feat: merge Release v1.8.0 into main

[vjgit96]

Summary by CodeRabbit

• New Features

  • Windows PostgreSQL event loop configuration to resolve compatibility issues
  • Double-bracket syntax support for prompt templating
  • Context ID propagation across chat components
  • Max tokens parameter for language model control
  • IBM WatsonX provider configuration options
  • Safari scroll stabilization for chat interface
  • File upload clear/cancel functionality
  • Flow locking capability

• Bug Fixes

  • Users prevented from deactivating their own accounts
  • Fixed asyncio event loop incompatibility on Windows with PostgreSQL

• Documentation

  • Windows PostgreSQL Event Loop Fix feature specification

• Tests

  • Comprehensive Windows PostgreSQL integration test suite
  • User account management tests
  • Template field validation tests
  • UI regression and component behavior tests

[github-actions]

Frontend Unit Test Coverage Report

Coverage Summary

Lines	Statements	Branches	Functions
	18.81% (6095/32395)	12.25% (3097/25264)	12.64% (879/6951)

Unit Test Results

Tests	Skipped	Failures	Errors	Time
2310	0 💤	0 ❌	0 🔥	31.836s ⏱️

[codecov]

Codecov Report

❌ Patch coverage is 20.39660% with 281 lines in your changes missing coverage. Please review.
✅ Project coverage is 35.34%. Comparing base (38082c1) to head (4c977a2).
⚠️ Report is 3 commits behind head on main.

Files with missing lines	Patch %	Lines
...nent/components/accordionPromptComponent/index.tsx	18.88%	71 Missing and 2 partials ⚠️
...ontend/src/components/common/safari-scroll-fix.tsx	0.00%	64 Missing ⚠️
...view/chat-header/hooks/use-session-has-messages.ts	0.00%	18 Missing ⚠️
...t-view/chat-input/components/text-area-wrapper.tsx	0.00%	15 Missing ⚠️
src/frontend/src/modals/flowLogsModal/index.tsx	0.00%	12 Missing ⚠️
...ntainer/components/flow-page-sliding-container.tsx	0.00%	9 Missing ⚠️
src/backend/base/langflow/langflow_launcher.py	0.00%	8 Missing ⚠️
.../chat-header/components/chat-sessions-dropdown.tsx	0.00%	8 Missing ⚠️
...aygroundComponent/chat-view/utils/message-utils.ts	0.00%	8 Missing ⚠️
src/backend/base/langflow/__main__.py	0.00%	6 Missing ⚠️
... and 19 more

❌ Your patch status has failed because the patch coverage (20.39%) is below the target coverage (40.00%). You can increase the patch coverage or adjust the target coverage.
❌ Your project status has failed because the head coverage (42.03%) is below the target coverage (60.00%). You can increase the head coverage or adjust the target coverage.

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #11827      +/-   ##
==========================================
+ Coverage   35.27%   35.34%   +0.06%     
==========================================
  Files        1521     1525       +4     
  Lines       73008    73292     +284     
  Branches    10951    11021      +70     
==========================================
+ Hits        25755    25904     +149     
- Misses      45858    45974     +116     
- Partials     1395     1414      +19     

Flag	Coverage Δ
backend	55.88% <68.18%> (-0.09%)	⬇️
frontend	16.99% <13.59%> (+0.33%)	⬆️
lfx	42.03% <ø> (-0.01%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
src/backend/base/langflow/api/v1/users.py	67.81% <100.00%> (+0.75%)	⬆️
...d/base/langflow/helpers/windows_postgres_helper.py	100.00% <100.00%> (ø)
...backend/base/langflow/services/database/service.py	63.19% <100.00%> (+2.69%)	⬆️
src/backend/base/langflow/services/utils.py	81.65% <100.00%> (+0.21%)	⬆️
...onent/chat-view/chat-input/components/no-input.tsx	100.00% <ø> (ø)
src/frontend/src/components/ui/textarea.tsx	87.50% <ø> (ø)
...controllers/API/queries/flows/use-post-add-flow.ts	79.16% <100.00%> (+79.16%)	⬆️
src/frontend/src/customization/constants.ts	100.00% <ø> (ø)
...rc/shared/components/delete-confirmation-modal.tsx	29.16% <ø> (ø)
...d/src/stores/helpers/filter-singleton-component.ts	100.00% <100.00%> (ø)
... and 30 more

... and 21 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[coderabbitai]

Actionable comments posted: 13

Note

Due to the large number of review comments, Critical, Major severity comments were prioritized as inline comments.

