From 9323f4b5caf63ba40d56c0a7dbdff60a919dcf22 Mon Sep 17 00:00:00 2001
From: "LocalAI [bot]" <139863280+localai-bot@users.noreply.github.com>
Date: Mon, 8 Jun 2026 23:17:50 +0200
Subject: [PATCH] feat(llama-cpp): video input support (mtmd #24269) (#10216)

* chore(llama-cpp): bump to 8f83d6c for mtmd video input support

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

* feat(llama-cpp): forward video input to mtmd (template + non-template paths)

Wire request->videos() into grpc-server.cpp mirroring the existing image
and audio handling: a video_data build + non-template files extraction, and
input_video chat chunks on the tokenizer-template path. allow_video is
auto-set at model load by the vendored upstream chat_params.

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

* feat(ui): add video attachment support to the chat UI

Mirror the image/audio attachment path for video: emit video_url content
parts, accept video/* in the picker, keep video files as base64, show a
film icon badge, and render attached video inline with a <video> player.

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

* fix(llama-cpp): patch mtmd video stdin double-close (heap crash)

Upstream mtmd video input (ggml-org/llama.cpp#24269) double-fcloses the
ffmpeg/ffprobe stdin FILE: feed_stdin() fclose()s the FILE returned by
subprocess_stdin() (which is sp->stdin_file), then subprocess_destroy()
fclose()s the same pointer again -> heap corruption that aborts the
backend on any base64 input_video request (the CLI --video file path is
unaffected). Vendor a one-line fix (null sp->stdin_file after fclose)
via prepare.sh's patches/ until upstream merges it.

Verified e2e with gemma-4-e2b-it-qat-q4_0: video frames decode via
ffmpeg and the model answers correctly (red clip -> 'Red', blue -> 'Blue').

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

* chore(llama-cpp): re-pin to upstream #24316, drop vendored stdin patch

Upstream replaced the ad-hoc video stdin handling with a proper RAII
refactor (ggml-org/llama.cpp#24316, "mtmd: refactor video subproc
handling"), which includes the same `sp->stdin_file = nullptr` guard our
patch added (plus join-before-destroy ordering). Re-pin LLAMA_VERSION to
that branch head and drop patches/0001 - it's now redundant.

Verified e2e with gemma-4-e2b-it-qat-q4_0: no crash, video frames decode
and the model answers correctly (red clip -> "Red", blue -> "Blue").

NOTE: #24316 is not yet merged, so this pins to its branch-head commit
(28ca1e60). Re-pin to the squash-merge commit on master once it lands,
otherwise `git fetch` may lose the commit after the branch is deleted.

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

---------

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Co-authored-by: Ettore Di Giacinto <mudler@localai.io>
---
 backend/cpp/llama-cpp/Makefile          |  2 +-
 backend/cpp/llama-cpp/grpc-server.cpp   | 73 ++++++++++++++++++++++++-
 core/http/react-ui/src/hooks/useChat.js |  6 ++
 core/http/react-ui/src/pages/Chat.jsx   | 11 ++--
 4 files changed, 85 insertions(+), 7 deletions(-)

diff --git a/backend/cpp/llama-cpp/Makefile b/backend/cpp/llama-cpp/Makefile
index a33f1ceaa..13f0de3ee 100644
--- a/backend/cpp/llama-cpp/Makefile
+++ b/backend/cpp/llama-cpp/Makefile
@@ -1,5 +1,5 @@
 
-LLAMA_VERSION?=9e3b928fd8c9d14dbf15a8768b9fdd7e5c721d66
+LLAMA_VERSION?=28ca1e600c5dac1854fb7e09611914013430b037
 LLAMA_REPO?=https://github.com/ggerganov/llama.cpp
 
 CMAKE_ARGS?=
diff --git a/backend/cpp/llama-cpp/grpc-server.cpp b/backend/cpp/llama-cpp/grpc-server.cpp
index 90a5477a9..9ec806cd1 100644
--- a/backend/cpp/llama-cpp/grpc-server.cpp
+++ b/backend/cpp/llama-cpp/grpc-server.cpp
@@ -381,6 +381,15 @@ json parse_options(bool streaming, const backend::PredictOptions* predict, const
             });
     }
 
