
const dynamicModels = data.map((m) => ({
value: m.value,
label: m.label,
provider: m.provider,
contextWindow: m.context_window ?? undefined,
}));
// Providers covered by the dynamic API (from OpenRouter mapping).
// For uncovered providers (Ollama, Groq, Bedrock, etc.) keep the
// hand-curated static suggestions so users still see model options.
const coveredProviders = new Set(dynamicModels.map((m) => m.provider));
const staticFallbacks = LLM_MODELS.filter((m) => !coveredProviders.has(m.provider));
return [...dynamicModels, ...staticFallbacks];
},
};
});

View file

@ -147,7 +147,7 @@ export function ChatShareButton({ thread, onVisibilityChange, className }: ChatS
onClick={() => router.push(`/dashboard/${params.search_space_id}/settings`)}
className="flex items-center justify-center h-8 w-8 rounded-md bg-muted/50 hover:bg-muted transition-colors"
>
<Globe className="h-4 w-4 text-muted-foreground" />
<Earth className="h-4 w-4 text-muted-foreground" />
</button>
</TooltipTrigger>
<TooltipContent>

View file

@ -13,10 +13,13 @@ import {
Sparkles,
} from "lucide-react";
import { AnimatePresence, motion } from "motion/react";
import { useEffect, useState } from "react";
import { useEffect, useMemo, useState } from "react";
import { useForm } from "react-hook-form";
import { z } from "zod";
import { defaultSystemInstructionsAtom } from "@/atoms/new-llm-config/new-llm-config-query.atoms";
import {
defaultSystemInstructionsAtom,
modelListAtom,
} from "@/atoms/new-llm-config/new-llm-config-query.atoms";
import { Badge } from "@/components/ui/badge";
import { Button } from "@/components/ui/button";
import { Collapsible, CollapsibleContent, CollapsibleTrigger } from "@/components/ui/collapsible";
@ -50,7 +53,6 @@ import { Separator } from "@/components/ui/separator";
import { Spinner } from "@/components/ui/spinner";
import { Switch } from "@/components/ui/switch";
import { Textarea } from "@/components/ui/textarea";
import { getModelsByProvider } from "@/contracts/enums/llm-models";
import { LLM_PROVIDERS } from "@/contracts/enums/llm-providers";
import type { CreateNewLLMConfigRequest } from "@/contracts/types/new-llm-config.types";
import { cn } from "@/lib/utils";
@ -66,7 +68,7 @@ const formSchema = z.object({
api_key: z.string().min(1, "API key is required"),
api_base: z.string().max(500).optional().nullable(),
litellm_params: z.record(z.string(), z.any()).optional().nullable(),
system_instructions: z.string().optional().default(""),
system_instructions: z.string().default(""),
use_default_system_instructions: z.boolean().default(true),
citations_enabled: z.boolean().default(true),
search_space_id: z.number(),
@ -74,7 +76,7 @@ const formSchema = z.object({
type FormValues = z.infer<typeof formSchema>;
export interface LLMConfigFormData extends CreateNewLLMConfigRequest {}
export type LLMConfigFormData = CreateNewLLMConfigRequest;
interface LLMConfigFormProps {
initialData?: Partial<LLMConfigFormData>;
@ -102,12 +104,14 @@ export function LLMConfigForm({
const { data: defaultInstructions, isSuccess: defaultInstructionsLoaded } = useAtomValue(
defaultSystemInstructionsAtom
);
const { data: dynamicModels } = useAtomValue(modelListAtom);
const [modelComboboxOpen, setModelComboboxOpen] = useState(false);
const [advancedOpen, setAdvancedOpen] = useState(false);
const [systemInstructionsOpen, setSystemInstructionsOpen] = useState(false);
const form = useForm<FormValues>({
resolver: zodResolver(formSchema),
// eslint-disable-next-line @typescript-eslint/no-explicit-any
resolver: zodResolver(formSchema) as any,