[coderabbitai]

⚠️ Potential issue | 🔴 Critical

🧩 Analysis chain

🏁 Script executed:

#!/bin/bash
set -euo pipefail

# Check whether uv.lock includes the new dependency
lockfile=$(fd -t f '^uv\.lock$' -a . | head -n1 || true)
if [ -n "${lockfile}" ]; then
  rg -n "mcp-server-fetch" "${lockfile}" || echo "mcp-server-fetch not found in ${lockfile}"
else
  echo "uv.lock not found"
fi

# Fetch PyPI metadata for version & python compatibility
python - <<'PY'
import json, urllib.request
try:
    url = "https://pypi.org/pypi/mcp-server-fetch/json"
    data = json.load(urllib.request.urlopen(url))
    print("latest_version:", data["info"]["version"])
    print("requires_python:", data["info"].get("requires_python"))
except Exception as e:
    print(f"Error fetching PyPI data: {e}")
PY

Repository: langflow-ai/langflow

Length of output: 291

---

🏁 Script executed:

#!/bin/bash
# Check pyproject.toml around line 73 for context
sed -n '60,85p' pyproject.toml

Repository: langflow-ai/langflow

Length of output: 698

---

Regenerate uv.lock to include the new dev dependency.

The mcp-server-fetch>=2025.1.17 dependency was added to pyproject.toml but is missing from uv.lock. Since the project uses uv for dependency management, the lockfile must be kept in sync. Run uv lock to update uv.lock with the new dependency.

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@pyproject.toml` at line 73, pyproject.toml added the new dev dependency
"mcp-server-fetch>=2025.1.17" but uv.lock wasn't updated; run the UV lock
command to regenerate the lockfile and commit the updated uv.lock so it includes
"mcp-server-fetch>=2025.1.17" (run `uv lock` in the project root, verify uv.lock
contains the mcp-server-fetch entry, then add/commit the updated uv.lock).

[coderabbitai]

⚠️ Potential issue | 🟠 Major

max_tokens is listed in field_order but absent from the template snapshot — field won't render in the starter project.

The field_order array references max_tokens, but the template object for LanguageModelComponent-htMuI (lines 976–1243) contains no max_tokens entry. When Langflow loads this starter project the UI renderer skips fields that exist in field_order but not in template, so users cannot configure max_tokens without first refreshing the component. Runtime is unaffected (build_model guards with getattr(self, "max_tokens", None)), but the intent of exposing the field is defeated.

The serialized max_tokens block should be added to the template section alongside the existing temperature entry, e.g.:

🛠 Proposed addition to the template section (after line 1242, before the closing brace)

+             "max_tokens": {
+               "_input_type": "IntInput",
+               "advanced": true,
+               "display_name": "Max Tokens",
+               "dynamic": false,
+               "info": "Maximum number of tokens to generate. Field name varies by provider.",
+               "list": false,
+               "list_add_label": "Add More",
+               "name": "max_tokens",
+               "placeholder": "",
+               "range_spec": {
+                 "max": 128000,
+                 "min": 1,
+                 "step": 1,
+                 "step_type": "int"
+               },
+               "required": false,
+               "show": true,
+               "title_case": false,
+               "tool_mode": false,
+               "trace_as_metadata": true,
+               "type": "int",
+               "value": ""
+             }

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@src/backend/base/langflow/initial_setup/starter_projects/Document` Q&A.json
around lines 909 - 911, The starter project's LanguageModelComponent-htMuI lists
"max_tokens" in field_order but the template lacks a corresponding serialized
field, so add a max_tokens entry into the component's template object (next to
the existing temperature entry) so the UI can render it; locate
LanguageModelComponent-htMuI in the JSON, update its "template" section to
include a serialized max_tokens block (same structure as temperature: keys like
"name":"max_tokens", "value":<default>, "type":"int" or matching existing field
schema, and any metadata such as "label" and "required") and keep "max_tokens"
in field_order unchanged.

[coderabbitai]

⚠️ Potential issue | 🔴 Critical

🧩 Analysis chain

🏁 Script executed:

#!/bin/bash
# Search for deprecated Gemini 1.5 model references across all starter project JSON files
rg -l 'gemini-1\.5' --glob '*.json' src/backend/base/langflow/initial_setup/starter_projects/

Repository: langflow-ai/langflow

Length of output: 483

---

🌐 Web query:

Are Google Gemini 1.5 pro and flash models still available for inference API in February 2026?

💡 Result:

No.

