feat(example): support server video inputs and Gemma text tool calls (abetlen#2291)

* feat(example): support server video inputs and Gemma text tool calls

* fix(example): pass media count to MTMD tokenizer

* fix(example): avoid MTMD video stdin double close

* fix(example): parse Gemma text tool bracket calls

Author: abetlen

CHANGELOG.md

@@ -7,6 +7,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+- feat(example): support server video inputs and Gemma text tool calls by @abetlen in #2291
 - feat: update llama.cpp to ggml-org/llama.cpp@8f83d6c27
 - fix(example): support multi-step Responses tool streaming by @abetlen in #2288
 - fix(ci): Repair Linux accelerator wheels for manylinux publishing

examples/server/configs/gemma-4-12b-it-qat.json

@@ -44,7 +44,7 @@
         },
         "tool_calls": {
           "type": "array",
-          "x-regex-iterator": "<\\|tool_call>(call:[^\\{]+\\{.*?\\})<tool_call\\|>",
+          "x-regex-iterator": "<\\|tool_call>(call:[^\\{\\[]+[\\{\\[].*?[\\}\\]])<tool_call\\|>",
           "items": {
             "type": "object",
             "properties": {
@@ -56,11 +56,11 @@
                 "properties": {
                   "name": {
                     "type": "string",
-                    "x-regex": "^call:(\\w+)\\{"
+                    "x-regex": "^call:(\\w+)[\\{\\[]"
                   },
                   "arguments": {
                     "type": "object",
-                    "x-regex": "^call:\\w+(\\{.*\\})$",
+                    "x-regex": "^call:\\w+(\\{.*\\}|\\[.*\\])$",
                     "x-parser": "gemma4-tool-call",
                     "additionalProperties": true
                   }