defaultValues: {
name: initialData?.name ?? "",
description: initialData?.description ?? "",
@ -138,7 +142,10 @@ export function LLMConfigForm({
const watchProvider = form.watch("provider");
const selectedProvider = LLM_PROVIDERS.find((p) => p.value === watchProvider);
const availableModels = watchProvider ? getModelsByProvider(watchProvider) : [];
const availableModels = useMemo(
() => (dynamicModels ?? []).filter((m) => m.provider === watchProvider),
[dynamicModels, watchProvider]
);
const handleProviderChange = (value: string) => {
form.setValue("provider", value);
@ -293,57 +300,58 @@ export function LLMConfigForm({
</FormControl>
</PopoverTrigger>
<PopoverContent className="w-full p-0" align="start">
<Command shouldFilter={false}>
<CommandInput
placeholder={selectedProvider?.example || "Type model name..."}
value={field.value}
onValueChange={field.onChange}
/>
<CommandList>
<CommandEmpty>
<div className="py-3 text-center text-sm text-muted-foreground">
{field.value ? `Using: "${field.value}"` : "Type your model name"}
</div>
</CommandEmpty>
{availableModels.length > 0 && (
<CommandGroup heading="Suggested Models">
{availableModels
.filter(
(model) =>
!field.value ||
model.value.toLowerCase().includes(field.value.toLowerCase())
)
.slice(0, 8)
.map((model) => (
<CommandItem
key={model.value}
value={model.value}
onSelect={(value) => {
field.onChange(value);
setModelComboboxOpen(false);
}}
className="py-2"
>
<Check
className={cn(
"mr-2 h-4 w-4",
field.value === model.value ? "opacity-100" : "opacity-0"
)}
/>
<div>
<div className="font-medium">{model.label}</div>
{model.contextWindow && (
<div className="text-xs text-muted-foreground">
Context: {model.contextWindow}
</div>
)}
</div>
</CommandItem>
))}
</CommandGroup>
)}
</CommandList>
</Command>
<Command shouldFilter={false}>
<CommandInput
placeholder={selectedProvider?.example || "Type model name..."}
value={field.value}
onValueChange={field.onChange}
/>
<CommandList className="max-h-[300px]">
<CommandEmpty>
<div className="py-3 text-center text-sm text-muted-foreground">
{field.value ? `Using: "${field.value}"` : "Type your model name"}
</div>
</CommandEmpty>
{availableModels.length > 0 && (
<CommandGroup heading="Suggested Models">
{availableModels
.filter(
(model) =>
!field.value ||
model.value.toLowerCase().includes(field.value.toLowerCase()) ||
model.label.toLowerCase().includes(field.value.toLowerCase())
)
.slice(0, 50)
.map((model) => (
<CommandItem
key={model.value}
value={model.value}
onSelect={(value) => {
field.onChange(value);
setModelComboboxOpen(false);
}}
className="py-2"
>
<Check
className={cn(
"mr-2 h-4 w-4",
field.value === model.value ? "opacity-100" : "opacity-0"
)}
/>
<div>
<div className="font-medium">{model.label}</div>
{model.contextWindow && (
<div className="text-xs text-muted-foreground">
Context: {model.contextWindow}
</div>
)}
</div>
</CommandItem>
))}
</CommandGroup>
)}
</CommandList>
</Command>
</PopoverContent>
</Popover>
{selectedProvider?.example && (
@ -376,7 +384,7 @@ export function LLMConfigForm({
</FormControl>
{watchProvider === "OLLAMA" && (
<FormDescription className="text-[10px] sm:text-xs">
Ollama doesn't require auth enter any value
Ollama doesn&apos;t require auth enter any value
</FormDescription>
)}
<FormMessage />
@ -537,7 +545,7 @@ export function LLMConfigForm({
/>
</FormControl>
<FormDescription className="text-[10px] sm:text-xs">
Use {"{resolved_today}"} to include today's date dynamically
Use {"{resolved_today}"} to include today&apos;s date dynamically
</FormDescription>
<FormMessage />
</FormItem>