• Gemini API (Google AI for Developers / AI Studio): Google shut down the Gemini 1.5 models (gemini-1.5-pro, gemini-1.5-flash, and gemini-1.5-flash-8b) on September 29, 2025, so they are not available for inference in February 2026. [1]
• Vertex AI (Gemini on Vertex): The specific Gemini 1.5 Pro/Flash model versions (e.g., gemini-1.5-pro-002, gemini-1.5-flash-002) have retirement dates in 2025 and are listed as retired models, so they are not available for inference in February 2026. [2] (Vertex had also already blocked access for new/no-prior-usage projects as of April 29, 2025.) [3]

Sources: [1] Gemini API release notes, [2] Vertex AI “Model versions and lifecycle”, [3] Vertex AI Generative AI release notes.

---

Remove or replace completely unavailable Gemini 1.5 models.

Google shut down Gemini 1.5 models (gemini-1.5-pro, gemini-1.5-flash, gemini-1.5-flash-8b) on September 29, 2025. These three entries at lines 1028, 1040, and 1052 are completely non-functional in February 2026. Users selecting these options will encounter failures. Replace them with current models like gemini-2.0-flash or gemini-3-pro-preview, or remove them entirely. This issue also appears in at least 4 other starter projects (Text Sentiment Analysis, Simple Agent, Portfolio Website Code Generator, Basic Prompting).

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@src/backend/base/langflow/initial_setup/starter_projects/Financial` Report
Parser.json around lines 1019 - 1054, Remove the three deprecated Gemini 1.5
entries ("gemini-1.5-pro", "gemini-1.5-flash", "gemini-1.5-flash-8b") from the
Financial Report Parser starter JSON and replace them with supported model
entries (for example "gemini-2.0-flash" or "gemini-3-pro-preview"), preserving
the existing metadata structure (metadata.model_class:
"ChatGoogleGenerativeAIFixed", metadata.api_key_param, metadata.context_length,
metadata.model_name_param) and the provider/icon/category fields; also apply the
same replacement or removal to the other affected starter projects (Text
Sentiment Analysis, Simple Agent, Portfolio Website Code Generator, Basic
Prompting) so users do not see nonfunctional model options.

[coderabbitai]

⚠️ Potential issue | 🟠 Major

Dead CSS rule: .playground-messages-wrapper nested inside .label will never match

CSS nesting inside a style rule creates a descendant selector. This block compiles to:

`@media` not all and (min-resolution: 0.001dpcm) {
  `@supports` (-webkit-appearance: none) {
    .label .playground-messages-wrapper { /* ← descendant selector */
      padding-left: 1rem;
      padding-right: 1rem;
    }
  }
}

In the actual DOM, .playground-messages-wrapper (the <div> at line 199 of flow-page-sliding-container.tsx) is never a child of .label (a user-select utility), so this Safari padding fix is entirely unreachable — it will never be applied.

Move the block to the top level outside .label:

🐛 Proposed fix

 .label {
   user-select: none;
   -webkit-user-select: none; /* Safari */
   -moz-user-select: none;   /* Firefox */
   -ms-user-select: none;
-
-  /* Safari-only fix for playground padding */
-  `@media` not all and (min-resolution: 0.001dpcm) {
-    `@supports` (-webkit-appearance: none) {
-      .playground-messages-wrapper {
-        padding-left: 1rem;
-        padding-right: 1rem;
-      }
-    }
-  }
 }

+/* Safari-only fix for playground padding */
+@media not all and (min-resolution: 0.001dpcm) {
+  `@supports` (-webkit-appearance: none) {
+    .playground-messages-wrapper {
+      padding-left: 1rem;
+      padding-right: 1rem;
+    }
+  }
+}

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@src/frontend/src/App.css` around lines 22 - 29, The nested CSS rule places
.playground-messages-wrapper inside .label so it compiles to a descendant
selector that never matches; move the Safari-specific padding block out of the
.label nesting so the selector targets .playground-messages-wrapper at the top
level inside the `@media/`@supports wrapper. Locate the nested block in App.css
where .label contains the `@media/`@supports rule and extract or duplicate the
.playground-messages-wrapper rule so it exists directly under `@media` not all and
(min-resolution: 0.001dpcm) { `@supports` (-webkit-appearance: none) { ... } },
ensuring the selector is simply .playground-messages-wrapper and not .label
.playground-messages-wrapper.

[coderabbitai]

⚠️ Potential issue | 🔴 Critical

🚨 Rules of Hooks violation: useQuery is called conditionally after an early return.

useQueryClient() is always called (Hook #1), but the return false at Line 19 causes useQuery (Hook #2) to be skipped whenever flowId is falsy. React tracks hooks by call order; when flowId transitions between defined and undefined, the hook count changes and React throws a "Rendered fewer/more hooks than during the previous render" error at runtime.