+    // for each video in the request, add the video data
+    for (int i = 0; i < predict->videos_size(); i++) {
+        data["video_data"].push_back(json
+            {
+                {"id", i},
+                {"data",    predict->videos(i)},
+            });
+    }
+
     data["stop"] = predict->stopprompts();
     // data["n_probs"] = predict->nprobs();
     //TODO: images,
@@ -1503,7 +1512,7 @@ public:
                     msg_json["role"] = msg.role();
 
                     bool is_last_user_msg = (i == last_user_msg_idx);
-                    bool has_images_or_audio = (request->images_size() > 0 || request->audios_size() > 0);
+                    bool has_images_or_audio = (request->images_size() > 0 || request->audios_size() > 0 || request->videos_size() > 0);
 
                     // Handle content - can be string, null, or array
                     // For multimodal content, we'll embed images/audio from separate fields
@@ -1554,6 +1563,16 @@ public:
                                     content_array.push_back(audio_chunk);
                                 }
                             }
+                            if (request->videos_size() > 0) {
+                                for (int j = 0; j < request->videos_size(); j++) {
+                                    json video_chunk;
+                                    video_chunk["type"] = "input_video";
+                                    json input_video;
+                                    input_video["data"] = request->videos(j);
+                                    video_chunk["input_video"] = input_video;
+                                    content_array.push_back(video_chunk);
+                                }
+                            }
                             msg_json["content"] = content_array;
                         } else {
                             // Use content as-is (already array or not last user message)
@@ -1588,6 +1607,16 @@ public:
                                 content_array.push_back(audio_chunk);
                             }
                         }
+                        if (request->videos_size() > 0) {
+                            for (int j = 0; j < request->videos_size(); j++) {
+                                json video_chunk;
+                                video_chunk["type"] = "input_video";
+                                json input_video;
+                                input_video["data"] = request->videos(j);
+                                video_chunk["input_video"] = input_video;
+                                content_array.push_back(video_chunk);
+                            }
+                        }
                         msg_json["content"] = content_array;
                     } else if (msg.role() == "tool") {
                         // Tool role messages must have content field set, even if empty
@@ -2039,6 +2068,16 @@ public:
                         files.push_back(decoded_data);
                     }
                 }
+
+                const auto &video_data = data.find("video_data");
+                if (video_data != data.end() && video_data->is_array())
+                {
+                    for (const auto &video : *video_data)
+                    {
+                        auto decoded_data = base64_decode(video["data"].get<std::string>());
+                        files.push_back(decoded_data);
+                    }
+                }
             }
 
             const bool has_mtmd = ctx_server.impl->mctx != nullptr;
@@ -2291,7 +2330,7 @@ public:
                     }
 
                     bool is_last_user_msg = (i == last_user_msg_idx);
-                    bool has_images_or_audio = (request->images_size() > 0 || request->audios_size() > 0);
+                    bool has_images_or_audio = (request->images_size() > 0 || request->audios_size() > 0 || request->videos_size() > 0);
 
                     // Handle content - can be string, null, or array
                     // For multimodal content, we'll embed images/audio from separate fields
@@ -2344,6 +2383,16 @@ public:
                                     content_array.push_back(audio_chunk);
                                 }
                             }
+                            if (request->videos_size() > 0) {
+                                for (int j = 0; j < request->videos_size(); j++) {
+                                    json video_chunk;
+                                    video_chunk["type"] = "input_video";
+                                    json input_video;
+                                    input_video["data"] = request->videos(j);
+                                    video_chunk["input_video"] = input_video;
+                                    content_array.push_back(video_chunk);
+                                }
+                            }
                             msg_json["content"] = content_array;
                         } else {
                             // Use content as-is (already array or not last user message)
@@ -2383,6 +2432,16 @@ public:
                                 content_array.push_back(audio_chunk);
                             }
                         }
+                        if (request->videos_size() > 0) {
+                            for (int j = 0; j < request->videos_size(); j++) {
+                                json video_chunk;
+                                video_chunk["type"] = "input_video";
+                                json input_video;
+                                input_video["data"] = request->videos(j);
+                                video_chunk["input_video"] = input_video;
+                                content_array.push_back(video_chunk);
+                            }
+                        }
                         msg_json["content"] = content_array;
                         SRV_INF("[CONTENT DEBUG] Predict: Message %d created content array with media\n", i);
                     } else if (!msg.tool_calls().empty()) {
@@ -2845,6 +2904,16 @@ public:
                         files.push_back(decoded_data);
                     }
                 }
+
+                const auto &video_data = data.find("video_data");
+                if (video_data != data.end() && video_data->is_array())
+                {
+                    for (const auto &video : *video_data)
+                    {
+                        auto decoded_data = base64_decode(video["data"].get<std::string>());
+                        files.push_back(decoded_data);
+                    }
+                }
             }
 
