UPSTREAM PR #17670: Add a couple of file types to the text section

[loci-dev]

Mirrored from ggml-org/llama.cpp#17670

Fixes ggml-org/llama.cpp#17556

I believe we should at least support the source files in the llama.cpp codebase ;)

[loci-review]

Explore the complete analysis inside the Version Insights

Performance Analysis Summary - PR #396

Analysis Overview

This PR adds file type support for CUDA, Vulkan, C++, and other source files in the llama-server web UI. All changes are confined to TypeScript configuration files in the web interface layer. No modifications were made to C++ inference code, GGML backends, or any performance-critical components.

Performance Impact

No performance changes detected. The power consumption analysis confirms:

• libllama.so: -0.24 nJ change (negligible, within measurement noise)
• All GGML backend libraries: 0.0% change
• All utility binaries: 0.0% change

No functions in the inference pipeline (llama_decode, llama_encode, llama_tokenize) were modified. Tokens per second remains unaffected.

[loci-review]

Explore the complete analysis inside the Version Insights

Performance Analysis Summary - PR #396

Analysis Overview

This PR adds file type support for CUDA, Vulkan, C++, and other source files in the llama-server web UI. All changes are confined to TypeScript configuration files in the web interface layer. No modifications were made to C++ inference code, GGML backends, or any performance-critical components.

Performance Impact

No performance changes detected. The power consumption analysis confirms:

• libllama.so: -0.24 nJ change (negligible, within measurement noise)
• All GGML backend libraries: 0.0% change
• All utility binaries: 0.0% change

No functions in the inference pipeline (llama_decode, llama_encode, llama_tokenize) were modified. Tokens per second remains unaffected.

[loci-review]

Explore the complete analysis inside the Version Insights

Performance Analysis Summary - PR #396

Analysis Overview

This PR adds file type support for CUDA, Vulkan, C++, and other source files in the llama-server web UI. All changes are confined to TypeScript configuration files in the web interface layer. No modifications were made to C++ inference code, GGML backends, or any performance-critical components.

Performance Impact

No performance changes detected. The power consumption analysis confirms:

• libllama.so: -0.24 nJ change (negligible, within measurement noise)
• All GGML backend libraries: 0.0% change
• All utility binaries: 0.0% change

No functions in the inference pipeline (llama_decode, llama_encode, llama_tokenize) were modified. Tokens per second remains unaffected.

[loci-review]

Explore the complete analysis inside the Version Insights

Performance Analysis Summary: PR #396

Analysis Overview

This PR adds file type support for CUDA, Vulkan, C++ headers, and other source formats to the web UI. All changes are isolated to TypeScript files in the web UI layer with no modifications to C++ inference code.

Performance Assessment

Performance analysis between versions e3c6aab9-aac4-4cd0-b34b-fef6a3bfeace and 68981b48-fd8f-49da-bea5-28766cd7402d shows zero measurable impact across all binaries and functions. Power consumption remains stable at 1.48 mJ total across all binaries with maximum deviation of 0.35 nJ in llama-run, representing 0.0002% change within measurement precision limits. No functions exhibit response time or throughput changes. The modifications affect only web UI file type detection logic with no interaction with inference pipeline, tokenization, or model execution paths.