The enabled: !!flowId option already exists — that is the correct React Query mechanism to conditionally disable the query while keeping the hook call unconditional. Move the guard after all hook calls:

🐛 Proposed fix

 export const useSessionHasMessages = ({
   sessionId,
   flowId,
 }: UseSessionHasMessagesParams): boolean => {
   const queryClient = useQueryClient();
 
-  if (!flowId) {
-    return false;
-  }
-
   const sessionCacheKey = [
     "useGetMessagesQuery",
     { id: flowId, session_id: sessionId },
   ];
 
   const { data: sessionMessages = [] } = useQuery<Message[]>({
     queryKey: sessionCacheKey,
     queryFn: () => {
       const cachedData =
         queryClient.getQueryData<Message[]>(sessionCacheKey) || [];
       return cachedData;
     },
     staleTime: Infinity,
     gcTime: 5 * 60 * 1000,
     structuralSharing: false,
     refetchOnMount: false,
     refetchOnWindowFocus: false,
     enabled: !!flowId,
   });
 
+  if (!flowId) return false;
+
   const hasMessages = sessionMessages.some((message: Message) => {

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In
`@src/frontend/src/components/core/playgroundComponent/chat-view/chat-header/hooks/use-session-has-messages.ts`
around lines 17 - 41, Remove the early return that conditionally skips the
useQuery hook to fix the Rules of Hooks violation: always call useQuery (keep
the existing useQueryClient(), sessionCacheKey, and the useQuery call with
enabled: !!flowId) and then after the hooks complete, check flowId and return
false when appropriate (e.g., if !flowId return false using the already-fetched
sessionMessages). In short, delete the "if (!flowId) return false" before
useQuery, rely on enabled: !!flowId to disable the query, and perform the
flowId-based return after the hooks so hook call order stays stable.

[coderabbitai]

⚠️ Potential issue | 🟠 Major

scapeJSONParse can throw and crash the component if any edge has a malformed targetHandle.

scapeJSONParse wraps JSON.parse, which throws a SyntaxError on invalid input. Because this runs inside a Zustand selector that fires on every edges update, a single bad targetHandle value in the store will bubble up as an uncaught exception, crashing the component tree.

🛡️ Proposed fix — guard with try/catch

  const isConnected = useFlowStore(
    useCallback(
      (state) =>
        state.edges.some(
          (edge) =>
            edge.target === data.id &&
            edge.targetHandle &&
-           scapeJSONParse(edge.targetHandle)?.fieldName === name,
+           (() => {
+             try {
+               return scapeJSONParse(edge.targetHandle!)?.fieldName === name;
+             } catch {
+               return false;
+             }
+           })(),
        ),
      [data.id, name],
    ),
  );

📝 Committable suggestion

‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.

Suggested change

	state.edges.some(
	(edge) =>
	edge.target === data.id &&
	edge.targetHandle &&
	scapeJSONParse(edge.targetHandle)?.fieldName === name,
	state.edges.some(
	(edge) =>
	edge.target === data.id &&
	edge.targetHandle &&
	(() => {
	try {
	return scapeJSONParse(edge.targetHandle!)?.fieldName === name;
	} catch {
	return false;
	}
	})(),

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In
`@src/frontend/src/pages/FlowPage/components/InspectionPanel/components/InspectionPanelEditField.tsx`
around lines 34 - 38, The selector in InspectionPanelEditField uses
scapeJSONParse inside state.edges.some which can throw on malformed
targetHandle; wrap the JSON parse call in a safe parsing guard (either update
scapeJSONParse to catch and return undefined/null on parse errors or replace the
inline call with a try/catch around scapeJSONParse) so malformed
edge.targetHandle values don't throw; ensure the some predicate treats parse
failures as non-matching (e.g., treat undefined result as no fieldName match) so
the selector never throws during edges updates.

[coderabbitai]

⚠️ Potential issue | 🟠 Major

🧩 Analysis chain

🏁 Script executed:

# Find ShadTooltip definition
find . -type f \( -name "*.ts" -o -name "*.tsx" \) | xargs grep -l "ShadTooltip" | head -20

Repository: langflow-ai/langflow

Length of output: 2143

---

🏁 Script executed:

# Search for tooltip component definition in common locations
fd -type f -name "*tooltip*" -o -name "*Tooltip*"

Repository: langflow-ai/langflow

Length of output: 234

---

🏁 Script executed:

# Look at the specific file under review
cat -n src/frontend/src/pages/FlowPage/components/InspectionPanel/components/InspectionPanelEditField.tsx | head -100

Repository: langflow-ai/langflow

Length of output: 3570

---

🏁 Script executed:

# Read ShadTooltip component implementation
cat -n src/frontend/src/components/common/shadTooltipComponent/index.tsx

Repository: langflow-ai/langflow

Length of output: 3183

---

Tooltip will not appear when the button is disabled, leaving users without explanation for why the control is unresponsive.