             // process files
diff --git a/core/http/react-ui/src/hooks/useChat.js b/core/http/react-ui/src/hooks/useChat.js
index 539190b99..5bd491e77 100644
--- a/core/http/react-ui/src/hooks/useChat.js
+++ b/core/http/react-ui/src/hooks/useChat.js
@@ -216,6 +216,12 @@ export function useChat(initialModel = '') {
             audio_url: { url: `data:${file.type};base64,${file.base64}` },
           })
           userFiles.push({ name: file.name, type: 'audio' })
+        } else if (file.type?.startsWith('video/')) {
+          messageContent.push({
+            type: 'video_url',
+            video_url: { url: `data:${file.type};base64,${file.base64}` },
+          })
+          userFiles.push({ name: file.name, type: 'video' })
         } else {
 			// Text/PDF files - append to content
 			if (file.textContent) {
diff --git a/core/http/react-ui/src/pages/Chat.jsx b/core/http/react-ui/src/pages/Chat.jsx
index a638aa3a8..2c51fe942 100644
--- a/core/http/react-ui/src/pages/Chat.jsx
+++ b/core/http/react-ui/src/pages/Chat.jsx
@@ -265,7 +265,7 @@ function UserMessageContent({ content, files }) {
         <div className="chat-message-files">
           {files.map((f, i) => (
             <span key={i} className="chat-file-inline">
-              <i className={`fas ${f.type === 'image' ? 'fa-image' : f.type === 'audio' ? 'fa-headphones' : 'fa-file'}`} />
+              <i className={`fas ${f.type === 'image' ? 'fa-image' : f.type === 'audio' ? 'fa-headphones' : f.type === 'video' ? 'fa-film' : 'fa-file'}`} />
               {f.name}
             </span>
           ))}
@@ -274,6 +274,9 @@ function UserMessageContent({ content, files }) {
       {Array.isArray(content) && content.filter(c => c.type === 'image_url').map((img, i) => (
         <img key={i} src={img.image_url.url} alt="attached" className="chat-inline-image" />
       ))}
+      {Array.isArray(content) && content.filter(c => c.type === 'video_url').map((vid, i) => (
+        <video key={i} src={vid.video_url.url} controls className="chat-inline-video" />
+      ))}
     </>
   )
 }
@@ -711,7 +714,7 @@ export default function Chat() {
     for (const file of e.target.files) {
       const base64 = await fileToBase64(file)
       const entry = { name: file.name, type: file.type, base64 }
-      if (!file.type.startsWith('image/') && !file.type.startsWith('audio/')) {
+      if (!file.type.startsWith('image/') && !file.type.startsWith('audio/') && !file.type.startsWith('video/')) {
         entry.textContent = await file.text().catch(() => '')
       }
       newFiles.push(entry)
@@ -1244,7 +1247,7 @@ export default function Chat() {
           <div className="chat-files">
             {files.map((f, i) => (
               <span key={i} className="chat-file-badge">
-                <i className={`fas ${f.type?.startsWith('image/') ? 'fa-image' : f.type?.startsWith('audio/') ? 'fa-headphones' : 'fa-file'}`} />
+                <i className={`fas ${f.type?.startsWith('image/') ? 'fa-image' : f.type?.startsWith('audio/') ? 'fa-headphones' : f.type?.startsWith('video/') ? 'fa-film' : 'fa-file'}`} />
                 {f.name}
                 <button onClick={() => setFiles(prev => prev.filter((_, idx) => idx !== i))}>
                   <i className="fas fa-xmark" />
@@ -1343,7 +1346,7 @@ export default function Chat() {
               ref={fileInputRef}
               type="file"
               multiple
-              accept="image/*,audio/*,application/pdf,.txt,.md,.csv,.json"
+              accept="image/*,audio/*,video/*,application/pdf,.txt,.md,.csv,.json"
               style={{ display: 'none' }}
               onChange={handleFileChange}
             />