View file

@ -258,30 +258,6 @@ export const LLM_MODELS: LLMModel[] = [
provider: "DEEPSEEK",
contextWindow: "128K",
},
{
value: "deepseek-chat",
label: "DeepSeek Chat V3",
provider: "DEEPSEEK",
contextWindow: "66K",
},
{
value: "deepseek-v3",
label: "DeepSeek V3",
provider: "DEEPSEEK",
contextWindow: "66K",
},
{
value: "deepseek-r1",
label: "DeepSeek R1",
provider: "DEEPSEEK",
contextWindow: "66K",
},
{
value: "deepseek-r1-0528",
label: "DeepSeek R1 (0528)",
provider: "DEEPSEEK",
contextWindow: "65K",
},
// xAI (Grok)
{ value: "grok-4", label: "Grok 4", provider: "XAI", contextWindow: "256K" },
@ -1134,7 +1110,7 @@ export const LLM_MODELS: LLMModel[] = [
contextWindow: "8K",
},
{
value: "mixtral-8x7B-Instruct-v0.1",
value: "mixtral",
label: "Ollama Mixtral 8x7B",
provider: "OLLAMA",
contextWindow: "33K",
@ -1236,13 +1212,13 @@ export const LLM_MODELS: LLMModel[] = [
// Zhipu (GLM)
{
value: "z-ai/glm-4.6",
value: "glm-4.6",
label: "GLM 4.6",
provider: "ZHIPU",
contextWindow: "203K",
},
{
value: "z-ai/glm-4.6:exacto",
value: "glm-4.6:exacto",
label: "GLM 4.6 Exacto",
provider: "ZHIPU",
contextWindow: "203K",
@ -1350,7 +1326,7 @@ export const LLM_MODELS: LLMModel[] = [
contextWindow: "128K",
},
{
value: "openai/gpt-oss-120b",
value: "gpt-oss-120b",
label: "Cerebras GPT-OSS-120B",
provider: "CEREBRAS",
contextWindow: "131K",

View file

@ -128,7 +128,7 @@ export const LLM_PROVIDERS: LLMProvider[] = [
{
value: "ZHIPU",
label: "Zhipu (GLM)",
example: "openrouter/z-ai/glm-4.6",
example: "glm-4.6, glm-4.6:exacto",
description: "GLM series models",
apiBase: "https://open.bigmodel.cn/api/paas/v4",
},

View file

@ -291,6 +291,22 @@ export const updateLLMPreferencesRequest = z.object({
export const updateLLMPreferencesResponse = llmPreferences;
// =============================================================================
// Model List (dynamic catalogue from OpenRouter API)
// =============================================================================
export const modelListItem = z.object({
value: z.string(),
label: z.string(),
provider: z.string(),
context_window: z.string().nullable().optional(),
});
export const getModelListResponse = z.array(modelListItem);
export type ModelListItem = z.infer<typeof modelListItem>;
export type GetModelListResponse = z.infer<typeof getModelListResponse>;
// =============================================================================
// Type Exports
// =============================================================================

View file

@ -10,6 +10,7 @@ import {
getDefaultSystemInstructionsResponse,
getGlobalNewLLMConfigsResponse,
getLLMPreferencesResponse,
getModelListResponse,
getNewLLMConfigRequest,
getNewLLMConfigResponse,
getNewLLMConfigsRequest,
@ -145,6 +146,13 @@ class NewLLMConfigApiService {
);
};
/**
* Get the dynamic LLM model catalogue (sourced from OpenRouter API)
*/
getModels = async () => {
return baseApiService.get(`/api/v1/models`, getModelListResponse);
};
/**
* Update LLM preferences for a search space
*/

View file

@ -33,6 +33,7 @@ export const cacheKeys = {
preferences: (searchSpaceId: number) => ["llm-preferences", searchSpaceId] as const,
defaultInstructions: () => ["new-llm-configs", "default-instructions"] as const,
global: () => ["new-llm-configs", "global"] as const,
modelList: () => ["models", "catalogue"] as const,
},
imageGenConfigs: {
all: (searchSpaceId: number) => ["image-gen-configs", searchSpaceId] as const,