The ShadTooltip component uses Radix UI's Tooltip with asChild={true} (default), which makes the button the direct trigger element. Disabled buttons don't fire pointer events, so the Radix trigger won't activate. Critically, the "Cannot change visibility of connected handles" message is the one users need to see precisely when isConnected=true—the state that disables the button. Without it, they receive no feedback about why the control is unresponsive.

🛠️ Proposed fix — wrap disabled button in a span

      <ShadTooltip
        content={
          isConnected
            ? "Cannot change visibility of connected handles"
            : isOnCanvas
              ? "Hide"
              : "Show"
        }
        avoidCollisions
      >
+       <span
+         tabIndex={isConnected ? 0 : undefined}
+         className={isConnected ? "cursor-not-allowed" : undefined}
+       >
          <button
            onClick={handleToggleVisibility}
            disabled={isConnected}
            className={cn(
              "flex h-6 w-6 shrink-0 items-center justify-center rounded-full transition-colors",
              isOnCanvas
                ? "bg-primary/10 text-primary hover:bg-primary/20"
                : "bg-muted text-muted-foreground hover:bg-muted-foreground/20 hover:text-foreground",
              isConnected
-               ? "cursor-not-allowed opacity-50 hover:bg-primary/10"
+               ? "pointer-events-none opacity-50"
                : "",
            )}
            data-testid={`show${name}`}
            id={`show${name}`}
            aria-checked={isOnCanvas}
            role="checkbox"
          >
            <IconComponent
              name={isOnCanvas ? "Minus" : "Plus"}
              className="h-3.5 w-3.5"
            />
          </button>
+       </span>
      </ShadTooltip>

📝 Committable suggestion

‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.

Suggested change

	<ShadTooltip
	content={
	isConnected
	? "Cannot change visibility of connected handles"
	: isOnCanvas
	? "Hide"
	: "Show"
	}
	avoidCollisions
	>
	<IconComponent
	name={isOnCanvas ? "Minus" : "Plus"}
	className="h-3.5 w-3.5"
	/>
	</button>
	<button
	onClick={handleToggleVisibility}
	disabled={isConnected}
	className={cn(
	"flex h-6 w-6 shrink-0 items-center justify-center rounded-full transition-colors",
	isOnCanvas
	? "bg-primary/10 text-primary hover:bg-primary/20"
	: "bg-muted text-muted-foreground hover:bg-muted-foreground/20 hover:text-foreground",
	isConnected
	? "cursor-not-allowed opacity-50 hover:bg-primary/10"
	: "",
	)}
	data-testid={`show${name}`}
	id={`show${name}`}
	aria-checked={isOnCanvas}
	role="checkbox"
	>
	<IconComponent
	name={isOnCanvas ? "Minus" : "Plus"}
	className="h-3.5 w-3.5"
	/>
	</button>
	</ShadTooltip>
	<ShadTooltip
	content={
	isConnected
	? "Cannot change visibility of connected handles"
	: isOnCanvas
	? "Hide"
	: "Show"
	}
	avoidCollisions
	>
	<span
	tabIndex={isConnected ? 0 : undefined}
	className={isConnected ? "cursor-not-allowed" : undefined}
	>
	<button
	onClick={handleToggleVisibility}
	disabled={isConnected}
	className={cn(
	"flex h-6 w-6 shrink-0 items-center justify-center rounded-full transition-colors",
	isOnCanvas
	? "bg-primary/10 text-primary hover:bg-primary/20"
	: "bg-muted text-muted-foreground hover:bg-muted-foreground/20 hover:text-foreground",
	isConnected
	? "pointer-events-none opacity-50"
	: "",
	)}
	data-testid={`show${name}`}
	id={`show${name}`}
	aria-checked={isOnCanvas}
	role="checkbox"
	>
	<IconComponent
	name={isOnCanvas ? "Minus" : "Plus"}
	className="h-3.5 w-3.5"
	/>
	</button>
	</span>
	</ShadTooltip>

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In
`@src/frontend/src/pages/FlowPage/components/InspectionPanel/components/InspectionPanelEditField.tsx`
around lines 63 - 95, The tooltip won't show when the button is disabled because
Radix's Tooltip (ShadTooltip) uses the button as the trigger and disabled
buttons don't fire pointer events; update the JSX so that when isConnected is
true the button is wrapped in a non-disabled inline wrapper (e.g., a <span
className="inline-block">) and pass the tooltip trigger to that wrapper instead
of the disabled button; keep the existing button props
(onClick={handleToggleVisibility}, disabled, aria-checked, role, data-testid/id)
and the IconComponent inside, and ensure the wrapper preserves layout
(inline-block) and any cursor/opacity styles so the UI remains identical while
allowing ShadTooltip to show the "Cannot change visibility of connected handles"
message.

[coderabbitai]

⚠️ Potential issue | 🟠 Major

🧩 Analysis chain

🏁 Script executed:

#!/bin/bash
# First, let's read the actual file to see the complete context around lines 97-129
cd src/frontend/src/pages/MainPage/pages/homePage
cat -n index.tsx | sed -n '90,135p'

Repository: langflow-ai/langflow

Length of output: 1694

---

🏁 Script executed:

#!/bin/bash
# Check if ENABLE_MCP flag exists and its value
fd feature-flags.ts -x cat -n {} | head -50

Repository: langflow-ai/langflow

Length of output: 1240

---

🏁 Script executed:

#!/bin/bash
# Search for usages of isEmptyFolder and EmptyFolder to understand the impact
rg "isEmptyFolder|EmptyFolder" -n -B2 -A2 \
  src/frontend/src/pages/MainPage/pages/homePage/

Repository: langflow-ai/langflow

Length of output: 4537

---

🏁 Script executed:

#!/bin/bash
# Check what flows are and understand the is_component property
rg "is_component|flows\?" -n -B2 -A2 \
  src/frontend/src/pages/MainPage/pages/homePage/ | head -60

Repository: langflow-ai/langflow

Length of output: 4260

---

isEmptyFolder incorrectly treats a components-only folder as empty when ENABLE_MCP is enabled

With ENABLE_MCP = true, the guard flow.is_component === false only counts flows — components are excluded from the emptiness check. Since the flows array is pre-filtered by the API based on the current tab (lines 60–68, 70), when you navigate to the components tab, the array contains only components. The check then searches for flows within a components-only array, finds nothing, and sets isEmpty = true. The folder displays the EmptyFolder view instead of the components list.

The tab-switching effect (lines 109–129) cannot recover because it guards with !isEmptyFolder (line 111) and thus never runs for this case.

If the intent is to show "empty only when there are genuinely no items in this folder," the condition should check for items regardless of type:

🐛 Proposed fix

  useEffect(() => {
    const isEmpty =
      flows?.find(
        (flow) =>
          flow.folder_id === (folderId ?? myCollectionId) &&
-         (ENABLE_MCP ? flow.is_component === false : true),
+         true,
      ) === undefined;
    setIsEmptyFolder(isEmpty);
  }, [flows, folderId, myCollectionId]);

Alternatively, if the intent is to exclude components from the "empty" check (treating them as a separate category), rename isEmptyFolder to hasNoFlows and add a corresponding hasNoComponents check to avoid conflating an empty-of-flows folder with an empty folder.

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@src/frontend/src/pages/MainPage/pages/homePage/index.tsx` around lines 97 -
105, The isEmptyFolder computation in the useEffect incorrectly ignores
components when ENABLE_MCP is true; update the logic in the effect that reads
flows to treat any item in the folder as non-empty (i.e., remove the
flow.is_component === false filter) so the existence check uses flows.find(flow
=> flow.folder_id === (folderId ?? myCollectionId)) and then calls
setIsEmptyFolder accordingly; alternatively, if you want to keep flows vs
components distinct, rename isEmptyFolder to hasNoFlows and add a separate
hasNoComponents flag and corresponding checks/setters so components-only folders
are not treated as empty by setIsEmptyFolder.

[coderabbitai]

⚠️ Potential issue | 🔴 Critical

🧩 Analysis chain

🏁 Script executed:

#!/bin/bash
# Find all call-sites of cleanEdges and clearHandlesFromAdvancedFields
echo "=== cleanEdges call sites ==="
rg -n --type=ts --type=tsx -P '\bcleanEdges\s*\(' -A2 -B2

echo ""
echo "=== clearHandlesFromAdvancedFields call sites ==="
rg -n --type=ts --type=tsx -P '\bclearHandlesFromAdvancedFields\s*\(' -A2 -B2

Repository: langflow-ai/langflow

Length of output: 217

---

🏁 Script executed:

# Read the relevant sections from reactflowUtils.ts to verify the logic
echo "=== Reading lines 150-210 (isAdvanced and cleanEdges logic) ==="
sed -n '150,210p' src/frontend/src/utils/reactflowUtils.ts | cat -n

echo ""
echo "=== Reading lines 250-295 (clearHandlesFromAdvancedFields) ==="
sed -n '250,295p' src/frontend/src/utils/reactflowUtils.ts | cat -n

echo ""
echo "=== Reading lines 2018-2030 (createNewFlow, locked property) ==="
sed -n '2018,2030p' src/frontend/src/utils/reactflowUtils.ts | cat -n

Repository: langflow-ai/langflow

Length of output: 4765

---

🏁 Script executed:

# Check the FlowType definition to understand locked property expectations
echo "=== Searching for FlowType definition ==="
rg -n -B3 -A10 'type\s+FlowType|interface\s+FlowType' --type=ts --type=tsx

Repository: langflow-ai/langflow

Length of output: 150

---

🏁 Script executed:

#!/bin/bash
# Search for cleanEdges and clearHandlesFromAdvancedFields call sites without file type restriction
echo "=== cleanEdges call sites ==="
rg -n 'cleanEdges\s*\(' -A2 -B2 2>/dev/null

echo ""
echo "=== clearHandlesFromAdvancedFields call sites ==="
rg -n 'clearHandlesFromAdvancedFields\s*\(' -A2 -B2 2>/dev/null

Repository: langflow-ai/langflow

Length of output: 6137

---

🏁 Script executed:

# Search for FlowType definition without file type restriction
echo "=== FlowType definition ==="
rg -n 'type FlowType|interface FlowType' -A15 2>/dev/null

echo ""
echo "=== Flow type with locked property ==="
rg -n 'locked\s*[:?]' -B5 -A2 2>/dev/null | head -80

Repository: langflow-ai/langflow

Length of output: 9842

---

🏁 Script executed:

# Check where cleanEdges is invoked - search in the file itself for calls
echo "=== Checking invocations in reactflowUtils.ts ==="
grep -n 'cleanEdges(' src/frontend/src/utils/reactflowUtils.ts

echo ""
echo "=== Checking imports/usage of clearHandlesFromAdvancedFields ==="
grep -rn 'clearHandlesFromAdvancedFields' src/frontend/src/ --include="*.ts" 2>/dev/null | head -20

Repository: langflow-ai/langflow

Length of output: 345

---

🏁 Script executed:

# Check the test file for cleanEdges to see if isAdvanced behavior is tested
echo "=== Checking cleanEdges test for isAdvanced ==="
grep -n "isAdvanced\|advanced" src/frontend/src/utils/__tests__/cleanEdges.test.ts -B2 -A2

echo ""
echo "=== Check what createNewFlow.test.ts says about locked ==="
grep -n "locked" src/frontend/src/utils/__tests__/createNewFlow.test.ts -B2 -A2

Repository: langflow-ai/langflow

Length of output: 171

---

🏁 Script executed:

# Check flowStore to understand the complete flow of cleanEdges invocations
echo "=== View context around cleanEdges calls in flowStore.ts ==="
sed -n '215,225p' src/frontend/src/stores/flowStore.ts | cat -n
sed -n '295,305p' src/frontend/src/stores/flowStore.ts | cat -n
sed -n '350,365p' src/frontend/src/stores/flowStore.ts | cat -n
sed -n '658,668p' src/frontend/src/stores/flowStore.ts | cat -n

Repository: langflow-ai/langflow

Length of output: 1860

---

🏁 Script executed:

# Check if there's a difference between "advanced" and "show" properties
echo "=== Searching for show property handling in cleanEdges or related ==="
grep -n "show\|advanced" src/frontend/src/utils/reactflowUtils.ts | head -30

Repository: langflow-ai/langflow

Length of output: 1316

---

isAdvanced ORed unconditionally removes all edges to advanced fields on every cleanEdges call, causing silent data loss

The code at line 158 computes isAdvanced and ORs it into the removal predicate at lines 201-202:

if (
  (scapedJSONStringfy(id) !== targetHandle ||
   (targetNode.data.node?.tool_mode && isToolMode) ||
   isAdvanced) &&
  !isLoopInput
)

This removes ANY edge whose target field has advanced: true, even when scapedJSONStringfy(id) === targetHandle (the handle is still perfectly valid). Since cleanEdges is called on every flow state update (in resetFlow, setNodes, paste operations, and unselectAll in flowStore.ts), edges to advanced fields are silently removed on load and paste cycles. This makes it impossible for users to maintain persistent connections to advanced fields.

This contradicts the existing clearHandlesFromAdvancedFields function (lines 257–289), which is explicitly event-driven—it removes edges only when code in the modal changes—and is now rendered redundant. The filterHiddenFieldsEdges function already handles the distinct show === false case; conflating that with advanced: true here is incorrect.

The condition should not OR isAdvanced at the top level. Consider removing it entirely and relying on the event-driven cleanup, or intersecting it with a handle mismatch to avoid unconditional removal.

locked property lacks explicit fallback, unlike other optional fields

At line 2024, locked: flow?.locked returns undefined when flow is undefined, but the type expects boolean | null. Other fields in the same object use explicit fallbacks (icon: flow?.icon ?? undefined, is_component: flow?.is_component ?? false). For consistency, use locked: flow?.locked ?? false or ?? undefined.

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@src/frontend/src/utils/reactflowUtils.ts` at line 158, The cleanEdges removal
predicate is incorrectly OR-ing isAdvanced (computed from
targetNode.data.node!.template[field]?.advanced) which causes all edges to
advanced fields to be removed unconditionally; update the predicate inside
cleanEdges (the block using scapedJSONStringfy(id) !== targetHandle,
targetNode.data.node?.tool_mode, isToolMode, isLoopInput) to remove isAdvanced
from the top-level OR and either delete it entirely or only apply it when the
handle mismatches (e.g., combine with scapedJSONStringfy(id) !== targetHandle)
so advanced-field edges are not dropped on every flow update; also keep the
event-driven cleanup in clearHandlesFromAdvancedFields and rely on
filterHiddenFieldsEdges for show===false cases. Separately, make the returned
locked property explicit by changing locked: flow?.locked to locked:
flow?.locked ?? false (or ?? undefined) so it never returns undefined
unexpectedly.

[coderabbitai]

⚠️ Potential issue | 🟠 Major

🧩 Analysis chain

🌐 Web query:

Google Gemini API available models 2025 2026 gemini-2.5-flash gemini-3-pro

💡 Result:

As of Feb 19, 2026, the official Gemini API model codes (Google AI for Developers) include:

• Gemini 2.5 Flash (stable): gemini-2.5-flash (latest update: June 2025) [1][2]
• Gemini 3 Pro (preview): gemini-3-pro-preview (latest update: Nov 2025) [1][3]
• Gemini 3.1 Pro (preview): gemini-3.1-pro-preview and gemini-3.1-pro-preview-customtools (latest update: Feb 2026) [1][4]
• Gemini 3 Flash (preview): gemini-3-flash-preview (latest update: Dec 2025) [1][5]

So, gemini-2.5-flash is available (stable), while “Gemini 3 Pro” is exposed in the API as gemini-3-pro-preview (preview)—not gemini-3-pro (without -preview) per the current model list. [1][3]

Sources
[1] https://ai.google.dev/gemini-api/docs/models
[2] https://ai.google.dev/gemini-api/docs/models/gemini-2.5-flash
[3] https://ai.google.dev/gemini-api/docs/models/gemini-3-pro-preview
[4] https://ai.google.dev/gemini-api/docs/models/gemini-3.1-pro-preview
[5] https://ai.google.dev/gemini-api/docs/models/gemini-3-flash-preview

---

Several model names lack verification in official Google API documentation, and latest Gemini 3.1 models are missing.

Confirmed via Google's official API (Feb 19, 2026):

• gemini-2.5-flash (stable) ✓
• gemini-3-pro-preview (preview) ✓
• gemini-3-flash-preview (preview) ✓

Cannot be verified in official Google API docs:

• gemini-2.5-pro, gemini-2.5-flash-lite, gemini-2.5-flash-preview-09-2025, gemini-2.5-flash-lite-preview-09-2025, gemini-2.5-flash-image, gemini-3-pro-image-preview

Missing from code (available since Feb 2026):

• gemini-3.1-pro-preview and gemini-3.1-pro-preview-customtools should be added as the latest preview models.

Clarify whether unverified model names are internal/region-specific variants or if they should be removed. Update preview model names to remove outdated September 2025 suffixes if they no longer reflect the actual release status.

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@src/lfx/src/lfx/base/models/google_generative_ai_constants.py` around lines 5
- 89, The list GOOGLE_GENERATIVE_AI_MODELS_DETAILED contains model entries
(created via create_model_metadata) that are either unverified against Google's
API (e.g., "gemini-2.5-pro", "gemini-2.5-flash-lite",
"gemini-2.5-flash-preview-09-2025", "gemini-2.5-flash-lite-preview-09-2025",
"gemini-2.5-flash-image", "gemini-3-pro-image-preview") and should be validated
or removed, and it is missing current Gemini 3.1 preview names; update the
constant by removing or flagging unverified/region-specific variants and add
entries for "gemini-3.1-pro-preview" and "gemini-3.1-pro-preview-customtools"
using create_model_metadata (mark preview=True), and rename/remove any outdated
"-09-2025" preview suffixes to match official model names.

[github-actions]

Build successful! ✅
Deploying docs draft.
Deploy successful! View draft

[Adam-Aghili]

LGTM